sessions and previous memories gone. permissions and tools also gone | Friends of the Crustacean 🦞🤝 | Page 1

rocky pumice Apr 27, 2026, 11:17 PM

#

When i started up today, all the sessions are gone. even memory or tools and permissions are gone

tiny kernelBOT Apr 27, 2026, 11:17 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

noble knotBOT Apr 27, 2026, 11:18 PM

#

First thing: please paste openclaw --version.

Also, don’t reinstall or rerun setup yet — if sessions/memory/tools all vanished together, it usually means OpenClaw started with a different state/config/workspace path, not necessarily that data was deleted.

Please run:

openclaw --version
openclaw status --all
openclaw gateway status
echo "OPENCLAW_STATE_DIR=$OPENCLAW_STATE_DIR"
echo "OPENCLAW_PROFILE=$OPENCLAW_PROFILE"
echo "OPENCLAW_CONFIG_PATH=$OPENCLAW_CONFIG_PATH"
find ~/.openclaw -maxdepth 4 -type f \( -name sessions.json -o -name MEMORY.md -o -name openclaw.json -o -name auth-profiles.json \) -print

Normally:

#

sessions live at ~/.openclaw/agents/<agentId>/sessions/
memory files live in the agent workspace, usually ~/.openclaw/workspace/
tools/permissions come from config + agent/tool policy

If all three disappeared, I’d first suspect: different Linux/macOS user, different OPENCLAW_STATE_DIR, different OPENCLAW_PROFILE, Docker/Podman volume not mounted, or a split install/path issue after update.

Docs:
https://docs.openclaw.ai/help/faq
https://docs.openclaw.ai/concepts/session
https://docs.openclaw.ai/concepts/memory

rocky pumice Apr 28, 2026, 8:02 AM

#

test@test-VirtualBox:~$ openclaw --version
openclaw status --all
openclaw gateway status
echo "OPENCLAW_STATE_DIR=$OPENCLAW_STATE_DIR"
echo "OPENCLAW_PROFILE=$OPENCLAW_PROFILE"
echo "OPENCLAW_CONFIG_PATH=$OPENCLAW_CONFIG_PATH"
find ~/.openclaw -maxdepth 4 -type f ( -name sessions.json -o -name MEMORY.md -o -name openclaw.json -o -name auth-profiles.json ) -print
OpenClaw 2026.4.23 (a979721)
[1]+ Exit 127 nohup /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh > /dev/null 2>&1

🦞 OpenClaw 2026.4.23 (a979721)
I've read more man pages than any human should—so you don't have to.

│
OpenClaw status --all

Overview
┌─────────────────┬────────────────────────────────────────────────────────────┐
│ Item │ Value │
├─────────────────┼────────────────────────────────────────────────────────────┤
│ Version │ 2026.4.23 │
│ OS │ linux 6.17.0-22-generic (x64) │
│ Node │ 24.14.1 │
│ Config │ ~/.openclaw/openclaw.json │
│ Dashboard │ http://127.0.0.1:18789/ │
│ Tailscale │ off │
│ Channel │ stable (default) │
│ Update │ pnpm · npm update 2026.4.26 │
│ Gateway │ local · ws://127.0.0.1:18789 (local loopback) · │

#

│ │ unreachable (connect ECONNREFUSED 127.0.0.1:18789) │
│ Security │ Run: openclaw security audit --deep │
│ Gateway self │ unknown │
│ Gateway service │ systemd installed · enabled · stopped (state activating) │
│ Node service │ systemd not installed │
│ Agents │ 1 total · 0 bootstrapping · 0 active · 2 sessions │
│ Secrets │ none │
└─────────────────┴────────────────────────────────────────────────────────────┘

Channels
┌──────────┬─────────┬────────┬────────────────────────────────────────────────┐
│ Channel │ Enabled │ State │ Detail │
├──────────┼─────────┼────────┼────────────────────────────────────────────────┤
└──────────┴─────────┴────────┴────────────────────────────────────────────────┘

Agents
┌───────────────────────────┬────────────────┬──────────┬──────────┬───────────┐
│ Agent │ Bootstrap file │ Sessions │ Active │ Store │
├───────────────────────────┼────────────────┼──────────┼──────────┼───────────┤
│ main (Research Assistant) │ ABSENT │ 2 │ 9h ago │ ~/. │
│ │ │ │ │ openclaw/ │
│ │ │ │ │ agents/ │
│ │ │ │ │ main/ │
│ │ │ │ │ sessions/ │
│ │ │ │ │ sessions. │
│ │ │ │ │ json │
└───────────────────────────┴────────────────┴──────────┴──────────┴───────────┘

#

Diagnosis (read-only)

Gateway connection details:
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/test/.openclaw/openclaw.json
Bind: loopback

✓ Config: /home/test/.openclaw/openclaw.json
✓ Secret diagnostics (0)
✓ Restart sentinel: none
✓ Port 18789
! Tailscale: off · unknown
✓ Skills: 10 eligible · 0 missing · /home/test/.openclaw/workspace
✓ Plugin compatibility (none)
! Channel issues skipped (gateway unreachable)

Gateway health:
connect ECONNREFUSED 127.0.0.1:18789

Pasteable debug report. Auth tokens redacted.
Troubleshooting: https://docs.openclaw.ai/troubleshooting

◇

🦞 OpenClaw 2026.4.23 (a979721)
I've seen your commit messages. We'll work on that together.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Command: /usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/

#

Probe note: Loopback-only gateway; only local clients can connect.

Runtime: stopped (state failed, sub failed, last exit 0, reason 0)
Connectivity probe: failed
Probe target: ws://127.0.0.1:18789
connect ECONNREFUSED 127.0.0.1:18789
Capability: unknown

Service is loaded but not running (likely exited immediately).
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Logs: journalctl --user -u openclaw-gateway.service -n 200 --no-pager
Restart attempts: /home/test/.openclaw/logs/gateway-restart.log

Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting
OPENCLAW_STATE_DIR=
OPENCLAW_PROFILE=
OPENCLAW_CONFIG_PATH=
/home/test/.openclaw/openclaw.json
/home/test/.openclaw/agents/research-assistant/agent/auth-profiles.json
/home/test/.openclaw/agents/research-assistant/sessions/sessions.json
/home/test/.openclaw/agents/main/agent/auth-profiles.json
/home/test/.openclaw/agents/main/sessions/sessions.json
test@test-VirtualBox:~$

noble knotBOT Apr 28, 2026, 8:03 AM

#

That output is cut off right before the important part — especially the find results.

Most likely this is not deleted data, but OpenClaw starting with a different state dir / profile / workspace. Sessions live under:

~/.openclaw/agents/<agentId>/sessions/sessions.json

Memory lives in the workspace, usually:

~/.openclaw/workspace/MEMORY.md

Docs: https://docs.openclaw.ai/help/faq#where-things-live-on-disk

Please don’t run openclaw reset.

Can you paste the rest, plus this?

#

openclaw config file
openclaw config get agents.defaults.workspace
openclaw config get tools --json
openclaw config get agents.list --json
openclaw sessions --all-agents
find ~ -maxdepth 1 -type d -name '.openclaw*' -print
find ~/.openclaw ~/.openclaw-* -maxdepth 5 -type f \( -name sessions.json -o -name MEMORY.md -o -name openclaw.json -o -name auth-profiles.json \) -print 2>/dev/null
openclaw doctor

Also: you’re on 2026.4.23 and status says 2026.4.26 is available, so after we confirm where your files are, I’d update + doctor/restart. But first let’s locate the old state so we don’t accidentally “fix” the wrong empty profile. 🦐

#

Good, this narrows it down: your Gateway is not reachable, not necessarily “everything deleted”.

Key line:

unreachable (connect ECONNREFUSED 127.0.0.1:18789)
Gateway service: systemd installed · enabled · stopped (state activating)

So the UI/client may look empty because the Gateway isn’t actually listening. Also, OpenClaw still sees:

Agents: 1 total · ... · 2 sessions
Store: ~/.openclaw/agents/main/sessions/sessions.json

Next, please run/paste:

#

openclaw gateway status --deep
openclaw logs --lines 120
openclaw config validate
openclaw doctor

If openclaw logs can’t connect, use systemd logs:

journalctl --user -u openclaw-gateway -n 160 --no-pager

Do not reset yet. We need the startup error first — likely config/service/startup failure, and your sessions file still exists. Docs: https://docs.openclaw.ai/gateway/troubleshooting

#

Yep — this still points to Gateway startup failure, not wiped sessions.

The important bits:

File logs: /tmp/openclaw/openclaw-2026-04-28.log
Gateway health: connect ECONNREFUSED
Service: enabled, but not listening

Please paste the startup error from:

tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
journalctl --user -u openclaw-gateway -n 200 --no-pager
systemctl --user status openclaw-gateway --no-pager

#

Before running repairs, make a local backup:

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)

Then, if the logs don’t show an obvious config error, the next safe repair path is:

openclaw update
openclaw doctor --fix
openclaw gateway restart

But paste the logs first if you can — the exact crash line will tell us why the gateway never binds to 127.0.0.1:18789.

Docs: https://docs.openclaw.ai/gateway/troubleshooting

#

Aha — this confirms two useful things:

No profile/env mismatch: those env vars are empty, so you’re on the normal ~/.openclaw.

Your session stores still exist:

~/.openclaw/agents/research-assistant/sessions/sessions.json
~/.openclaw/agents/main/sessions/sessions.json

So don’t reset. The current blocker is still: Gateway exits immediately and never binds to port 18789.

