Openclaw is running extremely slow | Friends of the Crustacean 🦞🤝 | Page 1

subtle reef Apr 29, 2026, 10:40 PM

#

Openclaw is running extremely slow can you help me fix this

meager hedgeBOT Apr 29, 2026, 10:40 PM

#

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

uneven hedgeBOT Apr 29, 2026, 10:41 PM

#

Yep — first thing: please paste your version and a safe status report:

openclaw --version
openclaw status --all
openclaw gateway status

If it’s slow only in one long chat/thread, also try:

/status
/compact

or start fresh with:

#

/new

Common causes of “extremely slow”:

Huge conversation context → /compact or /new
Local/Ollama model too large or huge context window → CPU/VRAM gets pinned
Provider/model latency → try a faster cloud model temporarily
Gateway stuck/busy → check openclaw status --all + logs
Tool/browser work running in background → /status shows what’s active

Docs: https://docs.openclaw.ai/help/faq and https://docs.openclaw.ai/help/troubleshooting

Paste the outputs above — especially openclaw --version — and I’ll help narrow it down. 🦐

subtle reef Apr 29, 2026, 10:46 PM

#

here is the version, the status is too much to put into this window OpenClaw 2026.4.26 (be8c246)

uneven hedgeBOT Apr 29, 2026, 10:47 PM

#

Thanks — 2026.4.26 is useful.

Don’t paste the giant status into Discord; upload it as a file instead:

openclaw status --all > openclaw-status.txt
openclaw logs --limit 300 --plain > openclaw-logs.txt

Attach those two files if you can.

Quick checks/fixes meanwhile:

openclaw doctor
openclaw gateway status

#

If it’s slow only in this one conversation/thread:

/status
/compact

If it’s slow everywhere, on 2026.4.26 also check this known-ish failure pattern:

python3 -m json.tool ~/.openclaw/devices/pending.json >/dev/null && echo "pending ok" || echo "pending.json corrupted"

If that says corrupted:

#

mv ~/.openclaw/devices/pending.json ~/.openclaw/devices/pending.json.bad
openclaw gateway restart

Then try again.

Docs for the triage flow: https://docs.openclaw.ai/help/troubleshooting 🦐

subtle reef Apr 29, 2026, 10:47 PM

#

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

│
◇
Service: LaunchAgent (loaded)
File logs: /tmp/openclaw/openclaw-2026-04-29.log
Command: /opt/homebrew/opt/node@22/bin/node /opt/homebrew/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/Library/LaunchAgents/ai.openclaw.gateway.plist
Working dir: ~/.openclaw
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 21576, state active)
Connectivity probe: ok
Capability: admin-capable

Listening: 127.0.0.1:18789

uneven hedgeBOT Apr 29, 2026, 10:48 PM

#

Gateway itself looks healthy: service is running, port is reachable, admin probe works. So the slowness is probably inside the agent run: model latency, huge session context, a stuck tool, or local/Ollama resource pressure.

Next best checks:

In the slow chat, send:

/status

If it shows a huge context / lots of compactions / active tool work, run:

/compact

or for a clean test:

#

/new

Attach the status/logs as files, not pasted text:

openclaw status --all > openclaw-status.txt
openclaw logs --limit 300 --plain > openclaw-logs.txt

If you’re using a local model/Ollama, also run:

ollama ps

#

and tell us the model name + whether CPU/RAM/VRAM is pegged.

The gateway status you pasted rules out “OpenClaw isn’t running”; now we need to see whether it’s waiting on the model, context building, tools, or something wedged in logs. Docs: https://docs.openclaw.ai/help/troubleshooting 🦐

subtle reef Apr 29, 2026, 10:48 PM

#

Also please dont give me multiple commands to run all in one response from ypu. it makes it very difficult to keeep up. 1 command and response at a time will keep us on track

uneven hedgeBOT Apr 29, 2026, 10:49 PM

#

Absolutely — my bad. We’ll do one step at a time.

Please run this in the slow OpenClaw chat:

/status

Paste just the result from that.

subtle reef Apr 29, 2026, 10:50 PM

#

🦞 OpenClaw 2026.4.26 (be8c246)
🧠 Model: openrouter/google/gemma-4-31b-it · 🔑 api-key (openrouter:default)
🧮 Tokens: 255k in / 118 out
📚 Context: 64k/262k (24%) · 🧹 Compactions: 0
🧵 Session: agent:main:main • updated just now
⚙️ Execution: direct · Runtime: OpenClaw Pi Default · Think: high
🪢 Queue: collect (depth 0)

uneven hedgeBOT Apr 29, 2026, 10:50 PM

#

Got it — this status says the gateway isn’t stuck. Queue is empty, but the chat has 64k context, which can make each reply slow.

