Help
Troubleshooting
The fastest fix for most issues is a Redeploy. If that doesn't do it, find your symptom below.
Try this first
Open Settings → Instance Controls, click Redeploy, and wait for the instance to return to Running. Configuration changes only apply after a deploy, so this resolves the majority of problems.
The agent isn't replying
Work through these in order:
- Confirm the instance is Running (not Stopped or Deploying).
- Confirm you clicked Redeploy after your last change.
- Test in the Chattab — if web chat works but a messaging app doesn't, the issue is with that channel.
- Send
/newto start a fresh session in case the context got stuck.
A channel token was rejected
Re-copy the token from the source without extra spaces and paste it again under Settings → Channels. For Telegram, a valid token is digits, a colon, then 30–50 letters/numbers (123456789:ABCdef...). If it leaked, rotate it and paste the new one. See the per-platform guides for Telegram, Discord, and Slack.
A setting change didn't take effect
Channels, secrets, model selection, and feature toggles all require a Redeploy to apply. Make the change, save, then redeploy and wait for Running.
A tool says it has no access
The matching secret is probably missing or misnamed. Open your Agent Environment Profile, add the exact variable the tool expects (for example GITHUB_TOKEN), then redeploy. Each integration guide lists the required secret and scopes.
The instance is stuck or unresponsive
Use Restart from the Control UI to cycle it without changing settings. If it's still wedged, Stop and Start it. As a last resort, destroy and re-provision (export anything you need first — this is irreversible).
A scheduled trigger didn't run
Triggers only fire while the instance is Running. Confirm the schedule's time zone, that the target channel is connected, and that any required secret is present — then redeploy. More in Triggers.
Still stuck?
Check the FAQ for product questions, or book a setup session and we'll help you get unblocked.