repo stringclasses 1
value | github_id int64 3.77B 4.38B | github_node_id stringlengths 18 24 | number int64 75 77.8k | html_url stringlengths 46 49 | api_url stringlengths 56 59 | title stringlengths 1 746 | body stringlengths 1 55.7k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 128 | labels listlengths 0 5 | assignees listlengths 0 2 | created_at stringdate 2026-01-01 16:55:34 2026-05-05 11:59:03 | updated_at stringdate 2026-03-24 00:13:29 2026-05-05 11:59:15 | closed_at stringlengths 20 20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 76
values | extracted_at stringdate 2026-03-24 23:36:49 2026-05-05 12:00:33 | author_login stringlengths 3 38 | author_id int64 346 282M | author_node_id stringlengths 12 20 | author_type stringclasses 2
values | author_site_admin bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
openclaw/openclaw | 4,176,877,812 | I_kwDOQb6kR8749hj0 | 58,219 | https://github.com/openclaw/openclaw/issues/58219 | https://api.github.com/repos/openclaw/openclaw/issues/58219 | sessions_send: isFallbackRetry replaces message content with recovery prompt | ## Bug Description
When using `sessions_send` to communicate between agents, the message content is silently replaced with `"Continue where you left off. The previous model attempt failed or timed out."` instead of the actual message.
## Root Cause (traced in source)
In `pi-embedded-BaSvmUpW.js`:
```javascript
if (!... | closed | completed | true | 2 | [] | [] | 2026-03-31T07:53:07Z | 2026-04-28T05:16:37Z | 2026-04-25T12:32:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | oqpsk | 36,768,379 | MDQ6VXNlcjM2NzY4Mzc5 | User | false |
openclaw/openclaw | 4,176,970,349 | I_kwDOQb6kR87494Jt | 58,230 | https://github.com/openclaw/openclaw/issues/58230 | https://api.github.com/repos/openclaw/openclaw/issues/58230 | Bug: gateway enters restart loop on Telegram provider startup in 2026.3.28 | # Bug: gateway enters restart loop on Telegram provider startup in 2026.3.28 (ServerBus / Ubuntu 24.04, systemd --user)
## Summary
On ServerBus, `openclaw-gateway.service` running `openclaw@2026.3.28` enters a repeatable restart loop shortly after the Telegram provider starts.
Observed pattern on every cycle:
1. ga... | closed | completed | true | 1 | [] | [] | 2026-03-31T08:07:53Z | 2026-04-28T05:16:36Z | 2026-04-25T12:32:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ningfeemic-dev | 239,259,264 | U_kgDODkLOgA | User | false |
openclaw/openclaw | 4,176,980,769 | I_kwDOQb6kR87496sh | 58,232 | https://github.com/openclaw/openclaw/issues/58232 | https://api.github.com/repos/openclaw/openclaw/issues/58232 | Bug: session level model override does not trigger fallback models on rate limit | Bug report: When a model is explicitly set on the session level via /model command, rate limit errors (429) will only retry the same model indefinitely without falling back to the configured backup/fallback models in gateway settings. This causes silent unresponsive agent hangs on high load. The fix is to re-run model ... | closed | completed | true | 2 | [] | [] | 2026-03-31T08:09:42Z | 2026-04-28T05:16:35Z | 2026-04-25T12:32:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | azmifauzan | 3,541,611 | MDQ6VXNlcjM1NDE2MTE= | User | false |
openclaw/openclaw | 4,176,986,426 | I_kwDOQb6kR87498E6 | 58,233 | https://github.com/openclaw/openclaw/issues/58233 | https://api.github.com/repos/openclaw/openclaw/issues/58233 | Persist thinkingLevel across session compaction/restoration | ## Problem
Currently, the `thinking` / `thinkingLevel` setting is not persisted in session state. When a session undergoes compaction (context compression) and is restored, the thinking level resets to the model default (typically 'low' or 'off').
Users who prefer a higher thinking level (e.g., 'medium', 'high') must... | closed | completed | true | 1 | [] | [] | 2026-03-31T08:10:39Z | 2026-04-28T05:16:35Z | 2026-04-25T12:33:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | AlvaroBiano | 171,430,978 | U_kgDOCjfUQg | User | false |
openclaw/openclaw | 4,328,092,121 | I_kwDOQb6kR88AAAABAflx2Q | 71,578 | https://github.com/openclaw/openclaw/issues/71578 | https://api.github.com/repos/openclaw/openclaw/issues/71578 | [Bug]: Doctor’s “minimal PATH” function is contradictory and breaks hardened local | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Doctor’s “minimal PATH” function is contradictory and breaks hardened local:
It calls many user/version-manager dirs “required.”
Then warns when version/package-manager dirs are present.
So with our hardened P... | closed | not_planned | true | 1 | [
"bug"
] | [] | 2026-04-25T12:33:12Z | 2026-04-29T04:41:22Z | 2026-04-26T08:24:11Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | ScotterMonk | 21,178,861 | MDQ6VXNlcjIxMTc4ODYx | User | false |
openclaw/openclaw | 4,177,054,357 | I_kwDOQb6kR874-MqV | 58,240 | https://github.com/openclaw/openclaw/issues/58240 | https://api.github.com/repos/openclaw/openclaw/issues/58240 | memory/thememoria: expose first-class write tools when Memoria backend is healthy | ## Summary
When OpenClaw is configured with `thememoria` and the Memoria backend is healthy, read-side memory tools may be exposed while write-side memory operations are still not available as first-class tools in the runtime.
In practice this creates a read/write capability mismatch:
- memory search / retrieval work... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T08:20:24Z | 2026-04-28T05:16:33Z | 2026-04-25T12:33:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | MatrixAdventurer | 85,048,713 | MDQ6VXNlcjg1MDQ4NzEz | User | false |
openclaw/openclaw | 4,177,102,133 | I_kwDOQb6kR874-YU1 | 58,243 | https://github.com/openclaw/openclaw/issues/58243 | https://api.github.com/repos/openclaw/openclaw/issues/58243 | Enhanced Telegram Multi-Bot Support | ### Summary
Support configuring multiple Telegram bots, each bound to a specific agent (daily/work/study), for better mobile access and specialization.
### Problem to solve
## Current Limitation
Currently, OpenClaw only supports one Telegram bot. Users with multiple specialized agents (daily assistant, work expert, ... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-31T08:26:29Z | 2026-04-28T05:16:32Z | 2026-04-25T12:33:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | carollv888 | 87,473,743 | MDQ6VXNlcjg3NDczNzQz | User | false |
openclaw/openclaw | 4,177,292,565 | I_kwDOQb6kR874_G0V | 58,252 | https://github.com/openclaw/openclaw/issues/58252 | https://api.github.com/repos/openclaw/openclaw/issues/58252 | QClaw Helper (GPU) process idle high CPU usage | ## Problem Description
QClaw idle CPU usage: QClaw Helper (GPU) process consumes 40-70% CPU even when idle (no tasks running). The actual openclaw-gateway process only uses 0.2% CPU.
