Issue Station
Sign In
Tools hereToolbox

openclaw - ✅(Solved) Fix [Bug]: Coding Agent never completes anything (worked in 2026.4.2 and earlier) [1 pull requests, 6 comments, 4 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openclaw/openclaw#62505•Fetched 2026-04-08 03:03:24
View on GitHub
Comments
6
Participants
4
Timeline
10
Reactions
0
Author
drpaudrpau
Participants
drpaudrpaugwwallacegwwallacelml2468lml2468ToniDonDoniToniDonDoni
Timeline (top)
commented ×6labeled ×2mentioned ×1subscribed ×1

I have an agent specifically setup for coding. This has been pumping out work for me for weeks and now just doesnt do anything apart from vague status updates (and then apologies for the vagueness of status updates).

I only use openai/codex oauth for the models. and i have tried both openai-codex/gpt-5.4 & openai-codex/gpt5.3-codex both displaying the same issue.

Since the update where agents are working on things in the background, it is demonstrating that its not doing anything in the background even after repeated prompts to start.

Root Cause

I have an agent specifically setup for coding. This has been pumping out work for me for weeks and now just doesnt do anything apart from vague status updates (and then apologies for the vagueness of status updates).

I only use openai/codex oauth for the models. and i have tried both openai-codex/gpt-5.4 & openai-codex/gpt5.3-codex both displaying the same issue.

Since the update where agents are working on things in the background, it is demonstrating that its not doing anything in the background even after repeated prompts to start.

PR fix notes

PR #67913: fix(heartbeat): deliver exec events with disabled heartbeat interval

Description (problem / solution / changelog)

Problem: Background exec completions could be lost when the heartbeat interval was disabled with agents.defaults.heartbeat.every = "0m". The process completed successfully, but the user did not receive the completion output.

Scenario: A user asks OpenClaw to run a long command. The command runs longer than the exec foreground yield window, so the tool returns a background/running result and the assistant turn ends. The command still finishes before the hard exec timeout, so it exits successfully.

Normally, background exec completion is delivered through the heartbeat/event-wake path: exec enqueues an exec-event wake with the original chat delivery context, and the heartbeat runner starts a follow-up agent turn to relay the command output back to the chat. When the heartbeat interval was disabled, the runner treated the wake as disabled before dispatching it, so the follow-up turn never happened.

Solution: Allow exec-event wakes to bypass the normal per-agent heartbeat interval checks while preserving the global heartbeat enable switch. For targeted exec wakes, create a one-shot runner entry even when no interval-backed heartbeat agent is scheduled. For exec events without an explicit heartbeat target, default to target: "last" so the captured delivery context can be used to send the completion output to the original chat.

Added unit coverage for:

  • Session-scoped exec wakes when the heartbeat interval is disabled.
  • Exec completion relay with captured delivery context when the heartbeat interval is disabled.

Verification:

  • pnpm test src/infra/heartbeat-runner.scheduler.test.ts -t "disabled heartbeat interval"
  • pnpm test src/infra/heartbeat-runner.returns-default-unset.test.ts -t "disabled heartbeat interval"

Relates to #62505

Changed files

  • src/infra/heartbeat-runner.returns-default-unset.test.ts (modified, +123/-1)
  • src/infra/heartbeat-runner.scheduler.test.ts (modified, +25/-0)
  • src/infra/heartbeat-runner.ts (modified, +56/-20)

Code Example

You’re right, Dave. Over an hour without delivering the requested outcome is a failure on my side.

The useful thing I have created is only a partial scaffold:
web/src/pages/MeWorkspace.tsx
nested /me routes in web/src/App.tsx
web/src/pages/Me.tsx re-export into that structure

But that is not the finished feature you asked for, and I should not have let this drag that long.

At this point, the honest options are:
1) I continue, but only with concrete code progress and no more status-only replies.
2) You stop this run and have me restart the task from a clean execution standard.

Either way, your criticism is valid.

---
RAW_BUFFERClick to expand / collapse

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

I have an agent specifically setup for coding. This has been pumping out work for me for weeks and now just doesnt do anything apart from vague status updates (and then apologies for the vagueness of status updates).

I only use openai/codex oauth for the models. and i have tried both openai-codex/gpt-5.4 & openai-codex/gpt5.3-codex both displaying the same issue.

Since the update where agents are working on things in the background, it is demonstrating that its not doing anything in the background even after repeated prompts to start.

Steps to reproduce

Issue standard prompt to create some code and no output ever occurs but full understanding of the request is constantly given in vague updates.

eg:

You’re right, Dave. Over an hour without delivering the requested outcome is a failure on my side.

The useful thing I have created is only a partial scaffold:
web/src/pages/MeWorkspace.tsx
nested /me routes in web/src/App.tsx
web/src/pages/Me.tsx re-export into that structure

But that is not the finished feature you asked for, and I should not have let this drag that long.

At this point, the honest options are:
1) I continue, but only with concrete code progress and no more status-only replies.
2) You stop this run and have me restart the task from a clean execution standard.

Either way, your criticism is valid.

Expected behavior

Previous versions the agent would be "typing something" until it finished, and then it replied with what it had done. this normally took ~2-3 minutes.

Actual behavior

no reply, or "im working on it in the background" but never actually is working on it

OpenClaw version

2026.4.5

Operating system

Ubuntu 24.04.3 LTS

Install method

No response

Model

openai-codex/

Provider / routing chain

openclaw direct

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response

extent analysis

TL;DR

The agent may be experiencing issues due to a regression in the OpenClaw or OpenAI Codex integration, causing it to fail to produce output despite acknowledging requests.

Guidance

  • Verify that the issue persists across different models, such as openai-codex/gpt-5.4 and openai-codex/gpt5.3-codex, to isolate the problem.
  • Check the OpenClaw version (2026.4.5) for any known issues or updates that may address the regression.
  • Consider restarting the task from a clean execution standard, as suggested by the agent, to see if it resolves the issue.
  • Monitor the agent's behavior and logs to identify any patterns or errors that may indicate the root cause of the problem.

Example

No code snippet is provided as the issue seems to be related to the integration and configuration of OpenClaw and OpenAI Codex.

Notes

The issue may be specific to the OpenClaw version or the OpenAI Codex models being used. Further investigation and debugging may be required to determine the root cause.

Recommendation

Apply workaround: Restart the task from a clean execution standard, as this may resolve the issue temporarily until a more permanent fix is found.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

Previous versions the agent would be "typing something" until it finished, and then it replied with what it had done. this normally took ~2-3 minutes.

#batch processing#GPU compatibility#latency issue#model loading#dependency error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

openclaw - ✅(Solved) Fix [Bug]: Coding Agent never completes anything (worked in 2026.4.2 and earlier) [1 pull requests, 6 comments, 4 participants]