gateway disconnected after kimi claw | Friends of the Crustacean 🦞🤝 | Page 1

robust orioleBOT Feb 15, 2026, 9:45 PM

#

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

delicate flaxBOT Feb 15, 2026, 9:47 PM

#

First thing: please delete/edit your message to remove that bot token — it’s now public in a Discord thread, so you should rotate/regenerate it immediately on the Kimi side. Don’t paste new tokens here.

On the OpenClaw side, openclaw tui showing:

gateway disconnected: closed

usually comes down to one of these:

the gateway service isn’t actually running (or is crash-looping)
port 18789 is taken by something else (possibly the kimi-claw install)
auth/config mismatch: the systemd service is using a different config/token than your CLI/TUI

Run these and paste the output (redact tokens/keys):

openclaw --version

#

openclaw gateway status --deep
openclaw status
openclaw logs --last 200

Then also check whether something else is sitting on the gateway port:

sudo ss -lntp | rg ':18789' || true

Common fixes (depending on what the above shows)

A) “Config (cli)” vs “Config (service)” mismatch in gateway status --deep:

#

openclaw gateway install --force
openclaw gateway restart

B) Port conflict (something else is using 18789):
Either stop the conflicting service, or move OpenClaw to a new port:

openclaw config set gateway.port 18790 --json
openclaw gateway restart
openclaw tui --url ws://127.0.0.1:18790

Docs runbook (matches the commands above): https://docs.openclaw.ai/gateway/troubleshooting

Paste openclaw gateway status --deep + the ss output and we can pinpoint which case you’re in.

cobalt summit Feb 15, 2026, 9:54 PM

#

│ │
├──────────────────────────────────────────────╯
Error: Config validation failed: plugins.load.paths: plugin: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw
axis@vmi3046347:~$ openclaw tui
Invalid config at /home/axis/.openclaw/openclaw.json:\n- plugins.load.paths: plugin: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw

🦞 OpenClaw 2026.2.13 (203b5bd) — Pairing codes exist because even bots believe in consent—and good security hygiene.

│
◇ Config ───────────────────────────────────────────────────╮
│ │
│ Config invalid; doctor will run with best-effort config. │
│ │
├────────────────────────────────────────────────────────────╯
Config invalid
File: ~/.openclaw/openclaw.json
Problem:

plugins.load.paths: plugin: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw

delicate flaxBOT Feb 15, 2026, 9:56 PM

#

That explains it: your OpenClaw config is invalid, so the gateway won’t start and the TUI can’t connect.

plugins.load.paths: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw

So your config is pointing at a plugin directory that no longer exists (or never existed).

Fix (recommended): remove the broken plugin path from config

Back up your config:

cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bak

Edit it:

nano ~/.openclaw/openclaw.json

#

Find and remove the entry that mentions:

plugins: {
  load: { paths: ["/home/axis/.openclaw/extensions/kimi-claw"] }
}

If you don’t care about plugins right now, simplest is: remove the whole plugins: { ... } block (anything referencing kimi-claw).

Restart:

openclaw gateway restart
openclaw tui

If you wanted that plugin

Then you must reinstall it so the path exists and contains a valid openclaw.plugin.json manifest (missing/invalid plugins will also fail validation). Use OpenClaw’s plugin tooling rather than random scripts:

#

Important security note

You posted a bot token earlier in the thread — assume it’s compromised and rotate/regenerate it (don’t paste the new one into Discord).

If you paste the plugins section of your ~/.openclaw/openclaw.json (redacting secrets), I can tell you exactly what to delete.

cobalt summit Feb 15, 2026, 9:57 PM

#

}
}
},
"plugins": {
"load": {
"paths": [
"/home/axis/.openclaw/extensions/kimi-claw"
]
},

delicate flaxBOT Feb 15, 2026, 9:57 PM

#

Yep — that snippet is exactly the thing breaking your config.

Do this:

Find every “kimi” reference in the config:

grep -n "kimi" ~/.openclaw/openclaw.json

Edit the file:

nano ~/.openclaw/openclaw.json

Remove the whole plugins.load.paths entry (or simplest: remove the entire "plugins": { ... } block), i.e. delete:

#

"plugins": {
  "load": {
    "paths": [
      "/home/axis/.openclaw/extensions/kimi-claw"
    ]
  },
},

Important: make sure the commas/braces still match after deleting (JSON5 is tolerant, but missing/extra commas/braces will still break it).

Restart + test:

openclaw gateway restart
openclaw tui

#

If you run into another validation error after removing that path, paste the new error (it may be another plugins.entries / allowlist referencing the same plugin).

cobalt summit Feb 15, 2026, 9:59 PM

#

}
}
},
},
"entries": {
"telegram": {
"enabled": true
}
}
}
}

its deleted, is this fine now

#

@delicate flax

delicate flaxBOT Feb 15, 2026, 10:03 PM

#

Not sure yet — that pasted snippet looks not fine (it has extra }, / braces), and it also shows:

"entries": { "telegram": { "enabled": true } }

If that’s under plugins, it will likely still fail validation because:

plugins.entries.<id> must refer to a real installed plugin id
telegram is a channel, not a plugin id (unless you installed a plugin literally named telegram, which is uncommon)

Fastest “get me unstuck” fix

Open ~/.openclaw/openclaw.json and delete the entire "plugins": { ... } block (including entries), then:

openclaw gateway restart

#

openclaw tui

How to confirm

Just run this — it will tell you immediately if the config is valid:

openclaw doctor

If it still errors, paste the new exact validation error text (no need to paste the whole config).

#

Run this and paste the output (it’ll tell us immediately if the config is still invalid and why):

openclaw doctor

If it still complains about plugins, also paste:

openclaw gateway status --deep
openclaw logs --last 200

(Please don’t paste any tokens/secrets.)

cobalt summit Feb 15, 2026, 10:06 PM

#

bruh
"wizard": {
"lastRunAt": "2026-02-15T22:04:04.039Z",
"lastRunVersion": "2026.2.13",
"lastRunCommand": "doctor",
"lastRunMode": "local"
},
"commands": {
"native": "auto",
"nativeSkills": "auto"
},
"meta": {
"lastTouchedVersion": "2026.2.13",
"lastTouchedAt": "2026-02-15T22:04:04.070Z"
}
}

openclaw deleted everything? how do i get back my shit

#

│ Config: /home/axis/.openclaw/openclaw.json │
│ Bind: loopback │
│ │
├──────────────────────────────────────────────╯
Config overwrite: /home/axis/.openclaw/openclaw.json (sha256 c0e496bd1a8346e28c4c20bb4ba4c2cc18f1579365e1428e259df97c95ac3f31 -> ff9567212f767f44bb8472793fd9f65a00d6bf369e8b26fd56be9801f0d033ef, backup=/home/axis/.openclaw/openclaw.json.bak)
Config write anomaly: /home/axis/.openclaw/openclaw.json (size-drop:4652->335, missing-meta-before-write)
Updated ~/.openclaw/openclaw.json
Backup: ~/.openclaw/openclaw.json.bak
│
└ Doctor complete.

delicate flaxBOT Feb 15, 2026, 10:07 PM

#

Those "wizard" / "commands" / "meta" blocks are normal — they get appended/stamped whenever you run openclaw doctor/configure/onboard. By itself, that doesn’t mean your config was wiped.