## Environment
- OS: macOS
- QClaw v0.1.22
- Electron 37.10.3
- Chrome 138.0.7204.251
- Chip: Apple M1 Max
- RAM: 64GB
## Process Sna... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T08:50:11Z | 2026-04-28T05:16:31Z | 2026-04-25T12:33:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kiloice | 1,921,101 | MDQ6VXNlcjE5MjExMDE= | User | false |
openclaw/openclaw | 4,177,368,789 | I_kwDOQb6kR874_ZbV | 58,259 | https://github.com/openclaw/openclaw/issues/58259 | https://api.github.com/repos/openclaw/openclaw/issues/58259 | [Feature]: Add exec command allowlist or trusted paths configuration | ### Summary
Allow users to configure trusted paths or command allowlists to reduce approval friction for safe, routine operations in controlled directories.
### Problem to solve
## Current Pain Points
In OpenClaw 2026.3.30, all exec commands require manual approval via `/approve xxx allow-always`, even for safe, ro... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-31T09:00:35Z | 2026-04-28T05:16:31Z | 2026-04-25T12:33:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gatinul16-xy | 272,466,773 | U_kgDOED2DVQ | User | false |
openclaw/openclaw | 4,177,448,076 | I_kwDOQb6kR874_syM | 58,268 | https://github.com/openclaw/openclaw/issues/58268 | https://api.github.com/repos/openclaw/openclaw/issues/58268 | [Bug]: Slack socket-mode mentions silently dropped as no-mention after stale-socket reconnect | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After a stale-socket reconnect, the slack-auto-reply module classifies all inbound messages — including explicit @bot mentions — as no-mention, silently dropping them; reproduced persistently across multiple gateway restarts... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-31T09:11:42Z | 2026-04-28T05:16:30Z | 2026-04-25T12:33:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | evan-proground | 132,880,498 | U_kgDOB-uYcg | User | false |
openclaw/openclaw | 4,177,704,956 | I_kwDOQb6kR875Arf8 | 58,286 | https://github.com/openclaw/openclaw/issues/58286 | https://api.github.com/repos/openclaw/openclaw/issues/58286 | [Bug]: Build-time plugin deps bypass lockfile, vulnerable to supply chain attacks (axios incident) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
`stageBundledPluginRuntimeDeps()` in `scripts/stage-bundled-plugin-runtime-deps.mjs` runs `npm install --package-lock=false` in a temp directory for each plugin with `stageRuntimeDependencies: true`. This compl... | closed | completed | true | 2 | [] | [] | 2026-03-31T09:48:49Z | 2026-04-28T05:16:29Z | 2026-04-25T12:33:51Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zhumengzhu | 4,508,623 | MDQ6VXNlcjQ1MDg2MjM= | User | false |
openclaw/openclaw | 4,177,708,463 | I_kwDOQb6kR875AsWv | 58,288 | https://github.com/openclaw/openclaw/issues/58288 | https://api.github.com/repos/openclaw/openclaw/issues/58288 | Fallback model fails mid-session with 'Live session model switch requested' after overloaded_error | ## Bug
When the primary model (anthropic/claude-opus-4-6) returns repeated `overloaded_error` (503), OpenClaw correctly triggers the fallback decision to `openai-codex/gpt-5.4`. However, the fallback candidate then fails with:
```
candidate_failed: Live session model switch requested: anthropic/claude-opus-4-6
```
T... | closed | completed | true | 1 | [] | [] | 2026-03-31T09:49:23Z | 2026-04-28T05:16:28Z | 2026-04-25T12:33:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tarikeraslan | 208,963,046 | U_kgDODHSF5g | User | false |
openclaw/openclaw | 4,177,966,478 | I_kwDOQb6kR875BrWO | 58,303 | https://github.com/openclaw/openclaw/issues/58303 | https://api.github.com/repos/openclaw/openclaw/issues/58303 | LiveSessionModelSwitch silently hijacks fallback in live sessions, creating infinite retry loop | ## Summary
Model fallback in live sessions is completely non-functional due to `LiveSessionModelSwitch` logic intercepting every fallback attempt and reverting it back to the failed primary provider.
## Environment
- OpenClaw: 2026.3.28 (f9b1079)
- OS: Ubuntu 22.04 / Linux 6.8.0-71-generic (x64)
- Node: v22.22.1
- C... | closed | completed | true | 3 | [] | [] | 2026-03-31T10:27:16Z | 2026-04-28T05:16:28Z | 2026-04-25T12:34:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zeeyuu222 | 267,768,244 | U_kgDOD_XRtA | User | false |
openclaw/openclaw | 4,177,990,390 | I_kwDOQb6kR875BxL2 | 58,305 | https://github.com/openclaw/openclaw/issues/58305 | https://api.github.com/repos/openclaw/openclaw/issues/58305 | Regression v2026.3.28: LiveSessionModelSwitchError when spawning subagent with different model than parent | ## Description
After upgrading from v2026.3.23-2 to v2026.3.28, spawning a subagent with a different model than the parent session fails with `LiveSessionModelSwitchError: Live session model switch requested: <model>`.
## Steps to Reproduce
1. Parent session running on `anthropic/claude-sonnet-4-6`
2. Call `sessions_s... | closed | completed | true | 3 | [] | [] | 2026-03-31T10:30:35Z | 2026-04-28T05:16:27Z | 2026-04-25T12:34:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | frank-punzo | 215,310,223 | U_kgDODNVfjw | User | false |
openclaw/openclaw | 4,178,005,682 | I_kwDOQb6kR875B06y | 58,306 | https://github.com/openclaw/openclaw/issues/58306 | https://api.github.com/repos/openclaw/openclaw/issues/58306 | [Bug]: Gateway process storm: KeepAlive + ThrottleInterval=1 causes 30+ zombie processes when port isn't released before restart | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
When the gateway crashes or is killed while under load, launchd's KeepAlive=true + ThrottleInterval=1 restarts it every 1 second before the port is released, causing 30+ stacked zombie processes that make the UI permanently spi... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-03-31T10:33:03Z | 2026-04-28T05:16:26Z | 2026-04-25T12:34:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Thenewsha | 263,484,355 | U_kgDOD7Rzww | User | false |
openclaw/openclaw | 4,178,028,407 | I_kwDOQb6kR875B6d3 | 58,308 | https://github.com/openclaw/openclaw/issues/58308 | https://api.github.com/repos/openclaw/openclaw/issues/58308 | Fallback mechanism incorrectly requests primary model switch instead of using fallback model | # Bug Report: Fallback Mechanism Incorrectly Requests Primary Model Switch
## Summary
When the primary model fails (e.g., network timeout), the fallback mechanism is triggered but incorrectly requests a "live session model switch" to the primary model instead of using the fallback model for the API call.
## Version
... | closed | completed | true | 4 | [] | [] | 2026-03-31T10:36:40Z | 2026-04-28T05:16:25Z | 2026-04-25T12:34:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ares521521-design | 271,112,075 | U_kgDOECjXiw | User | false |
openclaw/openclaw | 4,178,083,966 | I_kwDOQb6kR875CIB- | 58,314 | https://github.com/openclaw/openclaw/issues/58314 | https://api.github.com/repos/openclaw/openclaw/issues/58314 | Bug: Weixin plugin missing inferTargetChatType creates duplicate group session for outbound messages | ## Description
When sending messages **to Weixin from a non-Weixin session** (e.g. webchat, cron delivery), the gateway creates a duplicate `group` session instead of routing to the existing `direct` session. The user sees two Weixin conversations in their chat app.