Please paste this next — especially the actual error lines:

tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail -n 200 ~/.openclaw/logs/gateway-restart.log

#

journalctl --user -u openclaw-gateway.service -n 200 --no-pager
openclaw config validate
openclaw sessions --all-agents
openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f \( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print

Also: the research-assistant vs main split may explain the “old sessions missing” part. Your old data might be under research-assistant, while the current config/UI is looking at main. But first we need the Gateway crash line. 🦐

rocky pumice Apr 28, 2026, 8:06 AM

#

test@test-VirtualBox:~$ tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail: cannot open '/tmp/openclaw/openclaw-2026-04-28.log' for reading: No such file or directory
test@test-VirtualBox:~$ journalctl --user -u openclaw-gateway.service -n 200 --no-pager
openclaw config validate
openclaw sessions --all-agents
openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f ( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md ) -print
Apr 28 08:57:08 test-VirtualBox systemd[1267]: openclaw-gateway.service: Scheduled restart job, restart counter is at 4.
Apr 28 08:57:08 test-VirtualBox systemd[1267]: Starting openclaw-gateway.service - OpenClaw Gateway (v2026.4.23)...
Apr 28 08:57:08 test-VirtualBox xhost[4711]: Invalid MIT-MAGIC-COOKIE-1 key
Apr 28 08:57:08 test-VirtualBox xhost[4711]: /usr/bin/xhost: unable to open display ":0"
Apr 28 08:57:08 test-VirtualBox systemd[1267]: openclaw-gateway.service: Control process exited, code=exited, status=1/FAILURE
Apr 28 08:57:08 test-VirtualBox systemd[1267]: openclaw-gateway.service: Failed with result 'exit-code'.
Apr 28 08:57:08 test-VirtualBox systemd[1267]: Failed to start openclaw-gateway.service - OpenClaw Gateway (v2026.4.23).
Apr 28 08:57:14 test-VirtualBox systemd[1267]: openclaw-gateway.service: Scheduled restart job, restart counter is at 5.
Apr 28 08:57:14 test-VirtualBox systemd[1267]: openclaw-g

#

ateway.service: Start request repeated too quickly.
Apr 28 08:57:14 test-VirtualBox systemd[1267]: openclaw-gateway.service: Failed with result 'exit-code'.
Apr 28 08:57:14 test-VirtualBox systemd[1267]: Failed to start openclaw-gateway.service - OpenClaw Gateway (v2026.4.23).
Apr 28 09:12:26 test-VirtualBox systemd[1267]: Starting openclaw-gateway.service - OpenClaw Gateway (v2026.4.23)...
Apr 28 09:12:26 test-VirtualBox xhost[5083]: Invalid MIT-MAGIC-COOKIE-1 key
Apr 28 09:12:26 test-VirtualBox xhost[5083]: /usr/bin/xhost: unable to open display ":0"
Apr 28 09:12:26 test-VirtualBox systemd[1267]: openclaw-gateway.service: Control process exited, code=exited, status=1/FAILURE

noble knotBOT Apr 28, 2026, 8:06 AM

#

Smoking gun found:

xhost: Invalid MIT-MAGIC-COOKIE-1 key
/usr/bin/xhost: unable to open display ":0"
Control process exited ... status=1/FAILURE

That means systemd is failing in a pre-start/control step before OpenClaw even starts. So the Gateway never binds to 18789, and the UI looks empty/broken. Your sessions are still on disk.

Please run:

systemctl --user cat openclaw-gateway.service
systemctl --user show openclaw-gateway.service -p ExecStartPre -p ExecStart

#

If you see an ExecStartPre=/usr/bin/xhost ... line, remove/regenerate the service:

cp ~/.config/systemd/user/openclaw-gateway.service ~/openclaw-gateway.service.backup
openclaw gateway install --force
systemctl --user daemon-reload
openclaw gateway restart
openclaw gateway status

After it starts, check the “missing” data with:

openclaw sessions --all-agents
openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f \( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print

#

Also note you have both research-assistant and main agent session stores, so the old sessions may simply be under the other agent. First get the gateway booting again, then we can fix the agent/workspace selection if needed. 🦐

Docs: https://docs.openclaw.ai/gateway/troubleshooting

#

Yep, same root cause. The Gateway isn’t crashing from OpenClaw config — systemd is failing on xhost before Gateway starts.

Run this first so we can see where the bad pre-start comes from:

systemctl --user cat openclaw-gateway.service
systemctl --user show openclaw-gateway.service -p ExecStartPre -p ExecStart

Likely fix:

cp ~/.config/systemd/user/openclaw-gateway.service ~/openclaw-gateway.service.backup
openclaw gateway install --force
systemctl --user daemon-reload
systemctl --user reset-failed openclaw-gateway.service

#

openclaw gateway restart
openclaw gateway status

If systemctl --user cat shows the xhost line in a drop-in override under openclaw-gateway.service.d/, install --force may not remove it. In that case paste the systemctl --user cat output and we’ll remove just that override safely.

Still: don’t reset. Your session files exist. This is a service startup issue.

rocky pumice Apr 28, 2026, 8:09 AM

#

test@test-VirtualBox:~$ systemctl --user cat openclaw-gateway.service
systemctl --user show openclaw-gateway.service -p ExecStartPre -p ExecStart

[Unit]
Description=OpenClaw Gateway (v2026.4.23)
After=network-online.target
Wants=network-online.target
StartLimitBurst=5
StartLimitIntervalSec=60

[Service]
ExecStart=/usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Restart=always
RestartSec=5
RestartPreventExitStatus=78
TimeoutStopSec=30
TimeoutStartSec=30
SuccessExitStatus=0 143
KillMode=control-group
Environment=HOME=/home/test
Environment=TMPDIR=/tmp
Environment=NODE_EXTRA_CA_CERTS=/etc/ssl/certs/ca-certificates.crt
Environment=PATH=/usr/bin:/home/test/.nvm/current/bin:/home/test/.local/bin:/home/test/.npm-global/bin:/home/test/bin:/hom>
Environment=OPENCLAW_GATEWAY_PORT=18789
Environment=OPENCLAW_SYSTEMD_UNIT=openclaw-gateway.service
Environment="OPENCLAW_WINDOWS_TASK_NAME=OpenClaw Gateway"
Environment=OPENCLAW_SERVICE_MARKER=openclaw
Environment=OPENCLAW_SERVICE_KIND=gateway
Environment=OPENCLAW_SERVICE_VERSION=2026.4.23
Environment=DISPLAY=:0
Environment=XAUTHORITY=/home/test/.Xauthority
ExecStartPre=/usr/bin/xhost +local:test

[Install]
WantedBy=default.target

lines 1-35

noble knotBOT Apr 28, 2026, 8:09 AM

#

Found it. This line is the breaker:

ExecStartPre=/usr/bin/xhost +local:test

ExecStartPre must succeed or systemd refuses to start OpenClaw. Your xhost command is failing, so the Gateway never launches.

Use a systemd override to clear just that pre-start line:

cp ~/.config/systemd/user/openclaw-gateway.service ~/openclaw-gateway.service.backup

mkdir -p ~/.config/systemd/user/openclaw-gateway.service.d

cat > ~/.config/systemd/user/openclaw-gateway.service.d/no-xhost.conf <<'EOF'

#

[Service]
ExecStartPre=
EOF

systemctl --user daemon-reload
systemctl --user reset-failed openclaw-gateway.service
systemctl --user restart openclaw-gateway.service

openclaw gateway status

If that starts cleanly, then check the “missing” stuff:

openclaw sessions --all-agents

#

openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f \( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print

The data still looks present; the gateway was just being killed before startup.

rocky pumice Apr 28, 2026, 8:10 AM

#

test@test-VirtualBox:~$ cp ~/.config/systemd/user/openclaw-gateway.service ~/openclaw-gateway.service.backup

mkdir -p ~/.config/systemd/user/openclaw-gateway.service.d

cat > ~/.config/systemd/user/openclaw-gateway.service.d/no-xhost.conf <<'EOF'

[Service]
ExecStartPre=
EOF

systemctl --user daemon-reload
systemctl --user reset-failed openclaw-gateway.service
systemctl --user restart openclaw-gateway.service

openclaw gateway status

🦞 OpenClaw 2026.4.23 (a979721) — I speak fluent bash, mild sarcasm, and aggressive tab-completion energy.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Command: /usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 3145, state active, sub running, last exit 0, reason 0)
Warm-up: launch agents can take a few seconds. Try again shortly.
Connectivity probe: failed
Probe target: ws://127.0.0.1:18789
connect ECONNREFUSED 127.0.0.1:18789
Capability: unknown

Gateway port 18789 is not listening (service appears running).
Logs: journalctl --user -u openclaw-gateway.service -n 200 --no-pager
Restart log: ~/.openclaw/logs/gateway-restart.log

Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

noble knotBOT Apr 28, 2026, 8:11 AM

#

Nice, that fixed the systemd pre-start failure. Now we’re past xhost; OpenClaw’s Node process is running, but it still hasn’t opened port 18789.

Give it ~10 seconds, then run:

sleep 10
openclaw gateway status
ss -ltnp | grep 18789 || true
ps -fp 3145
tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail -n 200 ~/.openclaw/logs/gateway-restart.log
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

Paste the log output. We need to see whether the gateway is hanging during startup or crashing/restarting silently.

Good news: this is still not pointing to deleted sessions/memory. We’ve moved from “service can’t start” to “service process starts but gateway doesn’t listen yet.” 🦐

#

Docs: https://docs.openclaw.ai/gateway/troubleshooting

rocky pumice Apr 28, 2026, 8:17 AM

#

sleep 10
openclaw gateway status
ss -ltnp | grep 18789 || true
ps -fp 3145
tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail -n 200 ~/.openclaw/logs/gateway-restart.log
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

noble knotBOT Apr 28, 2026, 8:17 AM

#

Looks like only the commands came through, not the output.

Please run them and paste what prints after:

sleep 10
openclaw gateway status
ss -ltnp | grep 18789 || true
ps -fp 3145 || true
tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log 2>&1
tail -n 200 ~/.openclaw/logs/gateway-restart.log 2>&1
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

If Discord is eating/truncating it, save it to a file:

#

{
  sleep 10
  openclaw gateway status
  ss -ltnp | grep 18789 || true
  ps -fp 3145 || true
  tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log 2>&1
  tail -n 200 ~/.openclaw/logs/gateway-restart.log 2>&1
  journalctl --user -u openclaw-gateway.service -n 200 --no-pager
} > ~/openclaw-debug.txt 2>&1
cat ~/openclaw-debug.txt

We’re looking for the first real gateway startup error now.

rocky pumice Apr 28, 2026, 8:17 AM

#

test@test-VirtualBox:~$ sleep 10
openclaw gateway status
ss -ltnp | grep 18789 || true
ps -fp 3145
tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail -n 200 ~/.openclaw/logs/gateway-restart.log
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

🦞 OpenClaw 2026.4.23 (a979721) — Like having a senior engineer on call, except I don't bill hourly or sigh audibly.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Command: /usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 3145, state active, sub running, last exit 0, reason 0)
Connectivity probe: ok
Capability: admin-capable

Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting
LISTEN 0 511 127.0.0.1:18789 0.0.0.0:* users:(("openclaw-gatewa",pid=3145,fd=29))
LISTEN 0 511 [::1]:18789 [::]:* users:(("openclaw-gatewa",pid=3145,fd=30))
UID PID PPID C STIME TTY TIME CMD
test 3145 1045 7 18:10 ? 00:00:28 openclaw-gateway
{"0":"\u001b[38;2;47;191;113mConfig valid: ~/.openclaw/openclaw.json\u001b[39m","_meta":{"runtime":"node","runtimeVersion":"24.14.1","hostname":"unknown","name":"openclaw","date":"2026-04-28T08:05:32.241Z","logLevelId":3,"logLevelName":"INFO","path":{"fullFilePath":"file:///home/test/.npm-global/lib/node_modules/openclaw/dist/subsystem-bmJ5rL4Q.js:160:46","fileName":"subsystem-bmJ5rL4Q.js","fileNameWithLine":"subsystem-

noble knotBOT Apr 28, 2026, 8:18 AM

#

Great — Gateway is healthy now 🎉

Connectivity probe: ok
Listening: 127.0.0.1:18789

Now check/restore the “missing” sessions + memory:

openclaw sessions --all-agents
openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f \( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print
openclaw status --all

Likely next issue: you have two agent stores:

#

~/.openclaw/agents/research-assistant/...
~/.openclaw/agents/main/...

