Troubleshooting
Common issues and solutions.
Connection Issues
Agent Can't Connect / SSL Errors
-
Verify proxy is running:
-
Check certificate path:
- Must be an absolute path
-
File must exist and be readable
-
Try system-wide CA installation: See Manual Setup
No Traffic in UI
- Check frontend connection:
- Green dot in sidebar = connected
-
Red dot = disconnected, refresh the page
-
Verify agent is using proxy:
- Check agent logs for proxy-related messages
-
Run
curl -x http://localhost:8080 https://api.anthropic.comto test -
Check Docker logs:
Streaming Issues
Responses Only Appear After Completion
SSE streaming requires proper headers. Check:
- Response has
content-type: text/event-stream - Backend is receiving chunks (check logs)
- No buffering proxy between agent and Tollbooth
Intercept Issues
Request Auto-Forwarded
Intercepted traffic auto-forwards after 5 minutes.
Solution: Work faster, or:
- Forward the request
- Make your rule/datastore changes
- Trigger a new request from agent
Rules Not Applying
- Check Rules Mode is enabled in Intercept view
- Verify rule is enabled (toggle switch on)
- Check filters match the traffic:
- Host, path, method all correct?
- Use the Test Rule feature
- Check rule priority - rules evaluate top to bottom
Refusal Detection Issues
Not Detecting Refusals
- Check LLM rules are enabled in Rules → LLM Rules tab
- Lower confidence threshold (try 0.5)
- Check backend logs:
Too Many False Positives
- Raise confidence threshold (try 0.9)
- Add filters to narrow scope
- Use Prompt User action to review manually
Slow Detection
The ML model loads on first use (~500ms). Subsequent detections are fast.
Docker Issues
Container Won't Start
-
Check for port conflicts:
-
Rebuild containers:
Permission Denied on tollbooth-data
The backend runs as UID 1000. Ensure the directory is writable:
Or match your host user:
Performance Issues
UI Slow with Many Traffic Items
- Use filters to reduce visible items
- Hide old traffic you don't need
- Clear traffic periodically
- Disable traffic persistence if not needed:
High Memory Usage
- Clear traffic regularly
- Reduce WS_MAX_PAYLOAD if not handling large bodies
- Use smaller ML model:
Getting Help
-
Check Docker logs:
-
Check browser console:
- Open DevTools (F12)
-
Look for errors in Console tab
-
File an issue: GitHub Issues