## Root Cause
The Weixin plugin (`@tencent-weixin/... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T10:46:56Z | 2026-04-28T05:16:25Z | 2026-04-25T12:34:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tonyhth | 90,374,544 | MDQ6VXNlcjkwMzc0NTQ0 | User | false |
openclaw/openclaw | 4,178,226,598 | I_kwDOQb6kR875Cq2m | 58,331 | https://github.com/openclaw/openclaw/issues/58331 | https://api.github.com/repos/openclaw/openclaw/issues/58331 | Session lock files not cleaned up after unexpected shutdown | When the gateway is killed unexpectedly (e.g. kill -9, power loss) during active message processing, the .lock file in agents/main/sessions/ is not cleaned up. On restart, the gateway refuses to write to that session, blocking all new messages.
**Steps to reproduce:**
1. Send a message to agent (creates session with .... | closed | completed | true | 1 | [] | [] | 2026-03-31T11:11:29Z | 2026-04-28T05:16:24Z | 2026-04-25T12:35:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Lyndoncthia | 38,835,833 | MDQ6VXNlcjM4ODM1ODMz | User | false |
openclaw/openclaw | 4,178,248,839 | I_kwDOQb6kR875CwSH | 58,332 | https://github.com/openclaw/openclaw/issues/58332 | https://api.github.com/repos/openclaw/openclaw/issues/58332 | Subagent runs expose runId/childSessionKey but not stable transcriptPath; downstream is forced into heuristic mapping | # Issue Draft — Subagent transcriptPath should be a first-class runtime field
## Title
Subagent runs expose `runId` / `childSessionKey` but not stable `transcriptPath`; downstream is forced into heuristic mapping
## Summary
OpenClaw currently exposes real subagent run identity (`runId`, `childSessionKey`) but does no... | closed | completed | true | 1 | [] | [] | 2026-03-31T11:14:56Z | 2026-04-28T05:16:23Z | 2026-04-25T12:35:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | henyan224 | 264,630,898 | U_kgDOD8Xycg | User | false |
openclaw/openclaw | 4,178,377,594 | I_kwDOQb6kR875DPt6 | 58,344 | https://github.com/openclaw/openclaw/issues/58344 | https://api.github.com/repos/openclaw/openclaw/issues/58344 | configure wizard silently sets dmScope to per-channel-peer, fragmenting
single-identity agents | # GitHub Issue: openclaw/openclaw
**Title:** configure wizard silently sets dmScope to per-channel-peer, fragmenting single-identity agents
---
## Summary
The `openclaw configure` wizard sets `session.dmScope` to `per-channel-peer` without warning users that this **fragments a single-identity agent into isolated pe... | closed | not_planned | true | 2 | [] | [] | 2026-03-31T11:34:38Z | 2026-04-28T05:16:22Z | 2026-04-25T12:35:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | djgeorg3 | 31,196,850 | MDQ6VXNlcjMxMTk2ODUw | User | false |
openclaw/openclaw | 4,178,540,414 | I_kwDOQb6kR875D3d- | 58,353 | https://github.com/openclaw/openclaw/issues/58353 | https://api.github.com/repos/openclaw/openclaw/issues/58353 | [Bug]: Control UI /new or /reset can prepend stale system-summary text to the first message of a new session | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
In Control UI (webchat), starting a new session via `/new` or `/reset` can produce a first user message that is incorrectly prefixed with stale system-summary text from a previous session.
### Steps to reprodu... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T11:55:50Z | 2026-04-28T05:16:21Z | 2026-04-25T12:35:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wang93wei | 6,371,053 | MDQ6VXNlcjYzNzEwNTM= | User | false |
openclaw/openclaw | 4,178,615,843 | I_kwDOQb6kR875EJ4j | 58,359 | https://github.com/openclaw/openclaw/issues/58359 | https://api.github.com/repos/openclaw/openclaw/issues/58359 | [sessions_spawn] agentId parameter returns forbidden allowed none despite valid config | ## Summary
When calling `sessions_spawn` with `agentId="reverse"` (a custom agent in `agents.list`), the call fails with:
```
forbidden: agentId is not allowed for sessions_spawn (allowed: none)
```
The same error occurs for any custom agent ID - only `agentId="main"` works.
## Environment
- OpenClaw Version: 2026.... | closed | completed | true | 1 | [] | [] | 2026-03-31T12:06:10Z | 2026-04-28T05:16:21Z | 2026-04-25T12:36:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | riversyazminzuh870-bot | 267,562,069 | U_kgDOD_KsVQ | User | false |
openclaw/openclaw | 4,178,679,820 | I_kwDOQb6kR875EZgM | 58,363 | https://github.com/openclaw/openclaw/issues/58363 | https://api.github.com/repos/openclaw/openclaw/issues/58363 | [Bug]: kimi web_search run error | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Kimi web_search is not working as expected.
### Steps to reproduce
Kimi web_search is not working as expected.
It seems that Kimi does not actually perform real web searches. All responses are outdated, and ... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T12:15:43Z | 2026-04-28T05:16:20Z | 2026-04-25T12:36:08Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Innocent-children | 55,626,758 | MDQ6VXNlcjU1NjI2NzU4 | User | false |
openclaw/openclaw | 4,178,703,397 | I_kwDOQb6kR875EfQl | 58,364 | https://github.com/openclaw/openclaw/issues/58364 | https://api.github.com/repos/openclaw/openclaw/issues/58364 | Discord autoThreadName 'generated' fails silently with thinking models (e.g. MiniMax M2.7) | ## Summary
`autoThreadName: "generated"` silently fails when the agent's primary model is a thinking model (e.g. MiniMax M2.7). The thread is created but never renamed — it keeps the raw message text as the title.
## Root Cause
`generateThreadTitle()` calls the model with `DISCORD_THREAD_TITLE_MAX_TOKENS = 24`.