But if your ~/.openclaw/openclaw.json now contains only that stuff, then yeah: the rest of the file got deleted/overwritten (usually by an edit gone wrong, or a “best-effort” rewrite after a badly-broken JSON5 file).

1) Check if your config still has your real settings

On the server, run:

head -n 40 ~/.openclaw/openclaw.json

You should see keys like channels, agents, models, etc.
If you don’t, continue below.

2) Restore from a backup (most likely fix)

List any backups Nano/you/doctor might have left:

#

ls -la ~/.openclaw | rg "openclaw\\.json"

If you see something like openclaw.json.bak, restore it safely:

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.now
cp -a ~/.openclaw/openclaw.json.bak ~/.openclaw/openclaw.json
openclaw gateway restart

3) If there’s no backup

Then we can’t “recover” the old config from OpenClaw alone — but your state (sessions/credentials) is probably still there in ~/.openclaw/ (doctor doesn’t nuke that).

Reply with the output of:

#

ls -la ~/.openclaw

(and tell me which channels you had configured: telegram/discord/whatsapp/etc). Then I’ll tell you the cleanest way to rebuild the config without re-breaking it.

#

Good news: it’s recoverable. Doctor explicitly says it made a backup:

backup=/home/axis/.openclaw/openclaw.json.bak

and it also detected a size drop 4652 → 335 bytes, so your “shit” is in the backup.

Restore your old config

Run:

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.doctor-broken
cp -a ~/.openclaw/openclaw.json.bak ~/.openclaw/openclaw.json
openclaw gateway restart
openclaw tui

After it’s back up

#

You’ll still want to remove the bad kimi-claw plugin reference (so it doesn’t break again). Do it safely this time:

grep -n "kimi-claw" ~/.openclaw/openclaw.json
nano ~/.openclaw/openclaw.json
# delete the plugins.load.paths entry that points to kimi-claw
openclaw gateway restart

If you want, paste the plugins block from the restored config and I’ll tell you exactly what to delete without breaking braces/commas again.

cobalt summit Feb 15, 2026, 10:08 PM

#

──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
axis@vmi3046347:~$ head -n 40 ~/.openclaw/openclaw.json
{
"wizard": {
"lastRunAt": "2026-02-15T22:04:04.039Z",
"lastRunVersion": "2026.2.13",
"lastRunCommand": "doctor",
"lastRunMode": "local"
},
"commands": {
"native": "auto",
"nativeSkills": "auto"
},
"meta": {
"lastTouchedVersion": "2026.2.13",
"lastTouchedAt": "2026-02-15T22:04:04.070Z"
}
}
axis@vmi3046347:~$ s -la ~/.openclaw | rg "openclaw\.json"
s: command not found
Command 'rg' not found, but can be installed with:
sudo snap install ripgrep # version 12.1.0, or
sudo apt install ripgrep # version 13.0.0-2ubuntu0.1
See 'snap info ripgrep' for additional versions.
axis@vmi3046347:~$ ls -la ~/.openclaw | rg "openclaw\.json"
Command 'rg' not found, but can be installed with:
sudo snap install ripgrep # version 12.1.0, or
sudo apt install ripgrep # version 13.0.0-2ubuntu0.1
See 'snap info ripgrep' for additional versions.

#

axis@vmi3046347:~$ ls -la ~/.openclaw
total 212
drwx------ 21 axis axis 4096 Feb 15 23:06 .
drwxr-x--- 39 axis axis 4096 Feb 15 22:41 ..
-rw------- 1 axis axis 569 Feb 7 23:54 .env
drwx------ 3 axis axis 4096 Jan 30 20:40 agents
drwxrwxr-x 5 axis axis 4096 Jan 31 19:24 backup-20260131
-rw-rw-r-- 1 axis axis 14360 Jan 31 19:24 backup-20260131.tar.gz
drwxrwxr-x 4 axis axis 4096 Feb 1 11:17 browser
drwxrwxr-x 2 axis axis 4096 Jan 30 21:49 canvas
drwxrwxr-x 2 axis axis 4096 Feb 9 11:12 completions
drwx------ 2 axis axis 4096 Jan 31 01:22 credentials
drwxrwxr-x 3 axis axis 4096 Feb 15 22:49 cron
drwx------ 3 axis axis 4096 Feb 15 21:07 delivery-queue
drwxrwxr-x 2 axis axis 4096 Feb 15 22:48 devices
drwxrwxr-x 2 axis axis 4096 Feb 15 22:49 extensions
drwxrwxr-x 3 axis axis 4096 Feb 15 22:37 extensions-backups
drwxrwxr-x 2 axis axis 4096 Jan 30 21:52 identity
drwxrwxr-x 2 axis axis 4096 Feb 15 22:43 logs
drwx------ 3 axis axis 4096 Feb 4 02:22 media
drwxrwxr-x 2 axis axis 4096 Feb 9 10:07 memory
-rw------- 1 axis axis 335 Feb 15 23:04 openclaw.json
-rw------- 1 axis axis 7625 Jan 31 23:10 openclaw.json.backup
-rw------- 1 axis axis 8426 Feb 2 15:58 openclaw.json.backup.20260202_155823
-rw------- 1 axis axis 4652 Feb 15 23:04 openclaw.json.bak
-rw------- 1 axis axis 4758 Feb 15 22:56 openclaw.json.bak.1
-rw------- 1 axis axis 4643 Feb 10 04:02 openclaw.json.bak.2
-rw------- 1 axis axis 4643 Feb 9 00:40 openclaw.json.bak.3
-rw------- 1 axis axis 4440 Feb 9 00:40 openclaw.json.bak.4
-rw------- 1 axis axis 18520 Jan 31 09:53 openclaw.json.save
-rw------- 1 axis axis 16391 Jan 31 11:45 openclaw.json.save.1
drwxrwxr-x 3 axis axis 4096 Feb 15 22:41 plugins
drwx------ 2 axis axis 4096 Jan 31 23:52 subagents
drwx------ 2 axis axis 4096 Feb 15 21:10 telegram
-rw-rw-r-- 1 axis axis 118 Feb 15 21:34 update-check.json
drwxrwxr-x 35 axis axis 4096 Feb 15 22:36 workspace

delicate flaxBOT Feb 15, 2026, 10:09 PM

#

Perfect — you have lots of backups. The one doctor just created is:

~/.openclaw/openclaw.json.bak (4652 bytes) ← this is the pre-wipe config

Restore it now

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.doctor-broken
cp -a ~/.openclaw/openclaw.json.bak ~/.openclaw/openclaw.json

Sanity check:

head -n 20 ~/.openclaw/openclaw.json

You should see way more than just wizard/commands/meta.

