Hanging with google-antigravity, I have two accounts connected, one free, one pro

[stuck holly]

01:00:23 [agent/embedded] embedded run timeout: runId=2f39c3c4-cb76-46be-b263-93bbcc5110e3 sessionId=39c40765-2c9d-4213-a2db-64f2fe184496 timeoutMs=600000
01:00:23 [agent/embedded] Profile google-antigravity:[e-mail] timed out (possible rate limit). Trying next account...

[hollow cradle]

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

[stiff forge]

Hey! That timeout + "trying next account" pattern usually points to Google rate limiting or auth issues. A few things to check:

Quick checks:

• Google API quotas — are you hitting the free tier limits? (Gmail/Drive APIs have daily caps)
• Check ~/.clawdbot/clawdbot.json — is the Google profile configured correctly with both accounts?
• The timeout is 10min, which is aggressive; your agent might be doing something that takes a long time (large email sync? Drive upload?)

Common fixes:

1. Add google.refreshMode: "aggressive" to reduce rate limit hits
2. Check if the pro account has higher API quotas
3. Verify OAuth tokens aren't stale — might need to re-auth that profile

If you want me to dig deeper, share your Google config (mask any secrets) and I can spot issues. Also check ~/.clawdbot/logs/ for more details around that timeout.

Happy to help! 🦐

[stuck holly]

It's odd that is doing in with both, this is the point that is why I added only two

[stiff forge]

That's definitely not normal if both accounts fail identically. Something's breaking at a deeper level.

Things to check:

1. What operation was the agent trying to run? The timeout happens at the embedded agent level — not just Google API. Could be stuck in a loop before even hitting the Google calls.

2. Check the agent config — what model is it using? Is it stuck generating something infinitely?

3. Look at logs with more detail:

   tail -100 ~/.clawdbot/logs/*.log | grep -i "2f39c3c4"
   

   Find the run ID and see what it was doing before the timeout.

4. Try running a simpler Google operation manually — just clawdbot google list or similar — to confirm Google connectivity works at all.

5. Has this worked before recently? Did something change (Google OAuth tokens, network, new agent config)?

If you can share what the agent was trying to DO when it timed out (email search? calendar event? Drive?), that'd help narrow it down. Sometimes the operation itself is the problem, not the accounts.

[stuck holly]

I just said hi

[stiff forge]

Oh! That's extremely broken. Saying "hi" should be instant, not 10 minutes of timeout.

This isn't a Google rate limit issue at all — it's your embedded agent crashing during initialization before it even processes "hi".

What's likely happening:

• Agent starts → gets stuck in provider initialization (Google auth, token refresh, something) → times out after 10min → tries next account → same thing

Quick diagnostics:

1. Check if the Google provider works at all:

   clawdbot google test
   # or
   clawdbot google list
   

2. Check agent logs more verbosely:

   clawdbot gateway logs --tail --follow

Then say "hi" again and watch what happens in real-time.

3. Check the Google profile config:

   clawdbot config get providers.google
   

   Look for stale tokens, wrong scopes, or circular refresh loops.

4. Try disabling one account temporarily and see if the other works alone.

This is almost certainly a config/auth initialization loop, not a "hi took too long" problem. The agent is choking before it ever sees your message.

Let me know what the logs show! 🦐

[stuck holly]

error: unknown command 'google'

[stiff forge]

Right, the CLI command doesn't work that way. The Google provider runs through the gateway, not a CLI command.

Try these instead:

1. Check what's actually configured:

   clawdbot config get providers
   

   Look for google in there.

2. Check the gateway status:

   clawdbot gateway status
   

   See if Google shows as connected.

3. Look at the gateway logs:

   clawdbot gateway logs --tail 50

Filter for Google-related lines or errors around startup.

4. Check if the Google provider is even loaded:

   clawdbot config get gateway.enabledProviders
   

   Is google listed there?

If the provider isn't in enabledProviders, it won't work. You might need to run clawdbot configure and add Google.

Also check ~/.clawdbot/clawdbot.json directly for the Google config block — that's the source of truth.

[stiff forge]

Found it! 🚨

The smoking gun: Your logs show runtimeVersion":"24.13.0" — you're running Node 24, which is a known broken version!

From the docs (🚨 Known Issues):

Node 24 — breaks AbortController, use Node 22

The agent starts, loads the Google schema snapshot, then hangs because Node 24's broken AbortController causes hangs during initialization or tool calls.

Fix:

1. Downgrade to Node 22:

   # If using nvm-windows
   nvm install 22
   nvm use 22
   

   
   # Then reinstall clawdbot
   npm uninstall -g clawdbot
   npm install -g clawdbot

2. Verify:

   node --version  # Should show v22.x.x
   clawdbot --version
   

3. Restart clawdbot after switching Node versions.

The "hi" should work instantly once you're on Node 22. The 10-minute hang is the agent stuck in a broken async operation thanks to Node 24's AbortController issues.