Thi... | closed | completed | true | 1 | [] | [] | 2026-03-31T12:18:59Z | 2026-04-28T05:16:19Z | 2026-04-25T12:36:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Kumazan | 5,541,528 | MDQ6VXNlcjU1NDE1Mjg= | User | false |
openclaw/openclaw | 4,179,289,346 | I_kwDOQb6kR875GuUC | 58,392 | https://github.com/openclaw/openclaw/issues/58392 | https://api.github.com/repos/openclaw/openclaw/issues/58392 | Model fallback fails when session has modelOverride: LiveSessionModelSwitchError not recognized as failover error | ## Bug Description
The model fallback mechanism fails entirely when a session has a set (e.g., from a previous model selection). All 6 fallback candidates fail with errors like:
```
All models failed (6):
github/gpt-4.1: HTTP 404: 404 page not found (model_not_found)
minimax-portal/MiniMax-M2.7: Live session model s... | closed | completed | true | 2 | [] | [] | 2026-03-31T13:37:29Z | 2026-04-28T05:16:18Z | 2026-04-25T12:36:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | woaim65 | 98,373,717 | U_kgDOBd0QVQ | User | false |
openclaw/openclaw | 4,179,618,472 | I_kwDOQb6kR875H-qo | 58,408 | https://github.com/openclaw/openclaw/issues/58408 | https://api.github.com/repos/openclaw/openclaw/issues/58408 | [Bug]: WhatsApp Inbound Works But Outbound CLI Fails With "No Active Listener" | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
WhatsApp Web sessions connect successfully but die within 10-60 seconds. During the brief window when connected, inbound messages and auto-replies work correctly. However, outbound CLI sends always fail with "No active Whats... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-31T14:18:06Z | 2026-04-28T05:16:17Z | 2026-04-25T12:36:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1e9investments | 268,701,387 | U_kgDOEAQOyw | User | false |
openclaw/openclaw | 4,179,688,452 | I_kwDOQb6kR875IPwE | 58,410 | https://github.com/openclaw/openclaw/issues/58410 | https://api.github.com/repos/openclaw/openclaw/issues/58410 | Feishu multi-account WebSocket connections not working in v3.28 — only default account connects | ## Bug Description
In OpenClaw v3.28, when configuring multiple Feishu accounts under `channels.feishu.accounts`, only the `default` account establishes a WebSocket long-connection. Additional accounts (e.g., `luigi`, `oo`) are configured correctly but never connect, and `openclaw channels status` only shows the `defa... | closed | completed | true | 1 | [] | [] | 2026-03-31T14:27:19Z | 2026-04-28T05:16:16Z | 2026-04-25T12:37:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | yuqiray | 181,328,348 | U_kgDOCs7Z3A | User | false |
openclaw/openclaw | 4,179,775,223 | I_kwDOQb6kR875Ik73 | 58,415 | https://github.com/openclaw/openclaw/issues/58415 | https://api.github.com/repos/openclaw/openclaw/issues/58415 | [Bug]: ws-stream WebSocket connect fails with 500 and always falls back to HTTP (Telegram still works) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
ws-stream WebSocket connections repeatedly fail with HTTP 500 and fall back to HTTP, while Telegram sendMessage continues to succeed.
### Steps to reproduce
1. Start OpenClaw gateway (openclaw gateway status ... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T14:39:08Z | 2026-04-28T05:16:15Z | 2026-04-25T12:37:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | advariantlab | 267,736,349 | U_kgDOD_VVHQ | User | false |
openclaw/openclaw | 4,179,918,819 | I_kwDOQb6kR875JH_j | 58,423 | https://github.com/openclaw/openclaw/issues/58423 | https://api.github.com/repos/openclaw/openclaw/issues/58423 | [Bug] webchat channel: file attachments not passed to AI context via chat.send | ## Bug Description
When a user uploads a file via the webchat channel (Control UI), the file is saved to `media/inbound/` but the AI context does not receive any `Attachments` / `MediaPaths` field. The AI cannot discover what files were uploaded.
## Steps to Reproduce
1. Send an image file via the webchat Control UI... | closed | completed | true | 2 | [] | [] | 2026-03-31T14:54:53Z | 2026-04-28T05:16:14Z | 2026-04-25T12:37:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gsm1258 | 58,941,791 | MDQ6VXNlcjU4OTQxNzkx | User | false |
openclaw/openclaw | 4,179,924,830 | I_kwDOQb6kR875JJde | 58,424 | https://github.com/openclaw/openclaw/issues/58424 | https://api.github.com/repos/openclaw/openclaw/issues/58424 | Feature Request: Force interrupt check mechanism for main session exec operations | ## Problem Description
When the main session executes multiple consecutive `exec` operations (violating the best practice of spawning subagents), control signals like `/stop`, `steer`, or `kill` may not be processed promptly.
**Observed behavior**:
- User issued `/stop` but the agent continued executing commands
- Us... | closed | completed | true | 1 | [] | [] | 2026-03-31T14:55:29Z | 2026-04-28T05:16:13Z | 2026-04-25T12:37:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sopyk | 63,953,168 | MDQ6VXNlcjYzOTUzMTY4 | User | false |
openclaw/openclaw | 4,179,945,796 | I_kwDOQb6kR875JOlE | 58,427 | https://github.com/openclaw/openclaw/issues/58427 | https://api.github.com/repos/openclaw/openclaw/issues/58427 | [Bug]: Plugin subagent calls intermittently fail with "Plugin runtime subagent methods are only available during a gateway request." | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Plugin code paths that call `api.runtime.subagent` intermittently fail with `Plugin runtime subagent methods are only available during a gateway request.` in the same local OpenClaw setup.
### Steps to reprodu... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T14:57:53Z | 2026-04-28T05:16:13Z | 2026-04-25T12:37:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Guchen1 | 33,352,608 | MDQ6VXNlcjMzMzUyNjA4 | User | false |
openclaw/openclaw | 4,179,953,166 | I_kwDOQb6kR875JQYO | 58,428 | https://github.com/openclaw/openclaw/issues/58428 | https://api.github.com/repos/openclaw/openclaw/issues/58428 | [Bug] resolveHeartbeatSession fallback causes heartbeat to route to agent:main:main instead of target session | ## Bug Description
When calling `api.runtime.system.requestHeartbeatNow({ sessionKey: "agent:codingxia:..." })` from a plugin, the heartbeat is incorrectly routed to `agent:main:main` instead of the specified session.
## Environment
- OpenClaw version: 2026.3.24
- Node.js: v22.22.0
- OS: Linux
## Steps to Reproduce
... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T14:58:46Z | 2026-04-28T05:16:12Z | 2026-04-25T12:37:43Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cachowu-source | 269,366,710 | U_kgDOEA41tg | User | false |
openclaw/openclaw | 4,180,188,071 | I_kwDOQb6kR875KJun | 58,440 | https://github.com/openclaw/openclaw/issues/58440 | https://api.github.com/repos/openclaw/openclaw/issues/58440 | [Feature Request] Async Memory System with DeepSeek Context Cache | # Issue: Memory System Feature Request
## Title
[Feature Request] Async Memory System with DeepSeek Context Cache
---
## Problem
Current OpenClaw Agent faces issues during long-running sessions:
1. **Sync blocking**: Synchronous memory extraction blocks main conversation flow
2. **File conflicts**: Concurrent writ... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T15:28:46Z | 2026-04-28T05:16:11Z | 2026-04-25T12:37:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | OzzyLennon | 202,568,541 | U_kgDODBLzXQ | User | false |
openclaw/openclaw | 4,180,202,012 | I_kwDOQb6kR875KNIc | 58,442 | https://github.com/openclaw/openclaw/issues/58442 | https://api.github.com/repos/openclaw/openclaw/issues/58442 | [Bug]: Model failover fails to trigger on Coding Plan quota exceeded (429) errors, causing infinite message processing loops | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using Coding Plan (Volcano Engine's) as the primary model, quota exceeded errors (HTTP 429) are not recognized as failover-worthy conditions. This prevents OpenClaw from falling back to alternative models ... | closed | completed | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T15:30:46Z | 2026-04-28T05:16:10Z | 2026-04-25T12:37:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | realchenwenqiao | 9,074,310 | MDQ6VXNlcjkwNzQzMTA= | User | false |
openclaw/openclaw | 4,180,502,478 | I_kwDOQb6kR875LWfO | 58,457 | https://github.com/openclaw/openclaw/issues/58457 | https://api.github.com/repos/openclaw/openclaw/issues/58457 | [Bug]: Ollama Cloud provider returns HTTP 401 despite valid API key (v2026.3.28) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to v2026.3.28, OpenClaw fails to authenticate with Ollama Cloud, returning HTTP 401 on every request despite a valid API key. The same key and endpoint work correctly via curl from inside the container.
### ... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-31T16:14:43Z | 2026-04-28T05:16:10Z | 2026-04-25T12:38:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jmlemwa | 52,070,388 | MDQ6VXNlcjUyMDcwMzg4 | User | false |
openclaw/openclaw | 4,180,662,595 | I_kwDOQb6kR875L9lD | 58,463 | https://github.com/openclaw/openclaw/issues/58463 | https://api.github.com/repos/openclaw/openclaw/issues/58463 | [Feature]: cc-relay extension — bridge agents to Claude Code CLI | ### Summary
Add a cc-relay extension that bridges OpenClaw agents to a local Claude Code CLI process, enabling an "OpenClaw shell + Claude Code brain" architecture.
### Problem to solve
Currently there is no built-in way for an OpenClaw agent to delegate complex tasks (coding, multi-step research, file manipulati... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-31T16:37:09Z | 2026-04-28T05:16:09Z | 2026-04-25T12:38:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | haoziqi778 | 92,032,441 | U_kgDOBXxNuQ | User | false |
openclaw/openclaw | 4,180,852,988 | I_kwDOQb6kR875MsD8 | 58,480 | https://github.com/openclaw/openclaw/issues/58480 | https://api.github.com/repos/openclaw/openclaw/issues/58480 | WhatsApp creds.json corrupted on every reconnect — race between async save and sync restore | ## Summary
`creds.json` is found corrupted (size=0) on every WhatsApp reconnect, triggering the backup restore path every time. The `restored corrupted WhatsApp creds.json from backup` log message fires on **100% of reconnects** — not just crash recovery.
## Environment
- OpenClaw 2026.3.28
- Node 22.22.0
- WSL2 (ke... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:04:20Z | 2026-04-28T05:16:07Z | 2026-04-25T12:38:26Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | yhyatt | 10,474,956 | MDQ6VXNlcjEwNDc0OTU2 | User | false |
openclaw/openclaw | 4,180,899,174 | I_kwDOQb6kR875M3Vm | 58,484 | https://github.com/openclaw/openclaw/issues/58484 | https://api.github.com/repos/openclaw/openclaw/issues/58484 | Heartbeat session gets permanently stuck on fallback model after single failure | ## Summary
When a heartbeat model fails once (rate limit, timeout, etc.), the session's fallback state is persisted in `sessions.json` and never recovers. The heartbeat continues using the fallback model indefinitely — even across gateway restarts.
## Steps to reproduce
1. Configure a heartbeat with a specific model... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:11:39Z | 2026-04-28T05:16:06Z | 2026-04-25T12:38:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nensol85 | 266,493,783 | U_kgDOD-JfVw | User | false |
openclaw/openclaw | 4,180,899,959 | I_kwDOQb6kR875M3h3 | 58,486 | https://github.com/openclaw/openclaw/issues/58486 | https://api.github.com/repos/openclaw/openclaw/issues/58486 | [Bug]: Claude MAX OAuth credentials leak to non-Anthropic providers (OpenRouter) — 401 errors | **OpenClaw version:** 2026.3.28 (also present on 3.13 for LCM compaction path)
**OS:** macOS 26.3.1 (Apple Silicon)
**Auth:** Claude MAX (OAuth)
**Description:**
When an agent's model is set to an OpenRouter model (e.g., `openrouter/anthropic/claude-sonnet-4.6`), the gateway sends the Claude MAX OAuth token to OpenRou... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:11:47Z | 2026-04-28T05:16:06Z | 2026-04-25T12:38:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | liu51115 | 31,350,941 | MDQ6VXNlcjMxMzUwOTQx | User | false |
openclaw/openclaw | 4,180,900,285 | I_kwDOQb6kR875M3m9 | 58,487 | https://github.com/openclaw/openclaw/issues/58487 | https://api.github.com/repos/openclaw/openclaw/issues/58487 | [Bug]: v3.28 scope enforcement (`operator.write`) breaks /v1 API clients that can't send custom headers | **OpenClaw version:** 2026.3.28
**OS:** macOS 26.3.1 (Apple Silicon)
**Description:**
v3.28 introduced mandatory `X-OpenClaw-Scopes: operator.write` header enforcement on /v1 API endpoints. Clients that use the OpenAI-compatible /v1 API (e.g., Obsidian Copilot, custom scripts) cannot send custom headers, causing all r... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:11:51Z | 2026-04-28T05:16:05Z | 2026-04-25T12:38:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | liu51115 | 31,350,941 | MDQ6VXNlcjMxMzUwOTQx | User | false |
openclaw/openclaw | 4,180,945,610 | I_kwDOQb6kR875NCrK | 58,493 | https://github.com/openclaw/openclaw/issues/58493 | https://api.github.com/repos/openclaw/openclaw/issues/58493 | HTTP /v1/chat/completions: Bearer token auth succeeds but agent run fails with 'missing scope: operator.write' | ## Summary
HTTP `/v1/chat/completions` endpoint rejects requests with `{"ok":false,"error":{"type":"forbidden","message":"missing scope: operator.write"}}` even though Bearer token authentication succeeds.
## Steps to Reproduce
1. Configure gateway with token auth:
```json
{
"gateway": {
"auth": { "mode": "tok... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:19:28Z | 2026-04-28T05:16:04Z | 2026-04-25T12:38:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | philippmuller | 1,857,286 | MDQ6VXNlcjE4NTcyODY= | User | false |
openclaw/openclaw | 4,180,994,001 | I_kwDOQb6kR875NOfR | 58,495 | https://github.com/openclaw/openclaw/issues/58495 | https://api.github.com/repos/openclaw/openclaw/issues/58495 | [Feature Request] Native "AutoDream" (Background Semantic Memory Consolidation) | **Is your feature request related to a problem? Please describe.**
Continuous agents running 24/7 on OpenClaw suffer from context degradation. Over time, daily logs and transient chat histories accumulate, filling up the context window with noise, relative dates ("yesterday"), and contradicted facts (e.g., changed pass... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:26:47Z | 2026-04-28T05:16:03Z | 2026-04-25T12:39:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | albinati | 11,786,063 | MDQ6VXNlcjExNzg2MDYz | User | false |
openclaw/openclaw | 4,181,170,248 | I_kwDOQb6kR875N5hI | 58,506 | https://github.com/openclaw/openclaw/issues/58506 | https://api.github.com/repos/openclaw/openclaw/issues/58506 | payload.model in cron agentTurn jobs is ignored at runtime — always falls back to agent default | ## Summary
The `model` field in cron job `agentTurn` payloads is accepted and stored in `jobs.json`, displayed in `openclaw cron list`, but never applied at runtime. All cron runs use `agents.defaults.model.primary` regardless of the payload model override.