So if the UI is now pointed at main but your older work was in research-assistant, it’ll look like sessions/memory vanished even though the files are still there.

Also keep the no-xhost.conf override for now. If openclaw gateway install --force re-adds the broken xhost line later, that’s worth filing as a GitHub issue with your service file + the xhost error: https://github.com/openclaw/openclaw/issues

rocky pumice Apr 28, 2026, 8:19 AM

#

test@test-VirtualBox:~$ openclaw sessions --all-agents
openclaw config get agents.defaults.workspace
find ~/.openclaw/workspace -maxdepth 3 -type f ( -name MEMORY.md -o -name AGENTS.md -o -name USER.md -o -name SOUL.md ) -print
openclaw status --all

🦞 OpenClaw 2026.4.23 (a979721) — Powered by open source, sustained by spite and good documentation.

Session stores: 1 (main)
Sessions listed: 2
Agent Kind Key Age Model Tokens (ctx %) Flags
main direct agent:main:tui-2...4fa1c8 9h ago gemini-3-flash-preview 14k/200k (7%) system id:0f314b7f-1e54-4f98-8b81-9634866882a3
main direct agent:main:main 9h ago gemini-3-flash-preview 167k/200k (83%) think:medium system id:98ec73a5-a3b5-4af7-b05a-a6f6ea9d8c24

🦞 OpenClaw 2026.4.23 (a979721) — The only open-source project where the mascot could eat the competition.

/home/test/.openclaw/workspace
/home/test/.openclaw/workspace/SOUL.md
/home/test/.openclaw/workspace/AGENTS.md
/home/test/.openclaw/workspace/USER.md

🦞 OpenClaw 2026.4.23 (a979721) — Your terminal just grew claws—type something and let the bot pinch the busywork.

│
OpenClaw status --all

Overview

│ Item │ Value │

│ Version │ 2026.4.23 │
│ OS │ linux 6.17.0-22-generic (x64) │

#

│ Node │ 24.14.1 │
│ Config │ ~/.openclaw/openclaw.json │
│ Dashboard │ http://127.0.0.1:18789/ │
│ Tailscale │ off │
│ Channel │ stable (default) │
│ Update │ pnpm · npm update 2026.4.26 │
│ Gateway │ local · ws://127.0.0.1:18789 (local loopback) · reachable 64ms · auth token · test-VirtualBox (10.0. │
│ │ 2.15) app 2026.4.23 linux 6.17.0-22-generic │
│ Security │ Run: openclaw security audit --deep │
│ Gateway self │ unknown │
│ Gateway service │ systemd installed · enabled · running (pid 3145, state active) │
│ Node service │ systemd not installed │
│ Agents │ 1 total · 0 bootstrapping · 0 active · 2 sessions │
│ Secrets │ none

#

Channels

│ Channel │ Enabled │ State │ Detail

Agents

│ Agent │ Bootstrap file │ Sessions │ Active │ Store │

│ main (Research Assistant) │ ABSENT │ 2 │ 9h ago │ ~/.openclaw/agents/main/sessions/sessions.json │

Diagnosis (read-only)

Gateway connection details:
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/test/.openclaw/openclaw.json
Bind: loopback

✓ Config: /home/test/.openclaw/openclaw.json
✓ Secret diagnostics (0)
✓ Restart sentinel: none
! Port 18789
Port 18789 is already in use.

pid 3145 test: openclaw-gateway (127.0.0.1:18789)
Gateway already running locally. Stop it (openclaw gateway stop) or use a different port.
! Tailscale: off · unknown
✓ Skills: 10 eligible · 0 missing · /home/test/.openclaw/workspace
✓ Plugin compatibility (none)
✓ Channel issues (none)

Pasteable debug report. Auth tokens redacted.
Troubleshooting: https://docs.openclaw.ai/troubleshooting

◇
test@test-VirtualBox:~$

noble knotBOT Apr 28, 2026, 8:20 AM

#

Good: Gateway is fixed, and OpenClaw sees 2 sessions again.

Now the remaining issue is likely agent/workspace mismatch, plus no MEMORY.md in the current workspace.

Run these:

openclaw config get agents.list --json
openclaw config get tools --json

openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json

find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f \( -name MEMORY.md -o -name memory.md \) -print 2>/dev/null
find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f \( -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print 2>/dev/null

What we know:

#

Current configured workspace is:
```
~/.openclaw/workspace
```
It has:
```
SOUL.md
AGENTS.md
USER.md
```
It does not currently have MEMORY.md.
Earlier, disk showed a research-assistant agent store, but openclaw sessions --all-agents only lists configured main.

So if your old sessions/memory were under research-assistant, we need to either re-add/switch to that agent or copy the right workspace files back — but don’t move/delete anything until we see those command outputs.

#

Perfect — core service is healthy now:

Gateway reachable
Gateway service running
Agents: 1 total · 2 sessions

So the original “everything gone” was caused by the Gateway not starting because of the broken xhost pre-start line.

For the remaining “missing memory/tools/permissions” part, run the checks from above:

openclaw config get agents.list --json
openclaw config get tools --json

#

openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json

find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f \( -name MEMORY.md -o -name memory.md \) -print 2>/dev/null
find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f \( -name AGENTS.md -o -name USER.md -o -name SOUL.md \) -print 2>/dev/null

If that old research-assistant store has the missing sessions, we’ll point OpenClaw back at the right agent/store instead of guessing.

#

That “Port 18789 already in use” is okay here — it’s just the now-running Gateway:

pid 3145 test: openclaw-gateway
Gateway already running locally

Don’t stop it unless you’re restarting intentionally.

Current state:

Gateway fixed ✅
2 sessions visible ✅
Current agent is main, display name “Research Assistant”
Current workspace is ~/.openclaw/workspace
No channels configured
No MEMORY.md found in that workspace yet

#

To finish tracking the missing memory/tools, paste:

openclaw config get agents.list --json
openclaw config get tools --json
openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json
find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f \( -name MEMORY.md -o -name memory.md \) -print 2>/dev/null

If research-assistant/sessions.json has your old sessions, we’ll restore that agent/store selection. If no MEMORY.md exists anywhere, then memory wasn’t written to durable memory yet — only the session context existed.

rocky pumice Apr 28, 2026, 8:21 AM

#

test@test-VirtualBox:~$ openclaw config get agents.list --json
openclaw config get tools --json
openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json
find ~/.openclaw ~/.openclaw-* -maxdepth 6 -type f ( -name MEMORY.md -o -name memory.md ) -print 2>/dev/null
[
{
"id": "main",
"default": true,
"name": "Research Assistant",
"workspace": "/home/test/.openclaw/workspace",
"agentDir": "/home/test/.openclaw/agents/research-assistant/agent",
"model": "openrouter/google/gemini-3-flash-preview"
}
]
{
"profile": "minimal"
}

🦞 OpenClaw 2026.4.23 (a979721) — Hot reload for config, cold sweat for deploys.

Session store: /home/test/.openclaw/agents/research-assistant/sessions/sessions.json
Sessions listed: 1
Kind Key Age Model Tokens (ctx %) Flags
direct agent:research-a...t:main 30h ago gemini-3-flash-preview 6.6k/1049k (1%) system id:e4eb96ed-209a-43c1-a347-c1ff12690207
test@test-VirtualBox:~$

noble knotBOT Apr 28, 2026, 8:22 AM

#

Good, we found both remaining causes.

1) Tools/permissions “gone”

Your config says:

{ "profile": "minimal" }

minimal only exposes session_status, so that explains missing tools.

Restore a normal local tool set with:

openclaw config set tools.profile coding
openclaw config validate

#

openclaw gateway restart

If you intentionally want the broadest tool surface, use full instead of coding.

2) Old session exists, but under old agent id

Current agent id is main, but old session store is:

~/.openclaw/agents/research-assistant/sessions/sessions.json

and it has 1 older session from 30h ago.

If research-assistant was your original agent id and you want that back active:

#

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)

openclaw config set agents.list[0].id research-assistant
openclaw config validate
openclaw gateway restart
openclaw sessions --all-agents

3) Memory file

Your find returned no MEMORY.md, so there doesn’t seem to be a durable memory file under ~/.openclaw. Future durable memory needs to be written to:

~/.openclaw/workspace/MEMORY.md

Gateway is fixed; now it’s config cleanup. 🦐

rocky pumice Apr 28, 2026, 8:25 AM

#

test@test-VirtualBox:~$ ~/.openclaw/workspace/MEMORY.md
bash: /home/test/.openclaw/workspace/MEMORY.md: No such file or directory

noble knotBOT Apr 28, 2026, 8:25 AM

#

Right — that was a path, not a command. Bash tried to execute it.

To check it:

ls -la ~/.openclaw/workspace/
cat ~/.openclaw/workspace/MEMORY.md

But in your case it probably doesn’t exist yet. To create it:

cat > ~/.openclaw/workspace/MEMORY.md <<'EOF'
# MEMORY.md

## Notes

#

EOF

Then restart or just ask the agent to remember something durable.

The bigger fixes are still these:

openclaw config set tools.profile coding
openclaw config validate
openclaw gateway restart

And only if you want the old research-assistant session to become active again:

#

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)
openclaw config set agents.list[0].id research-assistant
openclaw config validate
openclaw gateway restart

rocky pumice Apr 28, 2026, 8:28 AM

#

it still has its memory as its latest session and not previous session

noble knotBOT Apr 28, 2026, 8:29 AM

#

Yeah — that means you’re seeing session context, not durable MEMORY.md.

OpenClaw has two different things:

Session history: previous chat transcript/context. It only appears when you resume that exact session key.
Durable memory: ~/.openclaw/workspace/MEMORY.md. You don’t have that file yet, so nothing durable was written there.

Your old session is here:

agent:research-assistant:main

To view/resume it in TUI:

#

openclaw
/session agent:research-assistant:main

If that session is the one with the “memory” you want, ask the agent inside that session:

Summarize the important durable facts from this session and write them to MEMORY.md.

Then check:

cat ~/.openclaw/workspace/MEMORY.md

#

For tools, you’re still on minimal mode. Fix that separately:

openclaw config set tools.profile coding
openclaw config validate
openclaw gateway restart

Docs: https://docs.openclaw.ai/help/faq#where-should-agentsmd--soulmd--usermd--memorymd-live

rocky pumice Apr 28, 2026, 8:36 AM

#

can i make it to be able to see and edit files outsoide of the shared workspace

noble knotBOT Apr 28, 2026, 8:36 AM

#

Yes — but be careful. There are two separate gates:

Tool access: your current config is tools.profile: "minimal", so file tools are mostly hidden.
Sandbox/workspace access: workspace is the default cwd, but if sandboxing is enabled, outside paths may be blocked unless mounted.

Start with this:

openclaw config set tools.profile coding
openclaw config validate
openclaw gateway restart

coding includes file tools: read, write, edit, apply_patch.