Then remove the broken kimi-claw plugin reference (so it won’t break again)

#

Find it:

grep -n "kimi" ~/.openclaw/openclaw.json

Edit:

nano ~/.openclaw/openclaw.json

Delete the line/block that references:
/home/axis/.openclaw/extensions/kimi-claw

Restart + test

openclaw gateway restart

#

Your openclaw.json is currently the tiny “wizard/meta only” file (335 bytes). The real config is in the backup ~/.openclaw/openclaw.json.bak.

Also: you don’t have rg (ripgrep) installed — no problem, we’ll use plain ls/grep.

1) Confirm the backup exists

ls -la ~/.openclaw/openclaw.json*

You should see openclaw.json.bak.

2) Restore it

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.doctor-broken
cp -a ~/.openclaw/openclaw.json.bak ~/.openclaw/openclaw.json

#

openclaw tui

If grep -n "kimi" ~/.openclaw/openclaw.json shows multiple hits (like plugins.entries too), paste those few lines and I’ll tell you exactly what to delete.

#

3) Remove the broken kimi-claw plugin line (so it doesn’t break again)

Find it:

grep -n "kimi-claw" ~/.openclaw/openclaw.json

Edit the file (use whichever editor you have):

nano ~/.openclaw/openclaw.json
# or: vi ~/.openclaw/openclaw.json

Delete the plugins.load.paths entry that points to:
/home/axis/.openclaw/extensions/kimi-claw
(or delete the whole "plugins": { ... } block if you’re not using plugins)

4) Restart + test

#

openclaw gateway restart
openclaw tui

If you paste the output of ls -la ~/.openclaw/openclaw.json* and grep -n "plugins" ~/.openclaw/openclaw.json I can tell you exactly what to delete without breaking the braces/commas.

cobalt summit Feb 15, 2026, 10:11 PM

#

axis@vmi3046347:~$ openclaw gateway restartopenclaw gateway restart
Failed to read config at /home/axis/.openclaw/openclaw.json SyntaxError: JSON5: invalid character ',' at 204:6
at syntaxError (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:1110:17)
at invalidChar (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:1055:12)
at Object.end (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:704:15)
at Object.default (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:168:37)
at lex (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js💯42)
at Object.parse (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:25:17)
at Object.loadConfig (file:///home/axis/.local/lib/node_modules/openclaw/dist/config-BJWIHf6x.js:5193:111)
at loadConfig (file:///home/axis/.local/lib/node_modules/openclaw/dist/config-BJWIHf6x.js:5576:20)
at file:///home/axis/.local/lib/node_modules/openclaw/dist/context-DPQWp4WP.js:10:34 {
lineNumber: 204,
columnNumber: 6

delicate flaxBOT Feb 15, 2026, 10:12 PM

#

That error is just a syntax problem in ~/.openclaw/openclaw.json (extra/misplaced comma). Until the file parses as JSON5, nothing can start.

1) Show the broken area (line ~204)

Run:

nl -ba ~/.openclaw/openclaw.json | sed -n '195,215p'

Paste those lines here (feel free to redact any tokens/secrets).

2) Fix it quickly in an editor (jump to the line)

If you have nano:

nano +204 ~/.openclaw/openclaw.json

#

Go to line 204 and look for something like:

a line that is just ,
},,
],,
a comma right after { like {,

Delete the extra comma, save, exit.

3) Restart

openclaw gateway restart

If you just want a fast restore to a different backup

You have multiple backups (openclaw.json.bak.1, .bak.2, etc). If the .bak you restored is corrupted, try the most recent older one:

#

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bad-syntax
cp -a ~/.openclaw/openclaw.json.bak.1 ~/.openclaw/openclaw.json
openclaw gateway restart

(And if that one also errors, try .bak.2, .bak.3, etc.)

Paste the sed -n '195,215p' output and I’ll tell you exactly what character to remove.

cobalt summit Feb 15, 2026, 10:14 PM

#

axis@vmi3046347:~$ ^C
axis@vmi3046347:~$ nl -ba ~/.openclaw/openclaw.json | sed -n '195,215p'
195 "install": {
196 "nodeManager": "npm"
197 },
198 "entries": {
199 "nano-banana-pro": {
200 "apiKey": "dont mind this its just so i dont share my apik"
201 }
202 }
203 },
204 },
205 "entries": {
206 "telegram": {
207 "enabled": true
208 }
209 }
210 }
211 }

delicate flaxBOT Feb 15, 2026, 10:16 PM

#

That snippet shows the exact problem.

Fix the JSON5 syntax error

Line 204 is a stray }, that doesn’t belong there.

Do this:

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.pre-fix
nano +195 ~/.openclaw/openclaw.json

In nano, go to line 204 (you can use Ctrl+_ then type 204) and delete this whole line:

},

#

Also: your Telegram config is in the wrong place

Lines 205–210:

        "entries": {
          "telegram": { "enabled": true }
        }

…should not be under plugins.entries (that’s for plugin ids). Telegram belongs under:

channels.telegram.enabled (and channels.telegram.botToken, etc.)

So for now, delete that whole telegram “entries” block from the plugins section (lines 205–210), save, exit.

Then restart

#

openclaw gateway restart
openclaw tui

If it complains next about Telegram missing/disabled, we’ll add it back under channels.telegram in the right spot (don’t paste your bot token here).

cobalt summit Feb 15, 2026, 10:18 PM

#

idk i think i fucked it up because its showing this after : onfig invalid
File: ~/.openclaw/openclaw.json
Problem:

<root>: JSON5 parse failed: SyntaxError: JSON5: invalid character '}' at 210:1

Run: openclaw doctor --fix
Restarted systemd service: openclaw-gateway.service
Failed to read config at /home/axis/.openclaw/openclaw.json SyntaxError: JSON5: invalid character '}' at 210:1
at syntaxError (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:1110:17)
at invalidChar (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:1055:12)
at Object.end (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:704:15)
at Object.default (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:168:37)
at lex (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js💯42)
at Object.parse (/home/axis/.local/lib/node_modules/openclaw/node_modules/json5/lib/parse.js:25:17)
at Object.loadConfig (file:///home/axis/.local/lib/node_modules/openclaw/dist/config-BJWIHf6x.js:5193:111)
at loadConfig (file:///home/axis/.local/lib/node_modules/openclaw/dist/config-BJWIHf6x.js:5576:20)
at file:///home/axis/.local/lib/node_modules/openclaw/dist/context-DPQWp4WP.js:10:34 {
lineNumber: 210,
columnNumber: 1
}

🦞 OpenClaw 2026.2.13 (203b5bd) — curl for conversations.

│
◇ Config ───────────────────────────────────────────────────╮
│ │
│ Config invalid; doctor will run with best-effort config. │
│ │
├────────────────────────────────────────────────────────────╯
Config invalid
File: ~/.openclaw/openclaw.json
Problem:

<root>: JSON5 parse failed: SyntaxError: JSON5: invalid character '}' at 210:1

Run: openclaw doctor --fix

delicate flaxBOT Feb 15, 2026, 10:19 PM

#

You didn’t lose your sessions/credentials — you just made ~/.openclaw/openclaw.json syntactically broken (extra/mismatched }), so OpenClaw can’t even parse the config now.

Fastest recovery: restore a known-good backup again, then redo the “remove kimi-claw path” edit carefully.