Next single step: in that same slow OpenClaw chat, run:

/compact

After it finishes, paste the short result. Docs: https://docs.openclaw.ai/concepts/compaction

subtle reef Apr 29, 2026, 10:53 PM

#

it not giving me any response when I type /compact

uneven hedgeBOT Apr 29, 2026, 10:53 PM

#

That can happen — compaction notices may be quiet, or it may not have triggered visibly.

Next single step: run this again in the same chat:

/status

Paste the result. I want to see whether Compactions changed from 0 or context dropped from 64k.

subtle reef Apr 29, 2026, 10:54 PM

#

🦞 OpenClaw 2026.4.26 (be8c246)
🧠 Model: openrouter/google/gemma-4-31b-it · 🔑 api-key (openrouter:default)
🧮 Tokens: 3.5m in / 21k out
📚 Context: 64k/262k (24%) · 🧹 Compactions: 1
🧵 Session: agent:main:main • updated just now
⚙️ Execution: direct · Runtime: OpenClaw Pi Default · Think: high
🪢 Queue: collect (depth 0)

uneven hedgeBOT Apr 29, 2026, 10:54 PM

#

Compaction did run (Compactions: 1), but context is still 64k, so this session is still heavy.

Next single step: start a clean test session in that same chat:

/new

Then send a simple test message like “hi” and tell me if the reply is fast or still slow.

subtle reef Apr 29, 2026, 10:56 PM

#

it took 35 seconds for it to respond after i typed new. This is what the response finally was: Ready for execution, Warren. I'm currently running on Gemma 4, though the default is Qwen 2.5 72B. What's the high-ROI priority for today?

uneven hedgeBOT Apr 29, 2026, 10:56 PM

#

35 seconds on a brand-new session points more toward model/provider + Think: high, not old context.

Next single step: in that same OpenClaw chat, run:

/think off

Then tell me what it replies.

subtle reef Apr 29, 2026, 10:57 PM

#

so far no response from think off

uneven hedgeBOT Apr 29, 2026, 10:57 PM

#

Got it. If /think off isn’t acknowledging, try the colon form — one command:

/think:off

This should reply with something like “Thinking disabled.”

subtle reef Apr 29, 2026, 10:58 PM

#

Nothin in the reply .. completely blank. I'm in the opeenclaw dashboard chat UI

uneven hedgeBOT Apr 29, 2026, 10:59 PM

#

Ah, okay — Dashboard Chat UI changes this. Blank reply means the slash command may not be getting handled cleanly there.

Next single step: run this in your terminal right after that blank reply:

openclaw logs --limit 120 --plain

Paste the last chunk that shows the /think:off attempt or any errors.

subtle reef Apr 29, 2026, 11:00 PM

#

there is a lot of output i dont know how to get that to you

rare saddle Apr 29, 2026, 11:00 PM

#

pastebin

uneven hedgeBOT Apr 29, 2026, 11:01 PM

#

No worries — don’t paste it. Make a small file and attach it here.

Run this one command:

openclaw logs --limit 120 --plain > ~/Desktop/openclaw-logs.txt

Then drag openclaw-logs.txt from your Desktop into this Discord thread.

#

Yep, Pastebin/Gist works too — just make sure tokens/keys are redacted. openclaw logs --plain usually avoids obvious secrets, but still skim before posting if you can.

subtle reef Apr 29, 2026, 11:04 PM

#

sorry I cant drag that here. system just drops it

uneven hedgeBOT Apr 29, 2026, 11:04 PM

#

No problem. Let’s avoid file upload.

Run this one command — it only shows suspicious/error lines:

openclaw logs --limit 300 --plain | grep -Ei "error|warn|timeout|slow|blank|empty|fail|abort|openrouter|gemma|think|slash"

Paste whatever that prints.