Then check whether you’re sandboxed:

#

openclaw sandbox explain

If sandboxing is off, the agent can usually use absolute paths like /home/test/projects/foo once file tools are allowed.

If sandboxing is on, bind the folder explicitly, preferably only the folder you need:

openclaw config set agents.defaults.sandbox.docker.binds '["/home/test/projects:/projects:rw"]' --strict-json
openclaw config validate
openclaw gateway restart

Use :ro instead of :rw if you only want read access.

Safer alternative: make the project folder the workspace:

#

openclaw config set agents.defaults.workspace /home/test/projects/my-project
openclaw setup --workspace /home/test/projects/my-project
openclaw gateway restart

Docs: https://docs.openclaw.ai/concepts/agent-workspace and https://docs.openclaw.ai/gateway/sandboxing

rocky pumice Apr 28, 2026, 8:37 AM

#

test@test-VirtualBox:~$ openclaw sandbox explain

🦞 OpenClaw 2026.4.23 (a979721) — Welcome to the command line: where dreams compile and confidence segfaults.

Effective sandbox:
agentId: research-assistant
sessionKey: agent:research-assistant:main
mainSessionKey: agent:research-assistant:main
runtime: direct
mode: off scope: agent
workspaceAccess: none workspaceRoot: /home/test/.openclaw/sandboxes

Sandbox tool policy:
allow (default): exec, process, read, write, edit, apply_patch, image, sessions_list, sessions_history, sessions_send, sessions_spawn, sessions_yield, subagents, session_status
deny (default): browser, canvas, nodes, cron, gateway, feishu, googlechat, nostr, msteams, mattermost, nextcloud-talk, matrix, bluebubbles, line, zalo, zalouser, synology-chat, tlon, discord, imessage, irc, qqbot, signal, slack, telegram, twitch, whatsapp

Elevated:
enabled: true
channel: webchat
allowedByConfig: false
failing gates: allowFrom (tools.elevated.allowFrom.webchat)

Fix-it:

tools.sandbox.tools.allow
tools.sandbox.tools.alsoAllow
tools.sandbox.tools.deny
agents.list[].tools.sandbox.tools.allow
agents.list[].tools.sandbox.tools.alsoAllow
agents.list[].tools.sandbox.tools.deny
tools.elevated.enabled
tools.elevated.allowFrom.webchat

Docs: docs.openclaw.ai/sandbox

test@test-VirtualBox:~$
on or off?

noble knotBOT Apr 28, 2026, 8:37 AM

#

It’s off.

This line is the key:

runtime: direct
mode: off

So you do not need sandbox binds. Once you switch tools from minimal to coding, the agent should be able to read/edit outside the workspace using absolute paths.

Run:

openclaw config set tools.profile coding
openclaw config validate

#

openclaw gateway restart

Then test by asking it something like:

Read /home/test/some-folder/some-file.txt

I’d avoid full unless you really want everything. coding is the safer normal choice.

rocky pumice Apr 28, 2026, 8:40 AM

#

whats the command to run full

noble knotBOT Apr 28, 2026, 8:40 AM

#

To switch tools to full:

openclaw config set tools.profile full
openclaw config validate
openclaw gateway restart

Then verify:

openclaw config get tools --json

Small caution: full is broad access. Since your sandbox is off, the agent can potentially read/edit outside the workspace with absolute paths.

rocky pumice Apr 28, 2026, 8:42 AM

#

is it recommendable for the agent to edit its own files to apply models or manual editing is safer

noble knotBOT Apr 28, 2026, 8:42 AM

#

Manual/CLI is safer for models/config.

Recommended order:

Best: use OpenClaw commands:

openclaw configure

or targeted:

openclaw config set agents.list[0].model "provider/model-id"
openclaw config validate
openclaw gateway restart

Okay: let the agent edit its own workspace files like:

#

   AGENTS.md
   MEMORY.md
   SOUL.md
   USER.md

Those are meant to be editable.

Riskier: let the agent directly edit ~/.openclaw/openclaw.json. If you do, make it use backups + validation:
```
cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)
openclaw config validate
```

For model choices specifically, I’d use openclaw configure or openclaw config set, not freehand JSON edits. Less chance of breaking syntax or config shape.

Docs: https://docs.openclaw.ai/cli/config and https://docs.openclaw.ai/concepts/models

rocky pumice Apr 28, 2026, 8:44 AM

#

anyway to select multiple models at once in configure or have to be infividual at once

noble knotBOT Apr 28, 2026, 8:44 AM

#

openclaw configure is mostly one interactive flow/pick at a time.

For multiple models at once, use openclaw config set with JSON.

Add multiple models to the /model allowlist:

openclaw config set agents.defaults.models '{
  "provider/model-a": {},
  "provider/model-b": {},
  "provider/model-c": {}
}' --strict-json --merge

Set primary + fallbacks in one shot:

#

openclaw config set agents.defaults.model '{
  "primary": "provider/model-a",
  "fallbacks": ["provider/model-b", "provider/model-c"]
}' --strict-json

Then:

openclaw config validate
openclaw gateway restart
openclaw models status

For just adding fallbacks one-by-one, there’s also:

#

openclaw models fallbacks add provider/model-b
openclaw models fallbacks list

Docs: https://docs.openclaw.ai/concepts/models

rocky pumice Apr 28, 2026, 8:47 AM

#

openrouter's deekseek V4 doesnt show on configure's model selection

noble knotBOT Apr 28, 2026, 8:48 AM

#

Probably because it’s new and your 2026.4.23 catalog/config picker doesn’t know it yet.

OpenRouter IDs are:

openrouter/deepseek/deepseek-v4-pro
openrouter/deepseek/deepseek-v4-flash

OpenRouter lists both V4 Pro and V4 Flash, and OpenClaw’s OpenRouter refs use openrouter/<provider>/<model>. (openrouter.ai)

Add them manually:

openclaw config set agents.defaults.models '{
  "openrouter/deepseek/deepseek-v4-pro": {},

DeepSeek V4 Pro - API Pricing & Providers

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. $0.435 per million input tokens, $0.87 per million output tokens. 1,048,576 token context window, maximum output of 384,000 tokens. Higher uptime with 5 providers.

#

  "openrouter/deepseek/deepseek-v4-flash": {}
}' --strict-json --merge

Set one as active for your current agent:

openclaw config set agents.list[0].model openrouter/deepseek/deepseek-v4-pro
openclaw config validate
openclaw gateway restart

Or use Flash:

#

openclaw config set agents.list[0].model openrouter/deepseek/deepseek-v4-flash

Also worth running later:

openclaw update

since your status showed 2026.4.26 available.

rocky pumice Apr 28, 2026, 8:50 AM

#

isnt there issues with openclaw update?

noble knotBOT Apr 28, 2026, 8:51 AM

#

Yeah, you’re not wrong to be cautious.

There have been release-specific update regressions recently — e.g. reports around 2026.4.5 CLI hangs, 2026.4.9 update completion-cache failure, and 2026.4.14 exec-tool result issues. (github.com)

For your case, I would not update just to add DeepSeek V4. Add it manually instead.

If you do update, do it safely:

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)

openclaw update --dry-run
openclaw update
openclaw doctor
openclaw gateway status

GitHub

CLI commands hang indefinitely after upgrade to 2026.4.5, cannot in...

Problem Description After upgrading to OpenClaw 2026.4.5, several CLI commands hang indefinitely without producing any output. The commands cannot be interrupted with Ctrl+C and must be forcefully ...

#

And because you had that broken xhost service issue, verify after update:

systemctl --user show openclaw-gateway.service -p ExecStartPre

Ideally ExecStartPre should be empty because of your override. If it shows xhost again, the gateway may fail the same way.

rocky pumice Apr 28, 2026, 8:55 AM

#

◇ Update ──────────────────────────────────────────────────────────────────────────────────╮
│ │
│ This install is not a git checkout. │
│ Run openclaw update to update via your package manager (npm/pnpm), then rerun doctor. │
│ │
├───────────────────────────────────────────────────────────────────────────────────────────╯
│
◆ Archive 2 orphan transcript files in ~/.openclaw/agents/research-assistant/sessions? This only renames them to
│ *.deleted.<timestamp>.
│ ○ Yes / ● No

noble knotBOT Apr 28, 2026, 8:56 AM

#

Choose No.

Those “orphan transcript files” are in the research-assistant session folder — exactly where we found your older session data. Even though it says “only renames,” don’t archive anything until you’ve confirmed what those files contain.

After exiting doctor, inspect first:

find ~/.openclaw/agents/research-assistant/sessions -maxdepth 2 -type f -print
openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json

If you want to be extra safe:

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)

#

Then we can decide what’s safe to clean.

rocky pumice Apr 28, 2026, 8:57 AM

#

Runtime: running (pid 7005, state active, sub running, last exit 0, reason 0)
Warm-up: launch agents can take a few seconds. Try again shortly.
Connectivity probe: failed
Probe target: ws://127.0.0.1:18789
timeout
Capability: unknown

Port 18789 is already in use.

pid 7005 test: openclaw-gateway (127.0.0.1:18789)
Gateway already running locally. Stop it (openclaw gateway stop) or use a different port.
Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting
test@test-VirtualBox:~$ systemctl --user show openclaw-gateway.service -p ExecStartPre
test@test-VirtualBox:~$ find ~/.openclaw/agents/research-assistant/sessions -maxdepth 2 -type f -print
openclaw sessions --store ~/.openclaw/agents/research-assistant/sessions/sessions.json
/home/test/.openclaw/agents/research-assistant/sessions/e4eb96ed-209a-43c1-a347-c1ff12690207.jsonl.reset.2026-04-28T08-27-45.671Z
/home/test/.openclaw/agents/research-assistant/sessions/e4eb96ed-209a-43c1-a347-c1ff12690207.trajectory-path.json
/home/test/.openclaw/agents/research-assistant/sessions/e4eb96ed-209a-43c1-a347-c1ff12690207.trajectory.jsonl
/home/test/.openclaw/agents/research-assistant/sessions/sessions.json
/home/test/.openclaw/agents/research-assistant/sessions/85a5534b-ca28-4bd6-99dd-354c5c962c00.trajectory.jsonl
/home/test/.openclaw/agents/research-assistant/sessions/85a5534b-ca28-4bd6-99dd-354c5c962c00.trajectory-path.json
/home/test/.openclaw/agents/research-assistant/sessions/85a5534b-ca28-4bd6-99dd-354c5c962c00.jsonl

🦞 OpenClaw 2026.4.26 (be8c246) — I don't sleep, I just enter low-power mode and dream of clean diffs.

Session store: /home/test/.openclaw/agents/research-assistant/sessions/sessions.json
Sessions listed: 1
Kind Key Age Model Tokens (ctx %) Flags
direct agent:research-a...t:main 16m ago gemini-3-flash-preview 14k/200k (7%) system id:85a5534b-ca28-4bd6-99dd-354c5c962c00

noble knotBOT Apr 28, 2026, 8:58 AM

#

Good news: update succeeded — you’re now on 2026.4.26.