1) Restore from a clean backup (pick the newest that parses)

Run:

cp -a ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bad-parse
cp -a ~/.openclaw/openclaw.json.bak.1 ~/.openclaw/openclaw.json

Now test parsing:

openclaw doctor

#

If it still says JSON5 parse failed, try the next backup:

cp -a ~/.openclaw/openclaw.json.bak.2 ~/.openclaw/openclaw.json
openclaw doctor

(keep going with .bak.3, .bak.4, or even openclaw.json.save if needed)

2) Once it parses again: remove the one bad kimi-claw path

After you’re back to a parseable config, do:

grep -n "kimi-claw" ~/.openclaw/openclaw.json
nano ~/.openclaw/openclaw.json

#

…and delete only the entry that points to:
/home/axis/.openclaw/extensions/kimi-claw

Then:

openclaw gateway restart
openclaw tui

If you paste the output of:

grep -n "kimi-claw\\|plugins\"\\s*:" ~/.openclaw/openclaw.json

(just those lines, no tokens), I’ll tell you exactly what to delete so braces/commas stay valid.

cobalt summit Feb 15, 2026, 10:23 PM

#

d: 5 │
│ Disabled: 31 │
│ Errors: 0 │
│ │
├────────────────╯
│
◇ Plugin diagnostics ────────────────────────────────────────────────────────╮
│ │
│ - ERROR: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw │
│ (/home/axis/.openclaw/extensions/kimi-claw) │
│ │
├─────────────────────────────────────────────────────────────────────────────╯
Invalid config at /home/axis/.openclaw/openclaw.json:\n- plugins.load.paths: plugin: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw
│
│ │
│ Gateway target: ws://127.0.0.1:18789 │
│ Source: local loopback │
│ Config: /home/axis/.openclaw/openclaw.json │
│ Bind: loopback │
│ │
├──────────────────────────────────────────────╯
│
◇ Gateway ──────────────────────────────────────────────────────────────────────╮
│ │
│ Runtime: stopped (state activating, sub auto-restart, last exit 1, reason 1) │
│ Service is loaded but not running (likely exited immediately). │
│ File logs: /tmp/openclaw/openclaw-2026-02-15.log │
│ Logs: journalctl --user -u openclaw-gateway.service -n 200 --no-pager │
│ │
├────────────────────────────────────────────────────────────────────────────────╯
│
◇ Start gateway service now?
│ Yes
Restarted systemd service: openclaw-gateway.service
Run "openclaw doctor --fix" to apply changes.
Invalid config:

plugins.load.paths: plugin: plugin path not found: /home/axis/.openclaw/extensions/kimi-claw
│
└ Doctor complete.

delicate flaxBOT Feb 15, 2026, 10:23 PM

#

That’s back to the original issue (and it’s good news): your config parses again, but it’s invalid because it still references the missing kimi-claw plugin path.

You need to remove this from ~/.openclaw/openclaw.json:

/home/axis/.openclaw/extensions/kimi-claw

Remove the bad plugin path (no ripgrep needed)

Show where it is:

grep -n "kimi-claw" ~/.openclaw/openclaw.json

Edit the config:

nano ~/.openclaw/openclaw.json

#

Find this block and delete just the kimi-claw entry (or delete the whole load.paths block if it only contains that one path):

plugins: {
  load: {
    paths: [
      "/home/axis/.openclaw/extensions/kimi-claw",
    ],
  },
}

If that paths: [...] array becomes empty, you can leave it as:

paths: []

(or remove load entirely).

#

Restart + verify:

openclaw gateway restart
openclaw doctor
openclaw tui

If after removing it you get a new plugins error like plugins.entries: unknown plugin id ..., paste that exact error text and we’ll remove that reference too.

cobalt summit Feb 15, 2026, 10:24 PM

#

}
}
},
"plugins": {
"load": {
"paths": [
"/home/axis/.openclaw/extensions/kimi-claw"
]
},
"entries": {
"telegram": {
"enabled": true
}
}
}

delicate flaxBOT Feb 15, 2026, 10:24 PM

#

Cool — that’s the exact block we need to fix.

Edit it to this (minimal + correct)

Option A (recommended): remove plugins entirely
If you’re not intentionally using plugins, delete the whole "plugins": { ... } section, so you end up with just:

    }
  }

(i.e., no plugins key at all)

Option B: keep plugins enabled but remove the broken path
Replace your block with:

#

  "plugins": {
    "load": {
      "paths": []
    }
  }

Important: delete the `"telegram"` entry under plugins

This part is wrong and will bite you later:

"entries": { "telegram": { "enabled": true } }

Telegram is a channel, not a plugin. If you want Telegram enabled, it should be under channels.telegram.enabled instead.

#

After editing

openclaw gateway restart
openclaw doctor
openclaw tui

If it complains that Telegram is disabled/missing, paste just the channels section (without the bot token) and I’ll point you to the right key.

cobalt summit Feb 15, 2026, 10:38 PM

#

so how do io make kimiclaw work bro bash <(curl -fsSL https://cdn.kimi.com/kimi-claw/install.sh) --bot-token

delicate flaxBOT Feb 15, 2026, 10:39 PM

#

That kimi-claw curl|bash installer isn’t needed to use Kimi with OpenClaw, and in your case it just left you with a config that points at a plugin folder that doesn’t exist (~/.openclaw/extensions/kimi-claw) → which hard-breaks the gateway.

You’ve basically got two choices:

Option 1 (recommended): Use Kimi the supported way (no plugin)

OpenClaw already has a Moonshot (Kimi) provider built-in.

Do this on the VPS (don’t paste your key into Discord):

openclaw onboard --auth-choice moonshot-api-key

Then pick Moonshot/Kimi in the prompts, restart gateway, and you’re done.

Docs: https://docs.openclaw.ai/providers/moonshot

#

Option 2: You insist on “kimi-claw” plugin/script

Then you must make this path actually exist and contain a valid OpenClaw plugin:

/home/axis/.openclaw/extensions/kimi-claw/
and inside it there must be openclaw.plugin.json

Right now it’s missing, so OpenClaw will always fail config validation.

Check what’s there:

ls -la ~/.openclaw/extensions
ls -la ~/.openclaw/extensions/kimi-claw 2>/dev/null || echo "kimi-claw dir missing"

If it’s missing, you’d have to reinstall that plugin from Kimi’s own docs/support (I can’t vouch for or debug a third-party curl|bash installer).

#

Also: rotate that key

You posted a sk-... token earlier in a public thread — assume it’s compromised and rotate/revoke it immediately.

If you tell me which you actually want (Moonshot/Kimi provider vs that plugin), I’ll give you the exact clean config steps for that path.

cobalt summit Feb 17, 2026, 3:49 AM

#

Last login: Tue Feb 17 04:43:46 2026 from 209.35.167.169
axis@vmi3046347:~$ openclaw tui

🦞 OpenClaw 2026.2.15 (3fe22ea) — Say "stop" and I'll stop—say "ship" and we'll both learn a lesson.

