You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Web server stops after extended idle period, breaking browser interaction #84
Description:
When a plan is left open in the browser for an extended period without interaction, the web server appears to stop. Once this occurs, the browser UI becomes unresponsive and plan approval/denial/feedback must be done via CLI instead.
Steps to Reproduce:
Enter plan mode and open the plan in browser
Leave the plan idle for an extended period (exact timeout TBD)
Attempt to interact with the plan in the browser
Expected: Browser UI remains functional regardless of idle time
Actual: Web server becomes unresponsive; forced to fall back to CLI
Workaround: Use CLI for plan interactions when browser stops responding
Instead of increasing the timeout to an arbitrary number that will eventually get surpassed also (I left one session open overnight), maybe put the command to start the server back up in the footer of the plan or something.
The root cause is that Claude Code hooks manage the shell process that plannotator runs in, and hooks require a timeout configuration. Previously this was set to 30 minutes, which would kill the server during extended idle periods.
We've increased the timeout to 4 days (345600 seconds), which should cover most use cases.
There may be more sophisticated solutions (e.g., background process management, reconnection logic), but this is the simplest fix given how Claude Code hooks work - there's no easy way to restart the server once the hook process is terminated.
Activity
backnotprop commentedon Jan 16, 2026
@bdbull which harness are you using, Claude Code or OpenCode
bdbull commentedon Jan 16, 2026
@backnotprop I'm using Claude Code.
Instead of increasing the timeout to an arbitrary number that will eventually get surpassed also (I left one session open overnight), maybe put the command to start the server back up in the footer of the plan or something.
fix: increase hook timeout to 4 days
backnotprop commentedon Jan 17, 2026
Fixed in #88.
The root cause is that Claude Code hooks manage the shell process that plannotator runs in, and hooks require a timeout configuration. Previously this was set to 30 minutes, which would kill the server during extended idle periods.
We've increased the timeout to 4 days (345600 seconds), which should cover most use cases.
There may be more sophisticated solutions (e.g., background process management, reconnection logic), but this is the simplest fix given how Claude Code hooks work - there's no easy way to restart the server once the hook process is terminated.
fix: increase hook timeout to 4 days (backnotprop#88)