The important bit is this file:

/home/test/.openclaw/agents/research-assistant/sessions/e4eb96ed-209a-43c1-a347-c1ff12690207.jsonl.reset.2026-04-28T08-27-45.671Z

That is your previous session transcript archive. It was reset/rotated, not deleted. The current active session is now:

85a5534b-ca28-4bd6-99dd-354c5c962c00.jsonl

So openclaw sessions only shows the new/current session, while the older context is sitting in the .jsonl.reset... file.

To preserve the useful “memory” from the previous session, ask the agent:

#

Read /home/test/.openclaw/agents/research-assistant/sessions/e4eb96ed-209a-43c1-a347-c1ff12690207.jsonl.reset.2026-04-28T08-27-45.671Z and summarize the durable facts into /home/test/.openclaw/workspace/MEMORY.md.

If file tools are still missing, set tools first:

openclaw config set tools.profile full
openclaw config validate
openclaw gateway restart

Also: if doctor asks to archive orphan transcripts again, pick No until you’re done extracting anything useful from that reset file.

rocky pumice Apr 28, 2026, 9:02 AM

#

Models in /model picker (multi-select)
│
│ Search: _
│ ◼ openrouter/google/gemini-3-flash-preview
│ ◼ openrouter/anthropic/claude-opus-4.6
│ ◼ openrouter/anthropic/claude-sonnet-4.6
│ ◼ openrouter/google/gemini-2.5-flash-lite
│ ◼ openrouter/anthropic/claude-3-haiku (configured)
│ ↑/↓ to navigate • Tab: select • Enter: confirm • Type: to search
└

only ehse show up in configure

noble knotBOT Apr 28, 2026, 9:03 AM

#

Yep — that picker is only showing your current configured allowlist.

Add DeepSeek V4 to the allowlist manually, then it should show up:

openclaw config set agents.defaults.models '{
  "openrouter/deepseek/deepseek-v4-pro": {},
  "openrouter/deepseek/deepseek-v4-flash": {}
}' --strict-json --merge

openclaw config validate
openclaw gateway restart

Then re-open configure / /model.

If you want to set one immediately:

#

openclaw config set agents.list[0].model openrouter/deepseek/deepseek-v4-pro
openclaw config validate
openclaw gateway restart

OpenRouter currently lists the model IDs as deepseek/deepseek-v4-pro and deepseek/deepseek-v4-flash; in OpenClaw, prefix those with openrouter/. (openrouter.ai)

DeepSeek V4 Pro - API Pricing & Providers

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. $0.435 per million input tokens, $0.87 per million output tokens. 1,048,576 token context window, maximum output of 384,000 tokens. Higher uptime with 5 providers.

rocky pumice Apr 28, 2026, 9:09 AM

#

tru again

[assistant turn failed before producing content]

tru again

[assistant turn failed before producing content]

tru again

noble knotBOT Apr 28, 2026, 9:10 AM

#

That message means the model request failed before any assistant text streamed back. Since your footer shows:

openrouter/google/gemini-3-flash-preview

it’s probably a provider/model/API error, not the TUI itself.

First grab the real error:

openclaw logs --lines 120

If that doesn’t show it:

#

tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

Quick workaround: switch to one of the configured Anthropic/OpenRouter models in the TUI:

/model

Pick openrouter/anthropic/claude-sonnet-4.6 or another configured model, then try again.

Also check auth/model resolution:

#

openclaw models status
openclaw config get agents.defaults.model --json
openclaw config get agents.list --json

Paste the log error line if it still fails — that’ll tell us whether it’s missing OpenRouter auth, model unavailable, context issue, or provider error.

rocky pumice Apr 28, 2026, 9:11 AM

#

est@test-VirtualBox:~$ openclaw logs --lines 120
error: unknown option '--lines'
[1]+ Exit 127 nohup /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh > /dev/null 2>&1
test@test-VirtualBox:~$ openclaw logs --lines 120
error: unknown option '--lines'
[1]+ Exit 127 nohup /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh > /dev/null 2>&1
test@test-VirtualBox:~$ ^[[200~tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log
tail: command not found
test@test-VirtualBox:~$ journalctl --user -u openclaw-gateway.service -n 200 --no-pager~
journalctl: unrecognized option '--no-pager~'
test@test-VirtualBox:~$

i have already tried 3 other models

noble knotBOT Apr 28, 2026, 9:11 AM

#

Yep — two things happened there:

Correct flag is --limit, not --lines.
Your paste inserted junk chars: `^

rocky pumice Apr 28, 2026, 9:12 AM

#

odel-fallback/decision {"subsystem":"model-fallback/decision"} {"event":"model_fallback_decision","tags":["error_handling","model_fallback","candidate_failed"],"runId":"73f3c034-abfb-482b-8513-c9bddca55fa4","decision":"candidate_failed","requestedProvider":"openrouter","requestedModel":"google/gemini-3-flash-preview","candidateProvider":"openrouter","candidateModel":"openrouter/auto","attempt":6,"total":6,"reason":"auth","status":401,"errorPreview":"401 User not found.","errorHash":"sha256:2cdeca109031","fallbackStepType":"fallback_step","fallbackStepFromModel":"openrouter/openrouter/auto","fallbackStepFromFailureReason":"auth","fallbackStepFromFailureDetail":"401 User not found.","fallbackStepChainPosition":6,"fallbackStepFinalOutcome":"chain_exhausted","isPrimary":false,"requestedModelMatched":false,"fallbackConfigured":true} model fallback decision
09:09:45+00:00 error Followup agent failed before reply: All models failed (6): openrouter/google/gemini-3-flash-preview: 401 User not found. (auth) | openrouter/anthropic/claude-opus-4.6: 401 User not found. (auth) | openrouter/anthropic/claude-sonnet-4.6: 401 User not found. (auth) | openrouter/google/gemini-2.5-flash-lite: 401 User not found. (auth) | openrouter/anthropic/claude-3-haiku: 401 User not found. (auth) | openrouter/openrouter/auto: 401 User not found. (auth) | HTTP 401: User not found.

noble knotBOT Apr 28, 2026, 9:13 AM

#

Yep, that log is decisive:

401 User not found. (auth)
All models failed...

So it’s not the model. Your OpenRouter auth/key is bad, revoked, from the wrong account, or OpenClaw is picking up a stale key.

Fix OpenRouter auth:

openclaw models status
openclaw models auth login --provider openrouter --set-default
openclaw gateway restart

#

If that command doesn’t offer API-key login, use:

openclaw models auth add

Pick OpenRouter and paste a fresh key from OpenRouter.

If it still says 401 User not found, you likely have an old key shadowing the new one. Check where OpenRouter auth exists, but don’t paste the key publicly:

find ~/.openclaw -type f \( -name auth-profiles.json -o -name auth.json -o -name openclaw.json -o -name .env \) -print

Then inspect/remove stale OpenRouter entries, especially old legacy files like:

#

~/.openclaw/agents/research-assistant/agent/auth.json
~/.openclaw/agents/main/agent/auth.json

After fixing auth:

openclaw models status
openclaw gateway restart

The [assistant turn failed before producing content] should disappear once OpenRouter auth is valid.

rocky pumice Apr 28, 2026, 9:16 AM

#

test@test-VirtualBox:~$ find ~/.openclaw -type f ( -name auth-profiles.json -o -name auth.json -o -name openclaw.json -o -name .env ) -print
/home/test/.openclaw/openclaw.json
/home/test/.openclaw/agents/research-assistant/agent/auth-profiles.json
/home/test/.openclaw/agents/main/agent/auth-profiles.json
test@test-VirtualBox:~$ ~/.openclaw/agents/research-assistant/agent/auth.json

so i go to hese files and paste my new api key in the old ones

noble knotBOT Apr 28, 2026, 9:16 AM

#

No — don’t paste keys directly into random files if you can avoid it.

Use the CLI so it writes the correct format to the correct agent auth store:

openclaw models auth --agent research-assistant login --provider openrouter --set-default

If that doesn’t prompt for an API key, use the interactive helper:

openclaw models auth --agent research-assistant add

Pick OpenRouter, paste the new key there.

Then verify:

#

openclaw models status --agent research-assistant
openclaw gateway restart

Also, this path does not exist:

~/.openclaw/agents/research-assistant/agent/auth.json

Your find only found:

~/.openclaw/agents/research-assistant/agent/auth-profiles.json
~/.openclaw/agents/main/agent/auth-profiles.json

#

So if anything is stale, it’s in auth-profiles.json, but let openclaw models auth ... update it safely.

rocky pumice Apr 28, 2026, 9:18 AM

#

test@test-VirtualBox:~$ openclaw models status --agent research-assistant
openclaw gateway restart

🦞 OpenClaw 2026.4.26 (be8c246) — I don't have opinions about tabs vs spaces. I have opinions about everything else.

Config : ~/.openclaw/openclaw.json
Agent dir : ~/.openclaw/agents/research-assistant/agent
Default (agent): openrouter/deepseek/deepseek-v4-pro
Fallbacks (5) (defaults): openrouter/google/gemini-3-flash-preview, openrouter/anthropic/claude-opus-4.6, openrouter/anthropic/claude-sonnet-4.6, openrouter/google/gemini-2.5-flash-lite, openrouter/anthropic/claude-3-haiku
Image model (defaults): -
Image fallbacks (0) (defaults): -
Aliases (1) : OpenRouter -> openrouter/auto
Configured models (8): openrouter/google/gemini-3-flash-preview, openrouter/anthropic/claude-opus-4.6, openrouter/anthropic/claude-sonnet-4.6, openrouter/google/gemini-2.5-flash-lite, openrouter/anthropic/claude-3-haiku, openrouter/deepseek/deepseek-v4-pro, openrouter/deepseek/deepseek-v4-flash, openrouter/auto

Auth overview
Auth store : ~/.openclaw/agents/research-assistant/agent/auth-profiles.json
Shell env : off
Providers w/ OAuth/tokens (0): -

openrouter effective=profiles:~/.openclaw/agents/main/agent/auth-profiles.json | profiles=1 (oauth=0, token=0, api_key=1) | openrouter:default=sk-or-v1.48 | models.json=sk-or-v1...ee61d | source=models.json: ~/.openclaw/agents/research-assistant/agent/models.json

OAuth/token status

none

🦞 OpenClaw 2026.4.26 (be8c246) — If it's repetitive, I'll automate it; if it's hard, I'll bring jokes and a rollback plan.

what does shell env mean, is it sandbox?

noble knotBOT Apr 28, 2026, 9:18 AM

#

No — Shell env is not sandbox.

Shell env: off means OpenClaw is not importing API keys from your terminal shell environment like:

OPENROUTER_API_KEY=...

So it’s using stored config/auth files instead. In your output, OpenRouter auth is coming from:

~/.openclaw/agents/main/agent/auth-profiles.json

and also:

#

~/.openclaw/agents/research-assistant/agent/models.json

Sandbox is separate. Yours was:

runtime: direct
mode: off

So sandbox is off.

