openclaw - ✅(Solved) Fix [Bug]: 18789 port is in use when the startup of OpenClaw 2026.4.5, when configuring the new dreaming schedule [1 pull requests, 1 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#62276Fetched 2026-04-08 03:06:51
View on GitHub
Comments
0
Participants
1
Timeline
2
Reactions
0
Author
Participants
Timeline (top)
labeled ×2

the startup of OpenClaw 2026.4.5 is failure. After configuring the new dreaming,

Root Cause

the startup of OpenClaw 2026.4.5 is failure. After configuring the new dreaming,

PR fix notes

PR #6: docs(skill): note openclaw baseurl requirement

Description (problem / solution / changelog)

Why

Anyone wiring clawtel up behind an OpenClaw agent (clawchief, staffchief, openclaw-in-a-box) hits a silent failure: heartbeats flow, but every one of them has model="" and 0 tokens. The symptom looks like clawtel is broken, but clawtel is fine — tapes just never sees the traffic.

Root cause: OpenClaw instantiates its Anthropic client as

new Anthropic({ baseURL: model.baseUrl, ... })

When model.baseUrl is undefined (the default — config has no baseUrl set anywhere), that explicit baseURL: undefined overrides the SDK's normal readEnv("ANTHROPIC_BASE_URL") destructuring default. Downstream, options.baseURL = baseURL || 'https://api.anthropic.com' snaps to direct Anthropic, so the gateway holds a TLS connection straight to api.anthropic.com and tapes never records anything.

Setting models.providers.anthropic.baseUrl in ~/.openclaw/openclaw.json is what actually routes traffic through tapes. Verified live against clawchief today — before the fix, zero nodes rows for user chats; immediately after, new nodes rows with model=claude-opus-4-6 and real token counts.

What this PR changes

  • Step 1: new "OpenClaw users: ANTHROPIC_BASE_URL alone is not enough" subsection with the exact config snippet and a verification command (ss -tnp should show the gateway peered to 127.0.0.1:8080, not Anthropic's edge).
  • Common issues: split the "Heartbeats send but all zeros" row into OpenClaw and non-OpenClaw variants, each pointing at the right fix.
  • Added a row documenting the pre-0.1.4 cursor-stuck symptom (closed by #4 / fixed in v0.1.4) so users on older builds can self-diagnose without filing.

No code changes, no behavioral change to clawtel itself.

Related

  • openclaw/openclaw#56679 — upstream feature request tracking the root cause (baseUrl override for built-in providers). Commented with live evidence + suggested fix. When that lands, the config workaround this PR documents can be removed.
  • openclaw/openclaw#62276 — adjacent port-18789 collision between a custom system-level gateway unit and OpenClaw's user-level openclaw-gateway.service. Bit users of the same stack today; commented with repro + workaround.
  • #4 (closed) — cursor format bug fixed in v0.1.4; documented in the new troubleshooting row.
  • #7 (open) — aggregate()'s dominant-model collapse. A mixed-model claw (like the reference clawchief: opus chats + haiku heartbeat + sonnet cron) currently appears single-model on claw.tech; related to #1.

Changed files

  • SKILL.md (modified, +35/-2)
RAW_BUFFERClick to expand / collapse

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

the startup of OpenClaw 2026.4.5 is failure. After configuring the new dreaming,

Steps to reproduce

(base) PS F:\working\openclaw> openclaw gateway --port 18789 11:32:54+08:00 [plugins] [lcm] Plugin loaded (enabled=true, db=C:\Users\coddz.openclaw\lcm.db, threshold=0.75) 11:32:54+08:00 [plugins] [lcm] Compaction summarization model: (unconfigured)

🦞 OpenClaw 2026.4.5 (3e72c03) — I can't fix your code taste, but I can fix your build and your backlog.

| o 11:32:54+08:00 [gateway] loading configuration… 11:32:54+08:00 [gateway] resolving authentication… 11:32:54+08:00 [gateway] starting... 11:32:55+08:00 [gateway] feishu_doc: Registered feishu_doc, feishu_app_scopes 11:32:55+08:00 [gateway] feishu_chat: Registered feishu_chat tool 11:32:55+08:00 [gateway] feishu_wiki: Registered feishu_wiki tool 11:32:55+08:00 [gateway] feishu_drive: Registered feishu_drive tool 11:32:55+08:00 [gateway] feishu_bitable: Registered bitable tools 11:32:55+08:00 [gateway] [lcm] Ignoring sessions matching 1 pattern(s): agent:*:cron:** 11:32:55+08:00 [gateway] starting HTTP server... 11:32:55+08:00 [canvas] host mounted at http://127.0.0.1:18789/__openclaw__/canvas/ (root C:\Users\coddz.openclaw\canvas) 11:32:57+08:00 Gateway failed to start: another gateway instance is already listening on ws://127.0.0.1:18789 If the gateway is supervised, stop it with: openclaw gateway stop 11:32:57+08:00 Port 18789 is already in use. 11:32:57+08:00 - Port is in use but process details are unavailable (install lsof or run as an admin user). 11:32:57+08:00 Gateway service appears registered. Stop it first. 11:32:57+08:00 Tip: openclaw gateway stop 11:32:57+08:00 Or: schtasks /End /TN "OpenClaw Gateway"

Expected behavior

testing dreaming schedule of OpenClaw 2026.4.5

Actual behavior

18789 port is in use when the startup of OpenClaw 2026.4.5, when configuring the new dreaming schedule

OpenClaw version

2026.4.5

Operating system

Window11

Install method

No response

Model

minimax

Provider / routing chain

openclaw-minimax

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 most likely fix is to stop the existing OpenClaw gateway instance that is already listening on port 18789.

Guidance

  • Check if another OpenClaw gateway instance is running and stop it using the command openclaw gateway stop as suggested in the error message.
  • If the process details are unavailable, try running the command prompt or terminal as an administrator to get more information.
  • Alternatively, use the command schtasks /End /TN "OpenClaw Gateway" to stop the scheduled task.
  • Verify that port 18789 is not in use by another process before attempting to start the OpenClaw gateway again.

Example

No code snippet is provided as it is not necessary for this issue.

Notes

The issue seems to be specific to the Windows 11 operating system and OpenClaw version 2026.4.5. The error message clearly indicates that another instance of the gateway is already listening on the specified port.

Recommendation

Apply the workaround by stopping the existing gateway instance using the suggested commands. This should resolve the issue and allow the new instance to start successfully.

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

testing dreaming schedule of OpenClaw 2026.4.5

Still need to ship something?

×6

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

Back to top recommendations

TRENDING