Error Message

Unhandled promise rejection: Error: Agent listener invoked outside active run at Agent.processEvents (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:533:10) at file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:380:21 at Object.onUpdate (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:539:7) at emitUpdate (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1524:8) at handleStdout (or handleStderr) at Object.onSupervisorStdout [as onStdout] (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1610:3)

Code Example

Unhandled promise rejection: Error: Agent listener invoked outside active run
    at Agent.processEvents (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:533:10)
    at file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:380:21
    at Object.onUpdate (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:539:7)
    at emitUpdate (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1524:8)
    at handleStdout (or handleStderr)
    at Object.onSupervisorStdout [as onStdout] (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1610:3)

OpenClaw bug report draft

Title

Unhandled promise rejection: Agent listener invoked outside active run

Summary

OpenClaw intermittently throws an unhandled promise rejection from the embedded agent event pipeline:

Error: Agent listener invoked outside active run

The stack points into OpenClaw internals, specifically @mariozechner/pi-agent-core plus the OpenClaw exec/supervisor update bridge. This does not appear to be caused by a bad workspace prompt or invalid cron content.

Environment

OpenClaw app: 2026.4.5
Update available at time of report: 2026.4.20
OS: macOS 26.2 arm64
Node: 25.8.0
Gateway mode: local loopback
Gateway service: LaunchAgent, running

Error text

Unhandled promise rejection: Error: Agent listener invoked outside active run
    at Agent.processEvents (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:533:10)
    at file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent.ts:380:21
    at Object.onUpdate (file:///opt/homebrew/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:539:7)
    at emitUpdate (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1524:8)
    at handleStdout (or handleStderr)
    at Object.onSupervisorStdout [as onStdout] (file:///opt/homebrew/lib/node_modules/openclaw/dist/exec-defaults-uj0McX2k.js:1610:3)

Observed occurrences

1) 2026-04-19 15:27:33 America/Los_Angeles

Source: /tmp/openclaw/openclaw-2026-04-19.log line 67

2) 2026-04-19 16:11:16 America/Los_Angeles

Source: /tmp/openclaw/openclaw-2026-04-19.log line 157

3) 2026-04-19 16:17:57 America/Los_Angeles

Source: /tmp/openclaw/openclaw-2026-04-19.log line 206

this variant came through handleStderr

4) 2026-04-20 11:19:29 America/Los_Angeles

Source: /tmp/openclaw/openclaw-2026-04-19.log line 1473

5) 2026-04-20 11:29:40 America/Los_Angeles

Source: /tmp/openclaw/openclaw-2026-04-20.log line 124

Trigger pattern / suspicion

This looks like a race in the embedded agent event stream when supervisor stdout/stderr updates arrive after the active run has already ended or transitioned.

Evidence:

stack always routes through pi-agent-core Agent.processEvents and agent-loop.ts:onUpdate
updates are coming from OpenClaw exec/supervisor bridge (emitUpdate, handleStdout or handleStderr, onSupervisorStdout)
at least one occurrence happened near cron queue/manual-run churn:
- manual:2f3e1274-8ddd-473b-9310-221293426dbc:1776709039050:2 was queued waiting for an execution slot
another occurrence happened seconds after cron: job removed
this suggests late supervisor output is still being delivered to a listener after run teardown/removal

User-visible impact

noisy unhandled promise rejection in logs
makes foundation health look worse than it is
unclear whether some agent completion updates could be dropped or mishandled in edge cases
no clear evidence yet of user-facing data corruption, but it weakens trust in runtime stability

What was ruled out locally

These were fixed separately and do not explain this error:

missing Telegram targets in cron prompts
wrong STRATEGY.md path in marketing jobs
disallowed model overrides on some crons
missing memory/epiphanies.md
missing same-day daily memory file assumptions

After those fixes, this runtime error still stands out as an internal OpenClaw issue.

Likely reproduction directions

Potential reproduction path:

Have multiple cron-driven isolated agent runs active or queued.
Force one manual cron run while another is waiting for an execution slot.
Remove or finish a related cron/run while supervisor stdout/stderr is still streaming.
Observe whether late stdout/stderr update reaches Agent.processEvents after active-run teardown.

A second likely reproduction path:

Start an embedded agent run that streams tool/process output.
End, replace, or restart the run quickly.
Let delayed stdout/stderr continue arriving from the supervisor socket.
Observe the same exception.

Suggested fix area

Guard the embedded agent update listener so post-teardown supervisor output is ignored or detached cleanly instead of throwing.

Places implicated by stack:

@mariozechner/pi-agent-core/src/agent.ts
@mariozechner/pi-agent-core/src/agent-loop.ts
OpenClaw dist/exec-defaults-*.js supervisor output bridge

Attachments / evidence paths

/tmp/openclaw/openclaw-2026-04-19.log
/tmp/openclaw/openclaw-2026-04-20.log

Local operator note

This bug report was prepared after a cron and messaging reliability cleanup pass. Remaining runtime issue appears platform-side rather than workspace-side.

extent analysis

TL;DR

Investigate and address the synchronization of agent state during WebSocket reconnects to prevent the Agent listener invoked outside active run error.

Guidance

Review the pi-agent-core library to understand how agent listeners are managed and identify potential issues with the current implementation.
Implement guard checks before invoking agent listeners to ensure they are only called within an active run context.
Investigate the handleStdout → emitUpdate → Agent.processEvents chain to identify potential race conditions and optimize the code to prevent them.
Consider adding proper cleanup of agent state when runs are interrupted to prevent inconsistent state.

Example

No specific code example can be provided without modifying the pi-agent-core library, but a potential solution could involve adding a check before invoking the agent listener:

if (agent.isActive()) {
  agent.processEvents();
} else {
  // Handle the case where the agent is not active
}

Notes

The provided analysis suggests that the issue is related to the agent lifecycle management during WebSocket reconnects and potential race conditions. However, without further investigation and modification of the pi-agent-core library, it is difficult to provide a definitive solution.

Recommendation

Apply a workaround by implementing guard checks and proper cleanup of agent state, as this is a more feasible solution than upgrading to a fixed version, which is not available. This approach can help mitigate the issue until a more permanent solution is found.

openclaw - 💡(How to fix) Fix Unhandled promise rejection: Agent listener invoked outside active run [1 comments, 1 participants]

Recommended Tools

GitHub issue graph ai analysis