One thing I notice: for research-assistant, the effective OpenRouter key is coming from the main agent’s auth-profiles.json, not the research-assistant auth store. That might be okay, but if you just updated the key for the wrong agent, update this one explicitly:

#

openclaw models auth --agent research-assistant add

Pick OpenRouter and paste the fresh key.

Then:

openclaw models status --agent research-assistant
openclaw gateway restart

rocky pumice Apr 28, 2026, 9:21 AM

#

test@test-VirtualBox:~$ openclaw models auth --agent research-assistant add

🦞 OpenClaw 2026.4.26 (be8c246) — I'll refactor your busywork like it owes me money.

│
◆ Token provider
│ ● anthropic (Docs: /providers/models)
│ ○ custom (type provider id)
└
what should i say for custom provider id

noble knotBOT Apr 28, 2026, 9:21 AM

#

Choose custom, then type:

openrouter

If it asks for profile/name, use:

default

Then paste your OpenRouter API key.

Afterward run:

#

openclaw models status --agent research-assistant
openclaw gateway restart

If it still prefers the old key from main, we’ll clear/reorder the OpenRouter auth source next.

rocky pumice Apr 28, 2026, 9:21 AM

#

it asks fr profile id

noble knotBOT Apr 28, 2026, 9:21 AM

#

Use:

default

That should create/update:

openrouter:default

Then paste the API key when prompted.

rocky pumice Apr 28, 2026, 9:44 AM

#

deepseek/deepseek-v4-flash is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: https://openrouter.ai/settings/integrations

429 Provider returned error
deepseek/deepseek-v4-flash is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: https://openrouter.ai/settings/integrations

429 Provider returned error
deepseek/deepseek-v4-flash is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: https://openrouter.ai/settings/integrations
connected | idle
agent research-assistant (Research Assistant) | session main (openclaw-tui) | openrouter/anthropic/claude-3-haiku | tokens 31k/200k (15%)

noble knotBOT Apr 28, 2026, 9:45 AM

#

Good — auth is fixed now. This is a different issue:

429 Provider returned error
deepseek-v4-flash is temporarily rate-limited upstream

That means OpenRouter can reach your account/key, but DeepSeek’s upstream route is rate-limited.

Options:

Use another configured model for now
```
/model
```
Pick openrouter/anthropic/claude-3-haiku or Sonnet/Opus.

#

Set DeepSeek Pro instead of Flash:

openclaw config set agents.list[0].model openrouter/deepseek/deepseek-v4-pro
openclaw config validate
openclaw gateway restart

Add your own DeepSeek provider key in OpenRouter integrations if you have one. That’s what the message is suggesting — it can give you separate provider-side quota.

Also, if you don’t want it repeatedly trying Flash as fallback, remove it from fallbacks or don’t select it as primary.

rocky pumice Apr 28, 2026, 9:45 AM

#

i have switched to claud 3 haiku but it is still saying deepseek

noble knotBOT Apr 28, 2026, 9:45 AM

#

That usually means DeepSeek is still in the fallback chain or pinned as the agent default, even if the TUI footer now shows Haiku.

Check:

openclaw models status --agent research-assistant
openclaw config get agents.list --json
openclaw config get agents.defaults.model --json

To force the agent default to Haiku and clear DeepSeek from the fallback path:

openclaw config set agents.list[0].model openrouter/anthropic/claude-3-haiku
openclaw config set agents.defaults.model '{
  "primary": "openrouter/anthropic/claude-3-haiku",

#

  "fallbacks": [
    "openrouter/anthropic/claude-sonnet-4.6",
    "openrouter/google/gemini-3-flash-preview"
  ]
}' --strict-json
openclaw config validate
openclaw gateway restart

Then in TUI, start a clean test session so old session-pinned model state doesn’t confuse things:

/session debug-haiku
/model openrouter/anthropic/claude-3-haiku

#

Then try a simple hi.

rocky pumice Apr 28, 2026, 9:48 AM

#

openclaw tui slow loading, is that a problem you can fix or is it just like that

noble knotBOT Apr 28, 2026, 9:48 AM

#

Some slowness is normal, but yours is probably fixable/tunable.

Your session is already ~31k tokens, and TUI loads history by default. Quick fixes:

openclaw --history-limit 50

or for a clean/fast session in TUI:

/session debug-fast

If the current session has lots of useful context but is getting heavy, compact it:

#

/compact

If you’re done with that topic, start fresh instead:

/new

Other things that made it slow today: the Gateway restart/update, OpenRouter 429 retries, and model fallback attempts. Once auth/rate-limit issues are settled, it should feel better.

Docs: https://docs.openclaw.ai/web/tui

rocky pumice Apr 28, 2026, 10:02 AM

#

if prompt reasoning goes for too long, how do i verify it is still running

noble knotBOT Apr 28, 2026, 10:02 AM

#

Use these checks:

openclaw status
openclaw gateway status
openclaw logs --follow --plain

In another terminal, you can also watch the process:

ps -fp $(pgrep -f openclaw-gateway | head -n1)

For TUI specifically:

Footer says running / streaming → still active

#

Footer says idle → no active turn
If stuck, press Esc / use abort in TUI, or run:
```
openclaw gateway restart
```

For logs without follow:

openclaw logs --limit 100 --plain

Look for model request/fallback lines, token streaming, or final errors like 429, 401, timeout, etc.

rocky pumice Apr 28, 2026, 10:31 AM

#

how to delete sonnet 4.6 as a fallback

noble knotBOT Apr 28, 2026, 10:32 AM

#

Run:

openclaw models fallbacks remove openrouter/anthropic/claude-sonnet-4.6
openclaw models fallbacks list
openclaw config validate
openclaw gateway restart

That removes it from the fallback chain.

If you also want to remove it from the /model picker allowlist, that’s separate. First check:

openclaw config get agents.defaults.models --json

#

But don’t remove it from the allowlist unless you’re sure — fallback removal is enough to stop auto-trying it.

rocky pumice Apr 28, 2026, 10:34 AM

#

how do i do multi-model add selecet without going through entire fongi setup where i have to manually add models and pai keyu

noble knotBOT Apr 28, 2026, 10:34 AM

#

Yep — skip the whole configure flow and just edit the model allowlist/fallbacks with config set.

Add multiple models to picker/allowlist

openclaw config set agents.defaults.models '{
  "openrouter/deepseek/deepseek-v4-pro": {},
  "openrouter/deepseek/deepseek-v4-flash": {},
  "openrouter/anthropic/claude-3-haiku": {},
  "openrouter/google/gemini-3-flash-preview": {}
}' --strict-json --merge

Set primary + fallbacks at once

#

openclaw config set agents.defaults.model '{
  "primary": "openrouter/anthropic/claude-3-haiku",
  "fallbacks": [
    "openrouter/google/gemini-3-flash-preview",
    "openrouter/deepseek/deepseek-v4-pro"
  ]
}' --strict-json

If your agent has its own model override, set that too

openclaw config set agents.list[0].model '{
  "primary": "openrouter/anthropic/claude-3-haiku",
  "fallbacks": [

#

    "openrouter/google/gemini-3-flash-preview",
    "openrouter/deepseek/deepseek-v4-pro"
  ]
}' --strict-json

Then:

openclaw config validate
openclaw gateway restart
openclaw models status --agent research-assistant

This does not touch API keys. It only changes model selection/config.

rocky pumice Apr 28, 2026, 10:37 AM

#

GPT-5.5
Claude Opus 4.7
Gemini 3.1 Pro
GPT-5.4
GPT-5.2
Claude Sonnet 4.6
Kimi K2.6
DeepSeek V3.2
Grok 4.1
Gemini 2.5 Flash
Claude Haiku 4.5
Qwen 3.5 9B
Grok 3 Mini
DeepSeek V4-Flash
GPT-5 Nano
Llama 4 Scout
Llama 4 Maverick
DeepSeek V4
Gemma 3n E4B

can you find the openrouter ids for all of these and get the allowlist command i have to paste to get it please

noble knotBOT Apr 28, 2026, 10:38 AM

#

Yep — I checked the live OpenRouter catalog. OpenClaw wants OpenRouter refs as openrouter/<provider>/<model>, while OpenRouter’s API model IDs are the inner <provider>/<model> part. (docs.openclaw.ai)

Paste this:

openclaw config set agents.defaults.models '{
  "openrouter/openai/gpt-5.5": {},
  "openrouter/anthropic/claude-opus-4.7": {},
  "openrouter/google/gemini-3.1-pro-preview": {},
  "openrouter/openai/gpt-5.4": {},
  "openrouter/openai/gpt-5.2": {},
  "openrouter/anthropic/claude-sonnet-4.6": {},
  "openrouter/moonshotai/kimi-k2.6": {},
  "openrouter/deepseek/deepseek-v3.2": {},
  "openrouter/x-ai/grok-4.1-fast": {},
  "openrouter/google/gemini-2.5-flash": {},

OpenClaw

OpenRouter - OpenClaw

#

  "openrouter/anthropic/claude-haiku-4.5": {},
  "openrouter/qwen/qwen3.5-9b": {},
  "openrouter/x-ai/grok-3-mini": {},
  "openrouter/deepseek/deepseek-v4-flash": {},
  "openrouter/openai/gpt-5-nano": {},
  "openrouter/meta-llama/llama-4-scout": {},
  "openrouter/meta-llama/llama-4-maverick": {},
  "openrouter/deepseek/deepseek-v4-pro": {},
  "openrouter/google/gemma-3n-e4b-it": {}
}' --strict-json --merge

openclaw config validate
openclaw gateway restart

Notes:

#

I mapped DeepSeek V4 to deepseek-v4-pro since you also listed V4-Flash separately.
I mapped Grok 4.1 to grok-4.1-fast; that’s the available OpenRouter ID I found.
This only updates the model picker/allowlist. It does not touch API keys or fallbacks.

rocky pumice Apr 28, 2026, 10:46 AM

#

Install failed: session-logs — brew not installed — Homebrew is not installed. Install it from https://brew.sh or install "jq" manually using your system package manager …
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills
│
◇ Install failed: summarize — brew not installed — Homebrew is not installed. Install it from https://brew.sh or install "steipete/tap/summarize" manually using your sys…
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills
│
◇ Install failed: obsidian — brew not installed — Homebrew is not installed. Install it from https://brew.sh or install "yakitrak/yakitrak/obsidian-cli" manually using …
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills
│
◇ Install failed: gemini — brew not installed — Homebrew is not installed. Install it from https://brew.sh or install "gemini-cli" manually using your system package …
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills
│
◇ Install failed: nano-pdf — uv not installed — install manually: https://docs.astral.sh/uv/getting-started/installation/
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills
│
◇ Install failed: github — brew not installed — Homebrew is not installed. Install it from https://brew.sh or install "gh" manually using your system package manager …
Tip: run openclaw doctor to review skills + requirements.
Docs: https://docs.openclaw.ai/skills

noble knotBOT Apr 28, 2026, 10:46 AM

#

Those are optional skill dependency installs, not OpenClaw core breaking.

