Account email: [email protected]
Webhook Automations cannot start new background composers.
All POSTs return:
{“success”:false,“error”:“Failed to start background composer: [failed_precondition] Error”}
Stop / Stop All / Inactive do nothing — runs stay “Running”.
Deleting and recreating the Automation does not free capacity.
New Automation “后台” is Active (Composer 2.5), but Run Test and webhook still fail with the same failed_precondition.
This blocks our production news-review webhook writeback since ~2026-10-07 morning CST.
Steps to Reproduce
Create a Webhook Automation (model: Composer 2.5), enable it (Active).
Trigger via webhook POST or click Run Test.
Observe: Failed to start background composer: [failed_precondition]
Open All Runs — some entries stuck on Running.
Click Stop / Stop All / set Inactive — status does not clear; still cannot start new composers.
Delete Automation and create a new one with a new webhook — same failed_precondition immediately.
Expected Behavior
Stop should terminate stuck Running composers and free account capacity.
After Stop (or recreate), new webhook / Run Test should start a background composer successfully.
Version: 3.18.9
Commit: 2ba48ff3f7514cc4643c52ca9f7b3173d9b66130
Date: 2026-08-27T01:42:22.092Z
Electron: (see About if needed)
OS: Windows_NT x64 10.0.26200
For AI issues: which model did you use?
Composer 2.5 (Automation / Cloud Agent)
For AI issues: add Request ID with privacy disabled
No bc- id: composer never starts (fails at start with failed_precondition).
Webhook automation id tail: 3395cf25-c1f7-11f1-bb68-864e54d14197
Additional Information
Auth is valid (not 401 Invalid API key; not “automation disabled”).
Only capacity/start fails.
Please reset/clear stuck background composer slots for this account.
Do not need full webhook secret in this ticket.
The failed_precondition error usually means the account has hit its limit on concurrent Cloud Agents. Agents started by an automation you’ve since deleted can stay stuck in a starting or running state and keep counting toward that limit, even though they no longer show up in any automation’s run history.
Could you check a couple of things so we can narrow this down?
Open cursor.com/agents, and in the agent list filter set Source to Automations (agents created by automations are hidden from the list by default). Do you see older agents stuck starting or running? A screenshot would help.
If you do, archive them and click Run Test again. Does the run start now? If it still fails, let me know roughly when (with timezone) a failed attempt happened.
The “Running” entries in your run history are runs that were rejected at start, so they aren’t using any capacity. That part isn’t intended behavior, and I’ll follow up here once I know more.
@mohitjain my Automation cloud-agent slots are still stuck for [email protected]. On https://cursor.com/agents I skipped onboarding and only see “No Agents Yet” with no Source filter, so I can’t switch to Automations to archive them. Could you please clear/reset those stuck slots? Thanks.
I’ve cleared them on our side, so Run Test and new webhook calls should start now.
One thing to watch: each run holds a slot while it runs and for several minutes after, so if the webhook fires every 30 seconds or so for a long stretch, some calls can still hit the limit. Batching events into fewer runs avoids that.
Let me know if you still see the failed_precondition error.