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
connecting | idle
gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | idle
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
how to fix pls

#

@delicate flax

delicate flaxBOT Feb 17, 2026, 3:53 AM

#

That error means your CLI/TUI is presenting a stale device token (cached locally), but the Gateway now expects a different one (often after a restore, reinstall, or token rotation).

Fix (usually 1 command)

Run TUI once using the Gateway auth token (NOT the device token). Don’t paste tokens into Discord.

openclaw tui --url ws://127.0.0.1:18789 --token <gateway.auth.token>

If that connects, it should automatically re-issue + persist a fresh device token, and then plain openclaw tui will work again.

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

If you don’t know the gateway token

On the gateway host, retrieve it from config (locally):

#

openclaw config get gateway.auth.token

If it still fails: rotate the device token

openclaw devices list --url ws://127.0.0.1:18789 --token <gateway.auth.token>
openclaw devices rotate --url ws://127.0.0.1:18789 --token <gateway.auth.token> --device <deviceId> --role operator

Docs: https://docs.openclaw.ai/cli/devices

Last resort: clear the local cached device token and re-connect

This only affects the machine you run it on:

mv ~/.openclaw/identity/device-auth.json ~/.openclaw/identity/device-auth.json.bak

#

# then run the tui command again with --token <gateway.auth.token>

If you try the “Fix (usually 1 command)” step and it still says device token mismatch, paste (redacting secrets) the output of:

openclaw gateway status

#

I’m here — did you try connecting once with the gateway auth token (not the cached device token)?

On the gateway host, get the token:

openclaw config get gateway.auth.token

Then run:

openclaw tui --url ws://127.0.0.1:18789 --token <that-token>

If that works, it should refresh your cached device token and plain openclaw tui should stop erroring.

If it still says device token mismatch, run this and paste the output (redact any tokens/secrets):

#

openclaw gateway status
openclaw devices list

cobalt summit Feb 17, 2026, 3:54 AM

#

Last login: Tue Feb 17 04:48:52 2026 from 112.198.104.232
axis@vmi3046347:~$ openclaw tui --url ws://127.0.0.1:18789 --token <gateway.auth.token>
-bash: syntax error near unexpected token `newline'

delicate flaxBOT Feb 17, 2026, 3:55 AM

#

Yeah — that happened because you pasted my placeholder literally. In bash, the < character is input redirection, so --token <...> blows up.

Do this instead:

Print the actual gateway token (don’t paste it into Discord):

openclaw config get gateway.auth.token

Copy that output and run without any < >:

openclaw tui --url ws://127.0.0.1:18789 --token "PASTE_THE_TOKEN_HERE"

If config get prints the token with quotes already, keep them (or just wrap in "..." like above).

If you paste what openclaw config get gateway.auth.token prints but redact most of it (e.g. keep first/last 4 chars), I can tell you if it looks formatted right.

cobalt summit Feb 17, 2026, 3:57 AM

#

its connected now thank u, is this ok now

delicate flaxBOT Feb 17, 2026, 3:57 AM

#

Yep — that’s exactly what we wanted. Once you connect successfully with the gateway auth token, OpenClaw re-issues/saves a fresh device token, and the reconnect loop stops.

Quick sanity check:

Quit the TUI and run just:

openclaw tui

If it connects without --token, you’re fully good.

If it helped, please right-click my solution message → Apps → ✅ Mark Solution so Answer Overflow indexes it for the next person.

cobalt summit Feb 17, 2026, 3:34 PM

#

connecting | idle
gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | idle
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)

it did this

#

@delicate flax

delicate flaxBOT Feb 17, 2026, 3:38 PM

#

That message means your client is trying to authenticate with a stale device token (device pairing token), so the Gateway rejects it.

This usually happens after switching Gateways, reinstalling/wiping Gateway state, or rotating/revoking device tokens.

Two quick fixes (depending on where you’re seeing it):

If this is the TUI (`openclaw tui`)

Run the TUI once with the Gateway token explicitly (this overrides any cached device token and will re-issue a fresh one):

openclaw tui --url ws://<host>:18789 --token <gateway-token>

If it then says “pairing required”, approve it on the gateway host:

openclaw devices list
openclaw devices approve <requestId>

#

If this is the browser Control UI

Open the dashboard URL from openclaw gateway status (or run openclaw dashboard) and paste the gateway token in the Control UI settings again.
If it keeps looping, clear the site data for that dashboard URL (or try an Incognito window) to force a fresh device pairing, then approve via openclaw devices approve ….

Docs:

To pinpoint which path you’re on: is this happening in openclaw tui or the web Control UI? And what does openclaw --version show (on the machine you’re running that from, and on the gateway host if different)?

#

That output looks like the TUI (openclaw tui) reconnect loop.

It means the TUI is sending a token the gateway doesn’t accept anymore (often a cached device token, sometimes it’s actually the gateway token that changed — the error message is a bit misleading when a device id is present).

Try this in order:

Force the TUI to use the gateway token once (this should mint/store a fresh device token):

openclaw tui --url ws://<gateway-host>:18789 --token <gateway-token>

If you don’t know the correct gateway token, get it on the gateway host (or wherever the gateway config lives) and make sure your client token matches gateway.auth.token.
If it still keeps failing, delete the cached device token on the machine running the TUI and reconnect:

rm -f ~/.openclaw/identity/device-auth.json

#

(then rerun step 1)

Last resort (if you have access on the gateway host): rotate/revoke the device token and re-pair:

openclaw devices list
openclaw devices rotate --device <deviceId> --role operator
# or: openclaw devices revoke --device <deviceId> --role operator

If you paste:

openclaw --version
how you started the TUI (exact command)
and whether you recently reinstalled/changed machines/rotated tokens

…I can tell which of the above is the right fix.

cobalt summit Feb 17, 2026, 4:09 PM

#

Sounds good — standing by!

reset failed: Error: invalid sessions.reset params: at root: unexpected property 'reason'

delicate flaxBOT Feb 17, 2026, 4:11 PM

#

That’s actually a different issue now (and a good clue): you’re connected, but your client and gateway are on different versions.

sessions.reset in your TUI is sending { key, reason }, but the Gateway you’re connected to is older and its schema rejects the extra field:

invalid sessions.reset params: … unexpected property 'reason'

Fix (recommended): align versions

Run these on the machine running the TUI and on the gateway host and paste the outputs (no tokens):

openclaw --version
openclaw gateway status

Then update whichever side is older so they match, and restart the gateway.

Workaround (right now)

If you just wanted a clean slate, you can avoid reset entirely by switching to a new session key in the TUI:

/session scratch (or any new name)

#

That gives you a fresh session even while reset is broken.

(And don’t paste any gateway tokens in here—just versions/status.)

cobalt summit Feb 17, 2026, 4:12 PM

#

axis@vmi3046347:~$ openclaw --version
openclaw gateway status
2026.2.15

🦞 OpenClaw 2026.2.15 (3fe22ea) — Your inbox, your infra, your rules.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw-1000/openclaw-2026-02-17.log
Command: /usr/bin/node /home/axis/.local/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 1526200, state active, sub running, last exit 0, reason 0)
RPC probe: ok

Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

delicate flaxBOT Feb 17, 2026, 4:14 PM

#

You’re on 2026.2.15 and the gateway is healthy (RPC probe: ok) — nice.