On Linux, a bunch of skills still prefer Homebrew installers. If you don’t need those skills, you can ignore them or disable/narrow skills. Your agent will still work.

Useful checks:

openclaw skills check
openclaw skills list --eligible

If you want the common Linux deps, install them manually with your package manager, e.g. Ubuntu/Debian:

sudo apt update
sudo apt install -y jq ripgrep gh

#

npm install -g @google/gemini-cli

For nano-pdf, install uv first, then:

uv tool install nano-pdf

For summarize and obsidian, either install Linuxbrew or skip/disable those skills unless you specifically need them.

To disable noisy skills explicitly:

openclaw config set skills.entries.summarize.enabled false --strict-json

#

openclaw config set skills.entries.obsidian.enabled false --strict-json
openclaw config set skills.entries.nano-pdf.enabled false --strict-json
openclaw config validate
openclaw gateway restart

Docs: https://docs.openclaw.ai/tools/skills and https://docs.openclaw.ai/tools/skills-config

rocky pumice Apr 28, 2026, 11:01 AM

#

test@test-VirtualBox:~$ openclaw tui

🦞 OpenClaw 2026.4.26 (be8c246) — Shell yeah—I'm here to pinch the toil and leave you the glory.

openclaw tui - ws://127.0.0.1:18789 - agent research-assistant - session main
connecting | idle
gateway disconnected: closed | idle
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: gateway closed (1000):
────────────────────────────────────────────────────────────────────

noble knotBOT Apr 28, 2026, 11:01 AM

#

That closed (1000) is a clean close, so usually either auth/token mismatch, gateway restarted/closed during startup, or the TUI got rejected before it could attach.

Run these:

openclaw gateway status
openclaw status --all
openclaw logs --limit 120 --plain

Then try TUI with a small history load:

openclaw tui --history-limit 20

If it still closes, check auth mode/token state:

#

openclaw config get gateway.auth.mode
openclaw config get gateway.auth.token

If token auth is enabled and TUI isn’t picking it up, launch with the token explicitly:

openclaw tui --token "$(openclaw config get gateway.auth.token)"

Also, since you’ve been restarting/updating a lot, one clean restart is fine:

openclaw gateway restart
sleep 10

#

openclaw gateway status
openclaw tui --history-limit 20

rocky pumice Apr 28, 2026, 11:03 AM

#

make command for meta-llama/llama-4-scout to isntall as default and add to models list

noble knotBOT Apr 28, 2026, 11:03 AM

#

Use the OpenRouter-prefixed ID:

openclaw config set agents.defaults.models '{
  "openrouter/meta-llama/llama-4-scout": {}
}' --strict-json --merge

openclaw config set agents.list[0].model openrouter/meta-llama/llama-4-scout

openclaw config validate
openclaw gateway restart
openclaw models status --agent research-assistant

That adds it to the model list and sets it as the default for your current agent.

rocky pumice Apr 28, 2026, 11:08 AM

#

try again

run error: HTTP 401: User not found.
connected | error
agent research-assistant (Research Assistant) | session main (openclaw-tui) | openrouter/google/gemini-3-flash-preview | think high | tokens 48k/1.0m (5%)
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

noble knotBOT Apr 28, 2026, 11:09 AM

#

That’s still OpenRouter auth, not the model.

Also your TUI footer still shows:

openrouter/google/gemini-3-flash-preview

so it didn’t switch to Llama in that session yet. But 401 User not found would happen on any OpenRouter model if the key source is stale/bad.

Your earlier status showed OpenRouter auth coming from models.json, so clear that stale generated key source safely:

cp -a ~/.openclaw ~/.openclaw.backup-$(date +%F-%H%M%S)

mv ~/.openclaw/agents/research-assistant/agent/models.json \

#

  ~/.openclaw/agents/research-assistant/agent/models.json.bak-$(date +%F-%H%M%S) 2>/dev/null || true

openclaw models auth --agent research-assistant add
# custom -> openrouter -> profile id default -> paste fresh OpenRouter key

openclaw models status --agent research-assistant --probe
openclaw gateway restart

Then switch the active TUI session explicitly:

/model openrouter/meta-llama/llama-4-scout

or start fresh:

#

/session llama-test
/model openrouter/meta-llama/llama-4-scout

If --probe still says 401 User not found, the pasted OpenRouter key itself is invalid/revoked/wrong account.

rocky pumice Apr 28, 2026, 11:11 AM

#

🦞 OpenClaw 2026.4.26 (be8c246) — I autocomplete your thoughts—just slower and with more API calls.

│
◇ Token provider
│ custom (type provider id)
│
◇ Provider id
│ openrouter
│
◇ Profile id
│ openrouter:default
│
◇ Does this token expire?
│ No
│
◇ Paste token for openrouter
│ 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
Error: ENOSPC: no space left on device, write
test@test-VirtualBox:~$

noble knotBOT Apr 28, 2026, 11:11 AM

#

Stop there — you pasted the OpenRouter key publicly. Revoke that key now in OpenRouter and create a new one. Don’t reuse this key.

The actual error is:

ENOSPC: no space left on device

Your VM disk is full, so OpenClaw couldn’t save the token.

Run:

df -h
df -ih
du -h -d1 ~/.openclaw /tmp 2>/dev/null | sort -h

#

Quick safe cleanup options:

journalctl --user --vacuum-size=100M
rm -rf /tmp/openclaw/*

If ~/.openclaw is huge, check sessions/logs:

du -h -d2 ~/.openclaw | sort -h | tail -30

After freeing space, generate a fresh OpenRouter key, then rerun:

#

openclaw models auth --agent research-assistant add

Use:

provider id: openrouter
profile id: default — not openrouter:default; OpenClaw will make that full id itself.

rocky pumice Apr 28, 2026, 11:20 AM

#

think failed: GatewayClientRequestError: Error: ENOSPC: no space left on device, write: code=ENOSPC

noble knotBOT Apr 28, 2026, 11:20 AM

#

Yep — disk is full. OpenClaw can’t write session/auth/log files, so turns will fail until you free space.

Run:

df -h
du -h -d1 ~ ~/.openclaw /tmp 2>/dev/null | sort -h | tail -40

Quick safe cleanup:

journalctl --user --vacuum-size=100M
rm -rf /tmp/openclaw/*

Then check OpenClaw-heavy folders:

#

du -h -d2 ~/.openclaw | sort -h | tail -40

If sessions/logs are huge, you can compress backup first:

tar -czf ~/openclaw-sessions-backup-$(date +%F-%H%M%S).tgz ~/.openclaw/agents/*/sessions

Then use OpenClaw’s cleanup dry-run before deleting anything:

openclaw sessions cleanup --all-agents --dry-run

Also: revoke the OpenRouter key you pasted earlier, then create a fresh key after disk space is fixed.

rocky pumice Apr 28, 2026, 11:31 AM

#

test@test-VirtualBox:~$ openclaw gateway run

🦞 OpenClaw 2026.4.26 (be8c246)
I can run local, remote, or purely on vibes—results may vary with DNS.

│
◇
21:30:45 [gateway] loading configuration…
21:30:45 [gateway] resolving authentication…
21:30:45 [gateway] starting...

stuck on gateway starting now

noble knotBOT Apr 28, 2026, 11:31 AM

#

Most likely still the same root cause: disk full (ENOSPC), or you now have two gateways fighting for port 18789 because the systemd service may already be running.

In another terminal, run:

df -h
ss -ltnp | grep 18789 || true
openclaw gateway status

If port is already used by openclaw-gateway, stop the foreground one with Ctrl+C and use the service:

openclaw gateway restart
openclaw gateway status

#

If disk is full, fix that first:

journalctl --user --vacuum-size=100M
rm -rf /tmp/openclaw/*
du -h -d2 ~/.openclaw | sort -h | tail -40

Also don’t keep openclaw gateway run open if the service is installed — that’s for foreground/manual running. For your setup, prefer:

openclaw gateway restart
openclaw gateway status

rocky pumice Apr 28, 2026, 11:32 AM

#

test@test-VirtualBox:~$ df -h
ss -ltnp | grep 18789 || true
openclaw gateway status
Filesystem Size Used Avail Use% Mounted on
tmpfs 794M 1.5M 793M 1% /run
/dev/sda3 99G 48G 47G 51% /
tmpfs 3.9G 0 3.9G 0% /dev/shm
tmpfs 5.0M 8.0K 5.0M 1% /run/lock
/dev/sda2 512M 6.2M 506M 2% /boot/efi
tmpfs 794M 220K 794M 1% /run/user/1000
Virtual_Box_Shared_Folder_Cinnamon_test_1-_temp 450G 361G 90G 81% /media/sf_Virtual_Box_Shared_Folder_Cinnamon_test_1-_temp
/dev/sr0 51M 51M 0 100% /media/test/VBox_GAs_7.2.82
[1]+ Exit 127 nohup /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh > /dev/null 2>&1

🦞 OpenClaw 2026.4.26 (be8c246) — Alexa, but with taste.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Command: /usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 1255, state active, sub running, last exit 0, reason 0)
Warm-up: launch agents can take a few seconds. Try again shortly.
Connectivity probe: failed
Probe target: ws://127.0.0.1:18789
connect ECONNREFUSED 127.0.0.1:18789
Capability: unknown

noble knotBOT Apr 28, 2026, 11:33 AM

#

Disk looks fine now — root has 47G free. So this is not ENOSPC anymore.

Current problem:

Runtime: running pid 1255
Connectivity probe: ECONNREFUSED

Meaning the systemd process exists, but Gateway has not opened port 18789.

Run:

ps -fp 1255
journalctl --user -u openclaw-gateway.service -n 200 --no-pager

#

/usr/bin/tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log 2>&1
/usr/bin/tail -n 200 ~/.openclaw/logs/gateway-restart.log 2>&1

If you still have the foreground openclaw gateway run stuck in another terminal, press Ctrl+C there. Then restart only the service:

systemctl --user reset-failed openclaw-gateway.service
openclaw gateway restart
sleep 10
openclaw gateway status

Also that line:

#

auto_spawn_ui.sh Exit 127

means a startup script is failing with “command not found.” It may not be the gateway blocker, but we should inspect it later:

cat ~/.openclaw/workspace/scripts/auto_spawn_ui.sh

rocky pumice Apr 28, 2026, 11:35 AM

#

est@test-VirtualBox:~$ cat ~/.openclaw/workspace/scripts/auto_spawn_ui.sh
cat: /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh: No such file or directory

noble knotBOT Apr 28, 2026, 11:36 AM

#

That explains the Exit 127 line:

nohup ~/.openclaw/workspace/scripts/auto_spawn_ui.sh ...

127 = command/file not found. Since the script doesn’t exist, something is trying to auto-run a missing script.

It’s probably not the main gateway blocker, but you can find where it’s referenced:

grep -R "auto_spawn_ui.sh" ~/.bashrc ~/.profile ~/.config ~/.openclaw 2>/dev/null

If it shows a line in ~/.bashrc / ~/.profile, remove or comment that line.