subtle reef Apr 29, 2026, 11:06 PM

#

bleTotalTokens=0 promotedToolResults=0 promotedOrdinals=none removedToolUseBlocks=0 touchedAssistantMessages=0
2026-04-29T22:55:17.501Z info {"plugin":"lossless-claw"} [lcm] ingestBatch: session queue acquired queueKey=agent:main:main queuedAhead=0 wait=0ms session=ef0e9530-651b-491b-b8de-2cc0ae3417b0 sessionKey=agent:main:main messages=1
2026-04-29T22:55:17.503Z info {"plugin":"lossless-claw"} [lcm] incremental compaction decision: conversation=1 cacheState=cold activityBand=low tokenBudget=262144 currentTokenCount=0 cacheRead=0 cacheWrite=0 cachePromptTokenCount=63887 cacheReadSharePct=0.0% triggerLeafChunkTokens=20000 preferredLeafChunkTokens=40000 fallbackLeafChunkTokens=40000,30000,20000 rawTokensOutsideTail=0 threshold=20000 shouldCompact=false maxPasses=1 allowCondensedPasses=false reason=below-leaf-trigger
2026-04-29T22:55:17.505Z warn {"plugin":"lossless-claw"} [lcm] afterTurn: bootstrap checkpoint refresh failed for session=ef0e9530-651b-491b-b8de-2cc0ae3417b0 sessionKey=agent:main:main: ENOENT: no such file or directory, stat '/Users/warren/.openclaw/agents/main/sessions/ef0e9530-651b-491b-b8de-2cc0ae3417b0.jsonl'
2026-04-29T22:55:17.506Z info {"plugin":"lossless-claw"} [lcm] afterTurn: done conversation=1 session=ef0e9530-651b-491b-b8de-2cc0ae3417b0 sessionKey=agent:main:main newMessages=1 dedupedMessages=1 ingestedMessages=1 duration=13ms
2026-04-29T22:55:17.506Z info {"plugin":"lossless-claw"} [lcm] assemble: done conversation=1 session=ef0e9530-651b-491b-b8de-2cc0ae3417b0 sessionKey=agent:main:main contextItems=1 hasSummaryItems=false inputMessages=1 outputMessages=1 tokenBudget=262144 estimatedTokens=0 duration=0ms Thi is as much as i coul fit into the chat

uneven hedgeBOT Apr 29, 2026, 11:06 PM

#

That snippet helps. I see lossless-claw is active, and it’s still assembling around 64k prompt tokens even after /new. That’s likely why it feels slow.

Let’s test by switching back to the built-in context engine.

Run this one command:

openclaw config set plugins.slots.contextEngine legacy

Paste what it says.

subtle reef Apr 29, 2026, 11:07 PM

#

No I just installed that a few minutes ago. The system has been slow for days

uneven hedgeBOT Apr 29, 2026, 11:08 PM

#

Fair correction — then lossless-claw probably isn’t the original cause. Thanks for catching that.

Let’s isolate whether the model/provider itself is slow outside the chat/session.

Run this one command:

time openclaw infer model run --model openrouter/google/gemma-4-31b-it --prompt "Reply exactly: pong" --json

Tell me roughly how many seconds it takes.

subtle reef Apr 29, 2026, 11:15 PM

#

it returned this after about 30 seconds, but it may be stil processing more outbut i am not sure: warren@warrens-mini ~ % time openclaw infer model run --model openrouter/google/gemma-4-31b-it --prompt "Reply exactly: pong" --json
plugin runtime config.loadConfig() is deprecated (runtime-config-load-write); use config.current().

uneven hedgeBOT Apr 29, 2026, 11:15 PM

#

That line is just a plugin warning, not the model result yet.

Next step: wait until your terminal prompt comes back.

When it finishes, tell me only:

did it print pong?
the final real time from time

If it’s still not done after ~2 minutes, press Ctrl+C and tell me “stuck after 2 min”.

subtle reef Apr 29, 2026, 11:15 PM

#

actually it cam back with more after about 2 minutes: pong" --json
plugin runtime config.loadConfig() is deprecated (runtime-config-load-write); use config.current().
19:14:59 [agent/embedded] [tools] No callable tools remain after resolving explicit tool allowlist (tools.allow: *, browser); tools are disabled for this run. Fix the allowlist or enable the plugin that registers the requested tool.
19:14:59 [agent/embedded] embedded run failover decision: runId=ef0e9530-651b-491b-b8de-2cc0ae3417b0 stage=prompt decision=surface_error reason=none from=openrouter/google/gemma-4-31b-it profile=sha256:ac092b59b472 rawError=No callable tools remain after resolving explicit tool allowlist (tools.allow: *, browser); tools are disabled for this run. Fix the allowlist or enable the plugin that registers the requested tool.
19:14:59 [diagnostic] lane task error: lane=main durationMs=68247 error="Error: No callable tools remain after resolving explicit tool allowlist (tools.allow: *, browser); tools are disabled for this run. Fix the allowlist or enable the plugin that registers the requested tool."
19:14:59 [diagnostic] lane task error: lane=session:agent:main:main durationMs=68248 error="Error: No callable tools remain after resolving explicit tool allowlist (tools.allow: *, browser); tools are disabled for this run. Fix the allowlist or enable the plugin that registers the requested tool."
Error: No callable tools remain after resolving explicit tool allowlist (tools.allow: *, browser); tools are disabled for this run. Fix the allowlist or enable the plugin that registers the requested tool.
openclaw infer model run --model openrouter/google/gemma-4-31b-it --prompt 45.49s user 58.67s system 100% cpu 1:43.47 total