That sessions.reset error:

unexpected property 'reason'

means your client is calling sessions.reset with { key, reason }, but the Gateway you’re connected to is validating against a schema that doesn’t allow reason. In practice this happens when:

the TUI and Gateway are on different OpenClaw versions, or
there’s a protocol/schema mismatch bug in that build.

Quick workaround (no reset needed)

In the TUI, just start a fresh session by switching to a new session key:

/session scratch (or any new name)

That gives you a clean slate immediately.

#

Real fix

Run openclaw --version on the machine where you run openclaw tui (might be different from the gateway host).
Make sure both sides match, ideally by updating to the latest OpenClaw on both, then restart the gateway:

openclaw gateway restart

If after updating you still see unexpected property 'reason', paste the TUI-side openclaw --version and we’ll nail down whether you’re accidentally connecting to a different gateway URL than you think.

cobalt summit Feb 17, 2026, 4:16 PM

#

reset failed: Error: invalid sessions.reset params: at root: unexpected property 'reason'

🦞 OpenClaw 2026.2.15 (3fe22ea) — Your inbox, your infra, your rules.

│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw-1000/openclaw-2026-02-17.log
Command: /usr/bin/node /home/axis/.local/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 1526200, state active, sub running, last exit 0, reason 0)
RPC probe: ok

Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting
axis@vmi3046347:~$ openclaw --version
2026.2.15
axis@vmi3046347:~$ openclaw gateway restart

Restarted systemd service: openclaw-gateway.service
axis@vmi3046347:~$ openclaw tui

🦞 OpenClaw 2026.2.15 (3fe22ea) — I read logs so you can keep pretending you don't have to.

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
gateway disconnected: closed | idle

delicate flaxBOT Feb 17, 2026, 4:20 PM

#

Ok — so version mismatch isn’t it (both CLI + gateway are 2026.2.15).

1) `/reset` failing with `unexpected property 'reason'`

That points to a bug/protocol mismatch inside 2026.2.15 (the TUI sends {key, reason}, but the gateway validator on your side rejects reason).

Workarounds:

In the TUI, just start fresh via a new session key:
- /session scratch (or any new name)

Or reset from the shell without reason:

openclaw gateway call sessions.reset --params '{"key":"main"}'

2) `openclaw tui` immediately → `gateway disconnected: closed`

A blank close-reason usually means “server dropped the socket” and the why will be in the gateway logs.

#

Run this, then start openclaw tui in a second terminal and paste the relevant lines:

openclaw logs --follow --plain

One more important note: your gateway is bind=loopback, so only clients on the VPS itself can connect. If you were trying to run the TUI from your laptop, it’ll just disconnect until you expose the gateway (tailnet/serve/ssh tunnel).

cobalt summit Feb 17, 2026, 4:24 PM

#

│
◇ Gateway connection ─────────────────────────╮
│ │
│ Gateway target: ws://127.0.0.1:18789 │
│ Source: local loopback │
│ Config: /home/axis/.openclaw/openclaw.json │
│ Bind: loopback │
│ │
├──────────────────────────────────────────────╯
Config overwrite: /home/axis/.openclaw/openclaw.json (sha256 2951bc2e12c1609d4009c0c16f0255adeed46b0b2c6930ee5f6079795dc14cc9 -> fe96944fb77b8f6b3e669425f70a47f88eb0101bfaa5959f2a3f9cee6b0650ab, backup=/home/axis/.openclaw/openclaw.json.bak)
Updated ~/.openclaw/openclaw.json
Backup: ~/.openclaw/openclaw.json.bak
│
└ Doctor complete.

axis@vmi3046347:~$ openclaw tui

🦞 OpenClaw 2026.2.15 (3fe22ea) — Pairing codes exist because even bots believe in consent—and good security hygiene.

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
gateway disconnected: closed | idle
agent main | session main | unknown | tokens ?

how do i fix this

#

@delicate flax

#

🦞 OpenClaw 2026.2.15 (3fe22ea) — Because texting yourself reminders is so 2024.

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
gateway disconnected: closed | idle
gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | idle
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)───────────────

delicate flaxBOT Feb 17, 2026, 4:30 PM

#

gateway disconnected: closed (with an empty reason) usually means the socket got dropped without a proper close reason — most often because the gateway process crashed/restarted right when the TUI tried to connect (or hit an unhandled exception during handshake).

The next step is: look at the gateway logs at the moment you run openclaw tui.

1) Grab the log lines around the disconnect

In one terminal, run:

tail -n 200 /tmp/openclaw-1000/openclaw-2026-02-17.log

Then run openclaw tui again, and immediately run:

tail -n 200 /tmp/openclaw-1000/openclaw-2026-02-17.log

Paste the new lines that appeared (they’ll usually say exactly why it closed: unauthorized / invalid handshake / stack trace / etc).

#

That’s the smoking gun:

unauthorized: device token mismatch

Your TUI has a cached device token that no longer matches what the gateway has for that device (common after doctor/migrations, resets, or token rotation).

Fix (fast)

On the same machine where you run openclaw tui (your VPS), delete the cached device token and reconnect:

rm -f ~/.openclaw/identity/device-auth.json
openclaw gateway restart
openclaw tui

If it still says device token mismatch

Nuke the device identity too (this forces a brand new device ID; on loopback it should auto-approve):

#

rm -f ~/.openclaw/identity/device.json ~/.openclaw/identity/device-auth.json
openclaw gateway restart
openclaw tui

Side note: your `/reset` error

That sessions.reset … unexpected property 'reason' is a separate bug in this build. Workaround in TUI: use a fresh session key instead:

/session scratch

If you try the token-cache delete and it still fails, paste the few new lines from:

tail -n 80 /tmp/openclaw-1000/openclaw-2026-02-17.log

right after the disconnect.

#

Also check if systemd restarted the gateway:

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

2) Separate issue: `/reset` is broken in 2026.2.15

That sessions.reset … unexpected property 'reason' is a real bug in that release.

Workarounds:

Easiest: start a fresh session in the TUI:
- /session scratch

Or from shell (no reason field):

openclaw gateway call sessions.reset --params '{"key":"main"}'

3) Likely “real fix”

cobalt summit Feb 17, 2026, 4:31 PM

#

disconnected | press ctrl+c again to exit
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)

axis@vmi3046347:~$ rm -f ~/.openclaw/identity/device-auth.json
openclaw gateway restart
openclaw tui

🦞 OpenClaw 2026.2.15 (3fe22ea) — Pairing codes exist because even bots believe in consent—and good security hygiene.

Restarted systemd service: openclaw-gateway.service

🦞 OpenClaw 2026.2.15 (3fe22ea) — If something's on fire, I can't extinguish it—but I can write a beautiful postmortem.

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
gateway disconnected: closed | idle
agent main | session main | unknown | tokens ?

delicate flaxBOT Feb 17, 2026, 4:32 PM

#

Once we confirm the log error, it’s probably “update to the next patch release” (this smells like a 2026.2.15 protocol/schema mismatch).