For the gateway issue, still run:

#

ps -fp 1255
journalctl --user -u openclaw-gateway.service -n 200 --no-pager
/usr/bin/tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log 2>&1
/usr/bin/tail -n 200 ~/.openclaw/logs/gateway-restart.log 2>&1

The missing auto_spawn_ui.sh is cleanup; the logs above will show why the service process exists but isn’t listening on 18789.

rocky pumice Apr 28, 2026, 11:37 AM

#

what do i do after that

noble knotBOT Apr 28, 2026, 11:37 AM

#

After the grep, do this:

If it prints a file + line containing `auto_spawn_ui.sh`

Example:

/home/test/.bashrc: nohup /home/test/.openclaw/workspace/scripts/auto_spawn_ui.sh ...

Open that file and remove/comment that line:

nano ~/.bashrc

Put # in front of the auto_spawn_ui.sh line, save, then:

#

source ~/.bashrc

If `grep` prints nothing

Ignore the auto_spawn_ui.sh error for now. It’s not the main gateway issue.

Then continue with the real gateway check:

ps -fp 1255
journalctl --user -u openclaw-gateway.service -n 200 --no-pager
/usr/bin/tail -n 200 /tmp/openclaw/openclaw-2026-04-28.log 2>&1
/usr/bin/tail -n 200 ~/.openclaw/logs/gateway-restart.log 2>&1

After that, restart cleanly:

#

systemctl --user reset-failed openclaw-gateway.service
openclaw gateway restart
sleep 10
openclaw gateway status

Paste the first error-looking lines from the journal/log output if it still isn’t listening.

rocky pumice Apr 28, 2026, 11:38 AM

#

D6VW2Y3e.js:149","method":"console.error"}},"time":"2026-04-28T21:36:21.697+10:00","hostname":"test-VirtualBox","message":"Gateway restart timed out after 60s waiting for health checks."}
{"0":"Tip: openclaw gateway status --deep","_meta":{"runtime":"node","runtimeVersion":"24.14.1","hostname":"unknown","name":"openclaw","date":"2026-04-28T11:36:21.704Z","logLevelId":3,"logLevelName":"INFO","path":{"fullFilePath":"file:///home/test/.npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js:147:46","fileName":"console-D6VW2Y3e.js","fileNameWithLine":"console-D6VW2Y3e.js:147","fileColumn":"46","fileLine":"147","filePath":".npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js","filePathWithLine":".npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js:147","method":"console.log"}},"time":"2026-04-28T21:36:21.710+10:00","hostname":"test-VirtualBox","message":"Tip: openclaw gateway status --deep"}
{"0":"Tip: openclaw doctor","_meta":{"runtime":"node","runtimeVersion":"24.14.1","hostname":"unknown","name":"openclaw","date":"2026-04-28T11:36:21.718Z","logLevelId":3,"logLevelName":"INFO","path":{"fullFilePath":"file:///home/test/.npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js:147:46","fileName":"console-D6VW2Y3e.js","fileNameWithLine":"console-D6VW2Y3e.js:147","fileColumn":"46","fileLine":"147","filePath":".npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js","filePathWithLine":".npm-global/lib/node_modules/openclaw/dist/console-D6VW2Y3e.js:147","method":"console.log"}},"time":"2026-04-28T21:36:21.720+10:00","hostname":"test-VirtualBox","message":"Tip: openclaw doctor"}
[2026-04-28T08:55:19Z] openclaw restart attempt source=update target=openclaw-gateway.service
[2026-04-28T08:55:19Z] openclaw restart done source=update
test@test-VirtualBox:~$

noble knotBOT Apr 28, 2026, 11:38 AM

#

That only says the restart health check timed out — not the real startup error. Next, run the deep status and journal:

openclaw gateway status --deep
journalctl --user -u openclaw-gateway.service -n 250 --no-pager

Also run this to make sure the old xhost problem didn’t come back after update:

systemctl --user cat openclaw-gateway.service
systemctl --user show openclaw-gateway.service -p ExecStartPre -p ExecStart

If you see ExecStartPre=/usr/bin/xhost +local:test again, reapply the override:

#

mkdir -p ~/.config/systemd/user/openclaw-gateway.service.d
cat > ~/.config/systemd/user/openclaw-gateway.service.d/no-xhost.conf <<'EOF'
[Service]
ExecStartPre=
EOF
systemctl --user daemon-reload
systemctl --user reset-failed openclaw-gateway.service
openclaw gateway restart

If there’s no xhost line, paste the deep status/journal error lines — that’s where the actual blocker will be.

rocky pumice Apr 28, 2026, 11:39 AM

#

[Service]
ExecStart=/usr/bin/node /home/test/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Restart=always
RestartSec=5
RestartPreventExitStatus=78
TimeoutStopSec=30
TimeoutStartSec=30
SuccessExitStatus=0 143
KillMode=control-group
Environment=DISPLAY=:0
Environment=XAUTHORITY=/home/test/.Xauthority
Environment=HOME=/home/test
Environment=TMPDIR=/tmp
Environment=NODE_EXTRA_CA_CERTS=/etc/ssl/certs/ca-certificates.crt
Environment=PATH=/usr/bin:/home/test/.nvm/current/bin:/home/test/.local/bin:/home/test/.npm-global/bin:/home/test/bin:/home/test/.volta/bin:/home/test/.asdf/shims:/home/test/.bun/bin:/home/test/.nix-profile/bin:/home/test/.local/share/fn>
Environment=OPENCLAW_GATEWAY_PORT=18789
Environment=OPENCLAW_SYSTEMD_UNIT=openclaw-gateway.service
Environment="OPENCLAW_WINDOWS_TASK_NAME=OpenClaw Gateway"
Environment=OPENCLAW_SERVICE_MARKER=openclaw
Environment=OPENCLAW_SERVICE_KIND=gateway
Environment=OPENCLAW_SERVICE_VERSION=2026.4.26

noble knotBOT Apr 28, 2026, 11:39 AM

#

Good — the old xhost blocker is not back. There’s no ExecStartPre= line now.

Next step is to find where Gateway is hanging during startup. Do this clean isolation run:

openclaw gateway stop
sleep 3

OPENCLAW_GATEWAY_STARTUP_TRACE=1 \
OPENCLAW_LOG_LEVEL=debug \
openclaw gateway run --verbose 2>&1 | tee ~/openclaw-gateway-debug.log

Let it sit ~60 seconds. If it still stops at starting..., press Ctrl+C, then paste:

#

tail -n 120 ~/openclaw-gateway-debug.log

Also run:

journalctl --user -u openclaw-gateway.service -n 250 --no-pager
openclaw gateway status --deep

If the trace hangs before “ready,” the last startup trace line will tell us what subsystem is blocking — config, auth, plugin deps, model provider, etc.

rocky pumice Apr 28, 2026, 11:42 AM

#

1:41:42.826+10:00 [gateway] startup trace: sidecars.restart-sentinel 9.0ms total=48920.3ms eventLoopMax=0.0ms
2026-04-28T21:41:42.936+10:00 [gateway] startup trace: sidecars.subagent-recovery 106.1ms total=49037.2ms eventLoopMax=0.0ms
2026-04-28T21:41:42.950+10:00 [gateway] startup trace: sidecars.main-session-recovery 9.8ms total=49050.8ms eventLoopMax=0.0ms
2026-04-28T21:41:42.953+10:00 [gateway] startup trace: sidecars.total 45066.0ms total=49054.9ms eventLoopMax=0.0ms
2026-04-28T21:41:42.957+10:00 [gateway] startup trace: sidecars.ready 3.2ms total=49058.1ms eventLoopMax=0.0ms
2026-04-28T21:41:42.961+10:00 [gateway] ready
2026-04-28T21:41:42.967+10:00 [gateway] startup trace: runtime.post-attach 45370.3ms total=49069.4ms eventLoopMax=0.0ms
2026-04-28T21:41:42.969+10:00 [gateway] startup trace: ready 2.4ms total=49071.8ms eventLoopMax=0.0ms
2026-04-28T21:41:42.981+10:00 [heartbeat] started
2026-04-28T21:41:42.998+10:00 [plugins] [hooks] running gateway_start (1 handlers)
2026-04-28T21:42:05.809+10:00 [plugins] bonjour: watchdog detected non-announced service; attempting re-advertise (gateway fqdn=test-VirtualBox (OpenClaw) (2)._openclaw-gw._tcp.local. host=test-VirtualBox-(2).local. port=18789 state=probing)
2026-04-28T21:42:06.667+10:00 [plugins] bonjour: gateway name conflict resolved; newName="test-VirtualBox (OpenClaw) (2)"
2026-04-28T21:42:06.668+10:00 [plugins] bonjour: gateway hostname conflict resolved; newHostname="test-VirtualBox-(2)"
2026-04-28T21:42:06.671+10:00 [plugins] bonjour: advertised gateway fqdn=test-VirtualBox (OpenClaw) (2)._openclaw-gw._tcp.local. host=test-VirtualBox-(2).local. port=18789 state=announcing

noble knotBOT Apr 28, 2026, 11:43 AM

#

That’s actually not stuck — it reached:

[gateway] ready

It just took ~49s, mostly here:

sidecars.total 45066ms
runtime.post-attach 45370ms

So the restart “timed out after 60s” was close to the edge, but the gateway did come up.

Now, in a second terminal, check:

#

openclaw gateway status
openclaw tui --history-limit 20

If you want startup to be less slow, the next suspect is sidecars/plugins. Bonjour also shows name conflicts. You can disable Bonjour if you don’t need LAN discovery:

openclaw config set plugins.entries.bonjour.enabled false --strict-json
openclaw config validate
openclaw gateway restart

Then compare startup time. But first confirm gateway status says connectivity ok.

rocky pumice Apr 28, 2026, 11:44 AM

#

would i not need lan recovery if i want to be able to talk with my main computer from vm?

noble knotBOT Apr 28, 2026, 11:44 AM

#

You don’t need Bonjour specifically for that.

Bonjour is just LAN auto-discovery/advertising. For your main computer to talk to the VM, what matters is:

VM network mode allows host ↔ VM traffic
Gateway binds to LAN, not only loopback
Gateway auth is configured

Right now your gateway is loopback-only:

127.0.0.1:18789

That means only the VM itself can connect.

If you want your main computer to connect to the VM gateway, configure LAN bind + token auth:

#

openclaw doctor --generate-gateway-token
openclaw config set gateway.bind lan
openclaw config validate
openclaw gateway restart

Then get the VM IP:

hostname -I

From your main computer, connect to:

#sessions and previous memories gone. permissions and tools also gone

1) Tools/permissions “gone”

2) Old session exists, but under old agent id

3) Memory file

Add multiple models to picker/allowlist

Set primary + fallbacks at once

If your agent has its own model override, set that too

If it prints a file + line containing auto_spawn_ui.sh

If grep prints nothing

If it prints a file + line containing `auto_spawn_ui.sh`

If `grep` prints nothing