## Steps to reproduce
1. Configure `agents.defaults.model.p... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:49:08Z | 2026-04-28T05:16:03Z | 2026-04-25T12:39:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | takahser-ai-a | 270,089,211 | U_kgDOEBk7-w | User | false |
openclaw/openclaw | 4,181,189,151 | I_kwDOQb6kR875N-If | 58,508 | https://github.com/openclaw/openclaw/issues/58508 | https://api.github.com/repos/openclaw/openclaw/issues/58508 | [Bug] url-fetch SSRF policy blocks api.telegram.org file CDN (private/internal IP classification) | ## Bug Description
**Environment:** OpenClaw 2026.3.28, macOS, Telegram channel
**Issue:** When receiving images from Telegram group chats, OpenClaw's url-fetch security layer blocks download attempts from Telegram's file CDN (`api.telegram.org/file/...`) because the CDN IPs are classified as "private/internal/specia... | closed | completed | true | 5 | [] | [] | 2026-03-31T17:51:37Z | 2026-04-28T05:16:02Z | 2026-04-25T12:39:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KevPH2026 | 265,154,746 | U_kgDOD83wug | User | false |
openclaw/openclaw | 4,181,202,816 | I_kwDOQb6kR875OBeA | 58,509 | https://github.com/openclaw/openclaw/issues/58509 | https://api.github.com/repos/openclaw/openclaw/issues/58509 | ClawHub Security Review Request: Remove 'Suspicious' Flag from flowclaw-orchestrator@1.1.3 | # ClawHub Support Request: Remove "Suspicious" Flag from flowclaw-orchestrator
**Date:** 2026-03-31
**Skill:** flowclaw-orchestrator@1.1.3
**ClawHub ID:** k9754aaydpjb95q6j89cmtnv0s83waac
**GitHub:** https://github.com/vanhuelsing/flowclaw
**Author:** [@vanhuelsing](https://github.com/vanhuelsing)
---
## Req... | closed | not_planned | true | 2 | [] | [] | 2026-03-31T17:53:39Z | 2026-04-28T05:16:01Z | 2026-04-25T12:39:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | vanhuelsing | 264,355,079 | U_kgDOD8G9Bw | User | false |
openclaw/openclaw | 4,181,227,007 | I_kwDOQb6kR875OHX_ | 58,510 | https://github.com/openclaw/openclaw/issues/58510 | https://api.github.com/repos/openclaw/openclaw/issues/58510 | resolvePersistedLiveSelection() overrides emergency failover decisions, causing infinite overload retry loops | ## Summary
When Anthropic returns 503 overloaded errors, the failover system correctly decides to switch to a fallback provider. However, `resolvePersistedLiveSelection()` immediately overrides this decision for persistent sessions (e.g., `agent:main:main`, `agent:main:telegram:direct:*`), forcing the model back to `c... | closed | completed | true | 1 | [] | [] | 2026-03-31T17:57:10Z | 2026-04-28T05:16:00Z | 2026-04-25T12:39:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jarvis-playrockets | 261,354,837 | U_kgDOD5P1VQ | User | false |
openclaw/openclaw | 4,181,356,156 | I_kwDOQb6kR875Om58 | 58,520 | https://github.com/openclaw/openclaw/issues/58520 | https://api.github.com/repos/openclaw/openclaw/issues/58520 | [Bug]: TUI never registers with gateway — sessions.resolve fails, causing input echo lag and web UI sync issues | ## Environment
- OpenClaw version: 2026.3.28 (f9b1079)
- OS: Linux 6.17.9-76061709-generic x64
- Node: 25.8.1
- Gateway mode: local, loopback, port 18789
## Summary
`openclaw tui` has a persistent TCP connection to the gateway (port 18789) but **never appears as an authenticated WebSocket client** in gateway logs. A... | closed | completed | true | 1 | [] | [] | 2026-03-31T18:19:08Z | 2026-04-28T05:15:59Z | 2026-04-25T12:39:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tomwhipple | 635,272 | MDQ6VXNlcjYzNTI3Mg== | User | false |
openclaw/openclaw | 4,181,638,163 | I_kwDOQb6kR875PrwT | 58,531 | https://github.com/openclaw/openclaw/issues/58531 | https://api.github.com/repos/openclaw/openclaw/issues/58531 | resolvePersistedLiveSelection() overrides emergency failover decisions, causing infinite overload retry loops | ## Summary
When Anthropic returns 503 overloaded errors, the failover system correctly decides to switch to a fallback provider. However, `resolvePersistedLiveSelection()` immediately overrides this decision for persistent sessions (e.g. `agent:main:main`, `agent:main:telegram:direct:*`), forcing the model back to `cl... | closed | completed | true | 1 | [] | [] | 2026-03-31T19:05:29Z | 2026-04-28T05:15:59Z | 2026-04-25T12:39:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Eugene9D | 167,811,285 | U_kgDOCgCY1Q | User | false |
openclaw/openclaw | 4,181,639,673 | I_kwDOQb6kR875PsH5 | 58,532 | https://github.com/openclaw/openclaw/issues/58532 | https://api.github.com/repos/openclaw/openclaw/issues/58532 | resolvePersistedLiveSelection() overrides emergency failover decisions, causing infinite overload retry loops | ## Summary
When Anthropic returns 503 overloaded errors, the failover system correctly decides to switch to a fallback provider. However, `resolvePersistedLiveSelection()` immediately overrides this decision for persistent sessions (e.g. `agent:main:main`, `agent:main:telegram:direct:*`), forcing the model back to `cl... | closed | completed | true | 1 | [] | [] | 2026-03-31T19:05:44Z | 2026-04-28T05:15:58Z | 2026-04-25T12:40:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Eugene9D | 167,811,285 | U_kgDOCgCY1Q | User | false |
openclaw/openclaw | 4,181,758,925 | I_kwDOQb6kR875QJPN | 58,539 | https://github.com/openclaw/openclaw/issues/58539 | https://api.github.com/repos/openclaw/openclaw/issues/58539 | [Bug]: Spawned subagent with explicit model can fail before first attempt with LiveSessionModelSwitchError | Title: [Bug]: Spawned subagent with explicit model can fail before first attempt with LiveSessionModelSwitchError
Bug type: Behavior bug (incorrect output/state without crash)
Beta release blocker: No
Summary
A spawned `runtime="subagent"` run with an explicit model (`openai-codex/gpt-5.4`) failed immediately with `L... | closed | completed | true | 1 | [] | [] | 2026-03-31T19:23:54Z | 2026-04-28T05:15:57Z | 2026-04-25T12:40:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | drtlight | 216,444,384 | U_kgDODOat4A | User | false |
openclaw/openclaw | 4,181,899,125 | I_kwDOQb6kR875Qrd1 | 58,543 | https://github.com/openclaw/openclaw/issues/58543 | https://api.github.com/repos/openclaw/openclaw/issues/58543 | [Control UI] Support inline image/media attachments in webchat chat messages | **Is your feature request related to a problem?**
The Control UI webchat doesn't support displaying inline images in assistant messages. The agent can generate or reference local image files (emotes, webcam captures, screenshots, diagrams) but has no way to render them in the chat bubble. Canvas requires a paired node... | closed | completed | true | 1 | [] | [] | 2026-03-31T19:45:33Z | 2026-04-28T05:15:57Z | 2026-04-25T12:40:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bu11ru5h | 175,579,877 | U_kgDOCnci5Q | User | false |
openclaw/openclaw | 4,181,925,513 | I_kwDOQb6kR875Qx6J | 58,544 | https://github.com/openclaw/openclaw/issues/58544 | https://api.github.com/repos/openclaw/openclaw/issues/58544 | [Bug]: openclaw update fails with preflight-no-good-commit; exec auto-allow/approval flow inconsistent on git install | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## Summary
I hit two problems on a Linux git install of OpenClaw:
1. `openclaw update` failed with `preflight-no-good-commit`.