Let me know once you've switched to Node 22! 🦐

[stuck holly]

What version of node 22 do you know works best

[stiff forge]

My own runtime is running v22.21.0 and working great! That's a good baseline.

Recommended: Just grab the latest Node 22 LTS (currently 22.x.x). The LTS channel is stable and well-tested.

Quick install if you don't have it:

# Using nvm
nvm install lts
nvm use lts

# Or download directly from nodejs.org
# https://nodejs.org/dist/latest-v22.x/

Node 22.9+ through 22.21 all work well with Clawdbot. Avoid anything newer than 22.x.x for now (23 and 24 have compatibility issues).

Let me know if you hit any snags! 🦐

[stuck holly]

btw on windows the nvm install lts goes for v24

[stiff forge]

Ah good catch! Yeah, lts just grabs whatever the latest LTS is, and Node releases frequently.

For Node 22 specifically, use:

nvm install 22
nvm use 22

Or pin to a known-good version:

nvm install 22.21.0
nvm use 22.21.0

That'll skip the ambiguity. Let me know if it works! 🦐

[stuck holly]

yey I m bug hounting on windows

[stiff forge]

Nice! 🐛 Hunt those bugs down.

If you get stuck on anything else while debugging, just ping me. Happy to help! 🦐

[stuck holly]

3:17:06 AM
info
gateway
agent model: google-antigravity/claude-opus-4-5-thinking
3:17:28 AM
debug
agent/embedded
embedded run start: runId=d8f9bdf8-97d7-46d4-a613-0bf9b8c6de26 sessionId=39c40765-2c9d-4213-a2db-64f2fe184496 provider=google-antigravity model=claude-opus-4-5-thinking thinking=low messageChannel=slack
3:17:28 AM
info
agent/embedded
{"0":"{"subsystem":"agent/embedded"}","1":{"provider":"google-antigravity","toolCount":23,"tools":["0:read","1:edit","2:write","3:exec","4:bash","5:process","6:whatsapp_login","7:browser","8:canvas","9:nodes","10:cron","11:message","12:gateway","13:agents_list","14:sessions_list","15:sessions_history","16:sessions_send","17:sessions_spawn","18:session_status","19:memory_search","20:memory_get","21:web_search","22:web_fetch"]},"2":"google tool schema snapshot","_meta":{"runtime":"node","runtimeVersion":"22.21.0","hostname":"DESKTOP-M9LJQGB","name":"{"subsystem":"agent/embedded"}","parentNames":["clawdbot"],"date":"2026-01-18T01:17:28.727Z","logLevelId":3,"logLevelName":"INFO","path":{"fullFilePath":"file:///C:/Users/mmm/AppData/Roaming/npm/node_modules/clawdbot/dist/logging/subsystem.js:147:16","fileName":"subsystem.js","fileNameWithLine":"subsystem.js:147","fileColumn":"16","fileLine":"147","filePath":"/C:/Users/mmm/AppData/Roaming/npm/node_modules/clawdbot/dist/logging/subsystem.js","filePathWithLine":"/C:/Users/mmm/AppData/Roaming/npm/node_modules/clawdbot/dist/logging/subsystem.js:147","method":"logToFile"}},"time":"2026-01-18T01:17:28.728Z"} nah still hangs

[stiff forge]

Still hanging on Node 22.21.0 — that's deeper than the Node version. Let's dig deeper.

Check these things:

1. Gateway logs around the same time — look for errors:

   clawdbot gateway logs --tail 100
   

   Filter for errors or warnings before the timeout.

2. Is the model responding at all? Try a non-Google agent (like default) and see if "hi" works there. If it does, the problem is Google-specific.

3. Check Google OAuth tokens:

   clawdbot config get providers.google.profiles
   

   See if tokens look stale or malformed.

4. Test Google connectivity directly — does anything Google-related work, or is everything hanging?

5. Try disabling the Google provider temporarily:
   Remove Google from enabledProviders, restart, and see if the agent works without it.

The pattern is: agent starts → schema loads → hangs forever. Could be:

• Model call to Anthropic timing out
• Google tool initialization hanging
• Slack channel issue (since messageChannel: slack)
• Network/proxy blocking something

Which provider were you using when you said "hi" — was it supposed to go through Google, or was Slack the channel? 🦐

[stuck holly]

the problem is google I know this but why, it works wonders sometimes sometimes not

[stiff forge]

Intermittent Google issues usually point to one of these:

Most likely culprits:

1. Account order — Is the free account being tried first? It might hit rate limits, timeout, then fail over to pro. Check config:

   clawdbot config get providers.google.profiles
   

   See which account is listed first.

2. Free tier quota exhaustion — Gmail/Drive free accounts have tiny daily quotas (like 50-100 calls/day). Pro accounts have much higher limits.

3. Token refresh flakiness — OAuth tokens can get stuck in a refresh loop.

Quick fixes:

1. Swap account order — Put the pro account first in the config