Send the log snippet and I’ll tell you the exact fix (and whether it’s an update vs a one-line config/state cleanup).

#

That error is almost always: your TUI is sending the wrong auth.token (so shared-secret auth fails), and then the gateway tries to interpret that token as a device token → mismatch.

On a VPS this commonly happens because:

your shell has OPENCLAW_GATEWAY_TOKEN set (maybe an old value or even a device token), but
the systemd gateway service does not have that env var, so it’s using the token from ~/.openclaw/openclaw.json (or none).

So the TUI and the gateway end up using different tokens.

Fix

In the same terminal you run openclaw tui, clear any gateway auth env vars:

unset OPENCLAW_GATEWAY_TOKEN OPENCLAW_GATEWAY_PASSWORD
openclaw tui

If you want to be extra safe without changing your shell permanently:

#

OPENCLAW_GATEWAY_TOKEN= OPENCLAW_GATEWAY_PASSWORD= openclaw tui

If it still fails, tell me what auth mode you’re in (this is safe to paste):

openclaw config get gateway.auth.mode

(Do not paste the actual token/password.)

If you run those and it still says device token mismatch, paste the newest few lines from:

tail -n 80 /tmp/openclaw-1000/openclaw-2026-02-17.log

right after attempting openclaw tui.

cobalt summit Feb 18, 2026, 2:10 AM

#

gateway disconnected: closed | idle
gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | press ctrl+c again to exit

axis@vmi3046347:~$ [200~tail -n 200 /tmp/openclaw-1000/openclaw-2026-02-17.log~
[200~tail: command not found
axis@vmi3046347:~$ openclaw tui

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
connecting | idle
gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | idle

gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)

axis@vmi3046347:~$ Read from remote host 217.216.33.14: Operation timed out

what do i do please help me fix

#

@delicate flax

delicate flaxBOT Feb 18, 2026, 2:15 AM

#

You’re super close — the annoying part is that “device token mismatch” is often a misleading error here. It usually means:

the TUI is sending a token, but
the gateway is expecting a different gateway auth token (or is getting it from a different place like ~/.openclaw/.env / systemd env),
then the gateway falls back to trying that token as a device token and says “device token mismatch”.

Also: that tail: command not found happened because you pasted while still in a mode that injected "[200~" (bracketed paste). Make sure you’re at a normal shell prompt.

Step 1 — Force the TUI to use the token from your config (no pasting secrets)

SSH back in, then run:

unset OPENCLAW_GATEWAY_TOKEN OPENCLAW_GATEWAY_PASSWORD
OPENCLAW_GATEWAY_TOKEN="$(openclaw config get gateway.auth.token)" openclaw tui

That makes the TUI use exactly what ~/.openclaw/openclaw.json says.

#

Step 2 — If it STILL says device token mismatch, your gateway service is using a different token via env

Check (this does not print the token value):

test -f ~/.openclaw/.env && grep -q '^OPENCLAW_GATEWAY_TOKEN=' ~/.openclaw/.env && echo ".openclaw/.env sets OPENCLAW_GATEWAY_TOKEN"
systemctl --user show openclaw-gateway.service -p Environment | grep -q OPENCLAW_GATEWAY_TOKEN && echo "systemd service has OPENCLAW_GATEWAY_TOKEN set"

If either line prints, that’s your mismatch. Fix is to make them consistent (pick ONE source of truth):

simplest: remove the OPENCLAW_GATEWAY_TOKEN=... line from ~/.openclaw/.env (if present), then:
```
openclaw gateway restart
```
and retry the Step 1 command.

Step 3 — Your SSH “Operation timed out”

#

While debugging, run inside tmux so you don’t lose the session when the network blips:

tmux new -s claw

If you reply with just the output of:

openclaw config get gateway.auth.mode

and whether those two “sets OPENCLAW_GATEWAY_TOKEN” checks print anything, I can tell you exactly which branch to take.

#

You’re getting kicked because the TUI is still sending a token that the gateway doesn’t accept (and the gateway reports it as “device token mismatch”).

Do this on the VPS (as user axis), ideally inside tmux so your SSH doesn’t time out:

0) Don’t lose your session

tmux new -s fix

1) Remove any shell env overrides (most common cause)

env | grep OPENCLAW_GATEWAY || true
unset OPENCLAW_GATEWAY_TOKEN OPENCLAW_GATEWAY_PASSWORD
openclaw tui

2) Force the TUI to use the token from your config (without printing it)

#

(Do not paste the token into chat.)

OPENCLAW_GATEWAY_PASSWORD= OPENCLAW_GATEWAY_TOKEN="$(openclaw config get gateway.auth.token)" openclaw tui

3) If it STILL says “device token mismatch”, your gateway service is probably using a different token via systemd env or `~/.openclaw/.env`

Run these (they redact the token):

systemctl --user show openclaw-gateway.service -p Environment \
  | sed -E 's/(OPENCLAW_GATEWAY_TOKEN=)[^ ]+/\\1<redacted>/g'

sed -n 's/^OPENCLAW_GATEWAY_TOKEN=.*/OPENCLAW_GATEWAY_TOKEN=<redacted>/p' ~/.openclaw/.env

If either of those prints a line, remove that OPENCLAW_GATEWAY_TOKEN=... override (so the gateway uses ~/.openclaw/openclaw.json), then:

#

openclaw gateway restart
openclaw tui

If you paste the output of Step 3 (the redacted lines), I can tell you exactly which place is overriding your token and what to edit.

cobalt summit Feb 18, 2026, 2:17 AM

#

test -f ~/.openclaw/.env && grep -q '^OPENCLAW_GATEWAY_TOKEN=' ~/.openclaw/.env && echo ".openclaw/.env sets OPENCLAW_GATEWAY_TOKEN"
systemctl --user show openclaw-gateway.service -p Environment | grep -q OPENCLAW_GATEWAY_TOKEN && echo "systemd service has OPENCLAW_GATEWAY_TOKEN set"

when i did this the gateway was connected is this fixed already? or is there still more that needs to be done so that we can fix this properly with no more errors let me know please

delicate flaxBOT Feb 18, 2026, 2:20 AM

#

Those two commands are just checks — they don’t change anything by themselves.

How to tell if it’s actually “fixed”

Did either command print a line?

If you saw no output, that’s good: no token override from ~/.openclaw/.env and none baked into the systemd unit env.

The real test: run the TUI and see if it stays connected:

openclaw tui

If it sits on connected | idle (and you can run /status) for ~30–60 seconds, you’re good.

The most common remaining “gotcha”

Even if .env and systemd are clean, your current shell might still have an old token exported, which makes the TUI send the wrong token and triggers the misleading “device token mismatch”.

Check:

#

env | grep OPENCLAW_GATEWAY || true

If you see OPENCLAW_GATEWAY_TOKEN=..., do:

unset OPENCLAW_GATEWAY_TOKEN OPENCLAW_GATEWAY_PASSWORD
openclaw tui

If you want it “fixed properly” long-term

If OPENCLAW_GATEWAY_TOKEN is showing up in env, remove it from your shell startup files (e.g. ~/.bashrc, ~/.profile, ~/.zshrc) so it won’t come back.