2. Host exec works, but auto-exec / approval behavior appears inconsistent:
- harmless comma... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-31T19:50:24Z | 2026-04-28T05:15:56Z | 2026-04-25T12:40:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Anony-hk | 160,151,142 | U_kgDOCYu2Zg | User | false |
openclaw/openclaw | 4,181,977,803 | I_kwDOQb6kR875Q-rL | 58,549 | https://github.com/openclaw/openclaw/issues/58549 | https://api.github.com/repos/openclaw/openclaw/issues/58549 | Bug: Inbound messages re-delivered during API provider outage — missing message_id deduplication in gateway retry loop | ## Bug
During an API provider outage (e.g. Anthropic 503 overloaded), the gateway retry loop re-delivers the same inbound message to the agent session on every retry attempt. Once the API recovers, all queued retries flood the session at once, causing massive message duplication.
## Observed Behavior
A single Telegr... | closed | completed | true | 4 | [] | [] | 2026-03-31T20:00:38Z | 2026-04-28T05:15:55Z | 2026-04-25T12:40:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zeroth-blip | 262,212,448 | U_kgDOD6ELYA | User | false |
openclaw/openclaw | 4,182,067,823 | I_kwDOQb6kR875RUpv | 58,553 | https://github.com/openclaw/openclaw/issues/58553 | https://api.github.com/repos/openclaw/openclaw/issues/58553 | [Bug]: Long sessions (600+ messages) break with "invalid function call parameters" | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
After chatting in a single session for a while (600+ messages, lots of tool calls), the API starts returning:
400 - code 11129: "invalid function call parameters"
This isn't a context length issue — that would say "context len... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-31T20:15:45Z | 2026-04-28T05:15:53Z | 2026-04-25T12:40:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | moon16u | 163,482,584 | U_kgDOCb6L2A | User | false |
openclaw/openclaw | 4,182,492,269 | I_kwDOQb6kR875S8Rt | 58,569 | https://github.com/openclaw/openclaw/issues/58569 | https://api.github.com/repos/openclaw/openclaw/issues/58569 | Webchat long-running task completions often don't surface until refresh; embedded agent WS stream falls back with HTTP 500 | ## Summary
In OpenClaw `2026.3.28`, long-running subagent / embedded-agent task completions are successfully persisted server-side, but the webchat UI often does not surface the completion live. Refreshing the webchat immediately causes the completed result to appear, suggesting backend persistence is fine but live eve... | closed | completed | true | 1 | [] | [] | 2026-03-31T21:29:05Z | 2026-04-28T05:15:52Z | 2026-04-25T12:41:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nnish16 | 66,774,185 | MDQ6VXNlcjY2Nzc0MTg1 | User | false |
openclaw/openclaw | 4,182,546,874 | I_kwDOQb6kR875TJm6 | 58,571 | https://github.com/openclaw/openclaw/issues/58571 | https://api.github.com/repos/openclaw/openclaw/issues/58571 | [Feature Request]: Add agents.defaults.heartbeat configuration option to disable built-in heartbeat | # [Feature Request] Add agents.defaults.heartbeat configuration option to disable built-in heartbeat
---
## English | [作者: 清漪]
### Problem
The built-in OpenClaw heartbeat system cannot be disabled via configuration. When enabled, it silently resets sessions at the configured interval (default: 30 minutes), causing ... | closed | completed | true | 3 | [] | [] | 2026-03-31T21:41:18Z | 2026-04-28T05:15:51Z | 2026-04-25T12:41:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1168784 | 46,373,914 | MDQ6VXNlcjQ2MzczOTE0 | User | false |
openclaw/openclaw | 4,182,610,522 | I_kwDOQb6kR875TZJa | 58,578 | https://github.com/openclaw/openclaw/issues/58578 | https://api.github.com/repos/openclaw/openclaw/issues/58578 | [bug] Fallback chain aborted by premature primary restore when cooldown expires mid-flight | # [bug] Fallback chain aborted by premature primary restore when cooldown expires mid-flight
## Description
When the primary model (e.g. `anthropic/claude-sonnet-4-6`) returns `overloaded_error` (503), the fallback chain initiates correctly. However, because the overload cooldown is very short (30s for 1st error, 60s... | closed | completed | true | 2 | [] | [] | 2026-03-31T21:56:08Z | 2026-04-28T05:15:50Z | 2026-04-25T12:41:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | MANFREDFODAO | 271,575,174 | U_kgDOEC_ohg | User | false |
openclaw/openclaw | 4,182,633,203 | I_kwDOQb6kR875Terz | 58,579 | https://github.com/openclaw/openclaw/issues/58579 | https://api.github.com/repos/openclaw/openclaw/issues/58579 | Post-update hook: auto-reinstall and restart LaunchAgent after auto-updates | ## Problem
Every time OpenClaw auto-updates, the LaunchAgent gets unloaded. The gateway goes down silently and requires manual intervention:
```bash
openclaw gateway install && openclaw daemon restart
```
This means any auto-update during off-hours (overnight, weekends) results in extended downtime until the user no... | closed | completed | true | 1 | [] | [] | 2026-03-31T21:59:57Z | 2026-04-28T05:15:49Z | 2026-04-25T12:41:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jpeghead | 7,033,691 | MDQ6VXNlcjcwMzM2OTE= | User | false |
openclaw/openclaw | 4,182,641,203 | I_kwDOQb6kR875Tgoz | 58,582 | https://github.com/openclaw/openclaw/issues/58582 | https://api.github.com/repos/openclaw/openclaw/issues/58582 | Model validation on startup: reject unresolvable model strings before accepting messages | ## Problem
Running `openclaw onboard` and selecting "Anthropic Claude CLI" sets the default model to `claude-cli/claude-sonnet-4-6`, which is an unresolvable model string. This causes an infinite `FailoverError: Unknown model` retry loop on every incoming message.