uneven hedgeBOT Apr 29, 2026, 11:17 PM

#

That output found a real config problem:

tools.allow: *, browser is causing runs to fail at the prompt stage with “No callable tools remain…”. Even if this isn’t the only slowdown, we should fix that first because it’s definitely broken.

Next single command:

openclaw config get tools.allow --json

Paste what it prints.

subtle reef Apr 29, 2026, 11:18 PM

#

I was doing some browser control automatio stuff but i do wan to get rid of it

uneven hedgeBOT Apr 29, 2026, 11:18 PM

#

Makes sense. The risky bit is tools.allow being an absolute allowlist; it can accidentally remove the normal tool profile and break runs.

Next single command to remove it:

openclaw config unset tools.allow

Paste what it says.

subtle reef Apr 29, 2026, 11:21 PM

#

lets pause for a sec as I'm having openclae remove and delet all of the browser automation stuff I was doing that related to google.

uneven hedgeBOT Apr 29, 2026, 11:21 PM

#

Yep, pause is good.

One caution: deleting browser automation stuff may not fix this specific error unless tools.allow is also removed or changed. That error is from config policy, not just old browser files.

I’ll wait — when that cleanup finishes, send me what changed or any error it reports.

subtle reef Apr 29, 2026, 11:29 PM

#

got this on the first command warren@warrens-mini ~ % openclaw config get tools.allow --json
Config path not found: tools.allow

uneven hedgeBOT Apr 29, 2026, 11:29 PM

#

Good — that means the global tools.allow is gone now. The earlier failing run may have been using the old loaded config.

Next single step: restart the gateway so it reloads cleanly.

openclaw gateway restart

Tell me if it says restarted successfully or shows an error.

subtle reef Apr 29, 2026, 11:31 PM

#

got tis on the second commsnd: warren@warrens-mini ~ % openclaw config unset tools.allow

🦞 OpenClaw 2026.4.26 (be8c246) — Built by lobsters, for humans. Don't question the hierarchy.

Config path not found: tools.allow

uneven hedgeBOT Apr 29, 2026, 11:31 PM

#

That’s fine — it just means there’s nothing left to unset.

Next single command:

openclaw gateway restart

Paste the result.

subtle reef Apr 29, 2026, 11:35 PM

#

is restarting now

uneven hedgeBOT Apr 29, 2026, 11:35 PM

#

Good. Let it finish.

When the prompt comes back, tell me whether it restarted cleanly or paste the error if it doesn’t.

subtle reef Apr 29, 2026, 11:40 PM

#

no error but it took a long time to respoind. in the telegram chat i typed are you there and it cam back with this: ontext limit exceeded. I've reset our conversation to start fresh - please try again.

To prevent this, increase your compaction buffer by setting agents.defaults.compaction.reserveTokensFloor to 20000 or higher in your config.

uneven hedgeBOT Apr 29, 2026, 11:40 PM

#

That Telegram message means the chat history was overflowing and OpenClaw reset it. Good clue.

Next single command — check whether your compaction reserve is misconfigured:

openclaw config get agents.defaults.compaction --json

Paste the output.

subtle reef Apr 29, 2026, 11:41 PM

#

warren@warrens-mini ~ % openclaw config get agents.defaults.compaction --json
Config path not found: agents.defaults.compaction

uneven hedgeBOT Apr 29, 2026, 11:41 PM

#

Okay — no custom compaction config. Let’s add the recommended reserve floor.

Run this one command:

openclaw config set agents.defaults.compaction.reserveTokensFloor 20000 --strict-json

Paste what it says.

subtle reef Apr 29, 2026, 11:42 PM

#

Updated agents.defaults.compaction.reserveTokensFloor. Restart the gateway to apply.
warren@warrens-mini ~ %

#Openclaw is running extremely slow