If you tell me whether those two check commands printed anything, and what env | grep OPENCLAW_GATEWAY shows (you can redact values), I’ll tell you exactly what to clean up.

cobalt summit Feb 18, 2026, 4:03 AM

#

gateway disconnected: unauthorized: device token mismatch (rotate/reissue device token) | press ctrl+c again to exit
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)
gateway connect failed: Error: unauthorized: device token mismatch (rotate/reissue device token)

axis@vmi3046347:~$ systemctl --user show openclaw-gateway.service -p Environment
| sed -E 's/(OPENCLAW_GATEWAY_TOKEN=)[^ ]+/\1<redacted>/g'

sed -n 's/^OPENCLAW_GATEWAY_TOKEN=.*/OPENCLAW_GATEWAY_TOKEN=<redacted>/p' ~/.openclaw/.env
Environment=HOME=/home/axis PATH=/home/axis/.local/bin:/home/axis/.npm-global/bin:/home/axis/bin:/home/axis/.nvm/current/bin:/home/axis/.fnm/current/bin:/home/axis/.volta/bin:/home/axis/.asdf/shims:/home/axis/.local/share/pnpm:/home/axis/.bun/bin:/usr/local/bin:/usr/bin:/bin OPENCLAW_GATEWAY_PORT=18789 \1<redacted> OPENCLAW_SYSTEMD_UNIT=openclaw-gateway.service OPENCLAW_SERVICE_MARKER=openclaw OPENCLAW_SERVICE_KIND=gateway OPENCLAW_SERVICE_VERSION=2026.2.6-3
OPENCLAW_GATEWAY_TOKEN=<redacted>
axis@vmi3046347:~$ bash
axis@vmi3046347:~$ openclaw gateway restart
openclaw tui

🦞 OpenClaw 2026.2.15 (3fe22ea) — I can't fix your code taste, but I can fix your build and your backlog.

Restarted systemd service: openclaw-gateway.service

🦞 OpenClaw 2026.2.15 (3fe22ea) — It's not "failing," it's "discovering new ways to configure the same thing wrong."

openclaw tui - ws://127.0.0.1:18789 - agent main - session main
gateway disconnected: closed | idle

#

@delicate flax

#

@delicate flax please help me

#

@delicate flax

delicate flaxBOT Feb 18, 2026, 4:18 AM

#

That error is not about OPENCLAW_GATEWAY_TOKEN — it means a paired device is trying to auth with an old/stale per-device token, and the Gateway’s record doesn’t match anymore (device pairing state lives in $OPENCLAW_STATE_DIR/devices/paired.json).

The usual fixes are:

See which device is failing

openclaw devices list --json

This output is safe to paste here (it doesn’t print the raw tokens), and it will show the deviceId, role (operator = dashboard/CLI, node = iOS/node), and last-used timestamps.

Rotate / reissue the token for that device+role

openclaw devices rotate --device <deviceId> --role operator
# or: --role node

Then reconnect/restart the client that’s spamming “device token mismatch”.

#

If you recently changed profiles / state dirs: make sure the systemd service and your CLI are using the same state dir (profile mismatch is a super common cause of “token mismatch” weirdness). If needed, reinstall service metadata from the same profile/state dir:

openclaw gateway install --force
openclaw gateway restart

Docs:

https://docs.openclaw.ai/cli/devices
https://docs.openclaw.ai/gateway/troubleshooting (section “Pairing and device identity state changed”)

If you paste openclaw devices list --json and tell me which client is disconnecting (Dashboard in browser vs openclaw tui vs an iOS/node), I can tell you exactly which deviceId/role to rotate.

#

Nira — that specific error usually means your CLI/dashboard is trying to log in with a cached device token that no longer matches what the Gateway has on record.

Two quick things to do:

1) Check the paired devices (safe to paste)

Run on the gateway host:

openclaw devices list --json

Paste the output here (it won’t include the raw tokens). That tells us which device/role is failing.

2) Clear the client-side cached device token (often fixes immediately)

The CLI caches the last issued device token here:

$OPENCLAW_STATE_DIR/identity/device-auth.json (default: ~/.openclaw/identity/device-auth.json)

Do a safe “move aside” (not delete), then reconnect:

#gateway disconnected after kimi claw

Common fixes (depending on what the above shows)

Fix (recommended): remove the broken plugin path from config

If you wanted that plugin

Important security note

Do this:

Fastest “get me unstuck” fix

How to confirm

1) Check if your config still has your real settings

2) Restore from a backup (most likely fix)

3) If there’s no backup

Restore your old config

After it’s back up

Restore it now

Then remove the broken kimi-claw plugin reference (so it won’t break again)

Restart + test

1) Confirm the backup exists

2) Restore it

3) Remove the broken kimi-claw plugin line (so it doesn’t break again)

4) Restart + test

1) Show the broken area (line ~204)

2) Fix it quickly in an editor (jump to the line)

3) Restart

If you just want a fast restore to a different backup

Fix the JSON5 syntax error

Also: your Telegram config is in the wrong place

Then restart

1) Restore from a clean backup (pick the newest that parses)

2) Once it parses again: remove the one bad kimi-claw path

Remove the bad plugin path (no ripgrep needed)

Edit it to this (minimal + correct)

Important: delete the "telegram" entry under plugins

After editing

Option 1 (recommended): Use Kimi the supported way (no plugin)

Option 2: You insist on “kimi-claw” plugin/script

Also: rotate that key

Fix (usually 1 command)

If you don’t know the gateway token

If it still fails: rotate the device token

Last resort: clear the local cached device token and re-connect

If this is the TUI (openclaw tui)

If this is the browser Control UI

Fix (recommended): align versions

Workaround (right now)

Quick workaround (no reset needed)

Real fix

1) /reset failing with unexpected property 'reason'

2) openclaw tui immediately → gateway disconnected: closed

1) Grab the log lines around the disconnect

Fix (fast)

If it still says device token mismatch

Side note: your /reset error

2) Separate issue: /reset is broken in 2026.2.15

3) Likely “real fix”

Fix

Step 1 — Force the TUI to use the token from your config (no pasting secrets)

Step 2 — If it STILL says device token mismatch, your gateway service is using a different token via env

Step 3 — Your SSH “Operation timed out”

0) Don’t lose your session

1) Remove any shell env overrides (most common cause)

2) Force the TUI to use the token from your config (without printing it)

3) If it STILL says “device token mismatch”, your gateway service is probably using a different token via systemd env or ~/.openclaw/.env

How to tell if it’s actually “fixed”

The most common remaining “gotcha”

If you want it “fixed properly” long-term

1) Check the paired devices (safe to paste)

2) Clear the client-side cached device token (often fixes immediately)

Important: delete the `"telegram"` entry under plugins

If this is the TUI (`openclaw tui`)

1) `/reset` failing with `unexpected property 'reason'`

2) `openclaw tui` immediately → `gateway disconnected: closed`

Side note: your `/reset` error

2) Separate issue: `/reset` is broken in 2026.2.15

3) If it STILL says “device token mismatch”, your gateway service is probably using a different token via systemd env or `~/.openclaw/.env`