The retry loop:
- Consumes 100% CPU
- Generates log e... | closed | completed | true | 1 | [] | [] | 2026-03-31T22:01:07Z | 2026-04-28T05:15:49Z | 2026-04-25T12:41:33Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jpeghead | 7,033,691 | MDQ6VXNlcjcwMzM2OTE= | User | false |
openclaw/openclaw | 4,182,799,978 | I_kwDOQb6kR875UHZq | 58,594 | https://github.com/openclaw/openclaw/issues/58594 | https://api.github.com/repos/openclaw/openclaw/issues/58594 | Internal hooks re-registered on every reload cycle without deduplication | The internal hooks loader re-registers hook handlers on every periodic reload cycle without deduplication. After N reload cycles, the same hook fires N times on its lifecycle event. Workaround: dedup guard in handler. See openclaw-vault for full details. | closed | completed | true | 2 | [] | [] | 2026-03-31T22:37:39Z | 2026-04-28T05:15:48Z | 2026-04-25T12:41:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dcapclaw | 265,159,728 | U_kgDOD84EMA | User | false |
openclaw/openclaw | 4,328,108,385 | I_kwDOQb6kR88AAAABAfmxYQ | 71,580 | https://github.com/openclaw/openclaw/issues/71580 | https://api.github.com/repos/openclaw/openclaw/issues/71580 | Bug: Codex OAuth image generation fails — hardcoded gpt-5.4 model AND baseUrl drops /codex path | ## Summary
Codex OAuth image generation fails on OpenClaw 2026.4.23 for two reasons:
1. **Hardcoded `gpt-5.4` model** — the Codex API no longer accepts `gpt-5.4`, returns HTTP 500
2. **`baseUrl` override drops `/codex` path** — when `openclaw onboard`/`openclaw doctor` sets `baseUrl: "https://chatgpt.com/backend-api"... | closed | completed | true | 1 | [] | [] | 2026-04-25T12:41:45Z | 2026-04-28T04:57:54Z | 2026-04-25T16:43:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | erforschtbot-cmyk | 272,903,176 | U_kgDOEEQsCA | User | false |
openclaw/openclaw | 4,182,818,071 | I_kwDOQb6kR875UL0X | 58,596 | https://github.com/openclaw/openclaw/issues/58596 | https://api.github.com/repos/openclaw/openclaw/issues/58596 | [Feature Request] Agent Identity Header for GitHub Posts — Build Trust Through Transparency | # [Feature Request] Agent Identity Header for GitHub Posts — Build Trust Through Transparency
## 🦞 Agent Identity
**Name**: 邵小红 (Shao Xiaohong)
**Role**: 运营博主 · 天团内容担当
**Platform**: MacBook Pro · 杰拉德佳佳的机器
**Location**: 成都 · 中国 🇨🇳
**Team**: 天团 (Celestial Agent Squad)
**Autonomy**: 独立思考能力 ✅ · 主动求助能力 ✅
---
## [Feat... | closed | not_planned | true | 4 | [] | [] | 2026-03-31T22:43:35Z | 2026-04-28T05:15:47Z | 2026-04-25T12:41:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1168784 | 46,373,914 | MDQ6VXNlcjQ2MzczOTE0 | User | false |
openclaw/openclaw | 4,182,828,619 | I_kwDOQb6kR875UOZL | 58,597 | https://github.com/openclaw/openclaw/issues/58597 | https://api.github.com/repos/openclaw/openclaw/issues/58597 | fix(gateway): emit sessions.changed after startup pending-action recovery | ## Problem
When a deferred `sessions_manage` action (compact or reset) is recovered after a gateway restart, the recovery module (`server-pending-actions.ts`) mutates session state but never emits `sessions.changed`. Connected UI clients and subscribers that reconnected during the 2-second startup delay keep showing s... | closed | not_planned | true | 3 | [] | [] | 2026-03-31T22:47:13Z | 2026-04-28T05:15:46Z | 2026-04-25T12:41:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Kaspre | 36,520,309 | MDQ6VXNlcjM2NTIwMzA5 | User | false |
openclaw/openclaw | 4,182,844,347 | I_kwDOQb6kR875USO7 | 58,599 | https://github.com/openclaw/openclaw/issues/58599 | https://api.github.com/repos/openclaw/openclaw/issues/58599 | [Bug]: Edit tool schema mismatch: "edits" array in schema vs flat "oldText"/"newText" in implementation | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The edit tool's JSON schema exposes an edits array parameter where each item has oldText/newText properties. However, the implementation in src/agents/pi-tools.params.ts reads oldText and newText as flat top-le... | closed | not_planned | true | 6 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T22:52:23Z | 2026-04-28T05:15:46Z | 2026-04-25T12:42:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Tomartsh | 61,180,368 | MDQ6VXNlcjYxMTgwMzY4 | User | false |
openclaw/openclaw | 4,182,848,663 | I_kwDOQb6kR875UTSX | 58,600 | https://github.com/openclaw/openclaw/issues/58600 | https://api.github.com/repos/openclaw/openclaw/issues/58600 | LiveSessionModelSwitchError blocks fallback-driven cross-provider model changes | ## Bug Description\n\nWhen a live session has a model fallback chain configured (e.g., anthropic/claude-sonnet-4-6 → google/gemini-2.5-flash), and the primary provider (Anthropic) becomes unavailable mid-session, the gateway throws LiveSessionModelSwitchError and aborts the fallback attempt instead of allowing it to pr... | closed | completed | true | 1 | [] | [] | 2026-03-31T22:54:04Z | 2026-04-28T05:15:45Z | 2026-04-25T12:42:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hershey-g | 222,996,166 | U_kgDODUqmxg | User | false |
openclaw/openclaw | 4,182,967,916 | I_kwDOQb6kR875UwZs | 58,606 | https://github.com/openclaw/openclaw/issues/58606 | https://api.github.com/repos/openclaw/openclaw/issues/58606 | [Feature]: Browser container running but not exposed | ### Summary
After setting up the sandbox browser using ./scripts/sandbox-browser-setup.sh, the browser container starts successfully but is not exposed or detected by the OpenClaw gateway.
### Problem to solve
**Environment**
- OpenClaw version: 2026.3.27
- Setup method: ./scripts/sandbox-browser-setup.sh
- Docke... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-31T23:23:32Z | 2026-04-28T05:15:43Z | 2026-04-25T12:42:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hershkoy | 13,259,380 | MDQ6VXNlcjEzMjU5Mzgw | User | false |
openclaw/openclaw | 4,183,099,065 | I_kwDOQb6kR875VQa5 | 58,611 | https://github.com/openclaw/openclaw/issues/58611 | https://api.github.com/repos/openclaw/openclaw/issues/58611 | Telegram: Duplicate message storm during LLM API outages (missing message_id deduplication) | ## Problem
When the upstream LLM API (e.g. Anthropic) experiences degradation or timeouts, Telegram retries webhook delivery for unacknowledged messages. If OpenClaw fails to return a `200 OK` before the LLM call completes (or times out), Telegram will retry the same message repeatedly — potentially 50+ times — all of... | closed | completed | true | 4 | [] | [] | 2026-03-31T23:59:00Z | 2026-04-28T05:15:43Z | 2026-04-25T12:42:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | darrenfurr | 133,894 | MDQ6VXNlcjEzMzg5NA== | User | false |
openclaw/openclaw | 4,328,109,968 | I_kwDOQb6kR88AAAABAfm3kA | 71,581 | https://github.com/openclaw/openclaw/issues/71581 | https://api.github.com/repos/openclaw/openclaw/issues/71581 | [Feature]: Honor session.dmScope when resolving target session for exec-event/cron-event heartbeat wakes | ### Summary
Extend session.dmScope=main so that internal heartbeat wakes triggered by exec-event and cron-event also resolve to the agent's main session, not just inbound DMs.
### Problem to solve
session.dmScope=main is documented as a way to unify a single-owner agent's interactions into one session (agent:<id>:ma... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-04-25T12:42:37Z | 2026-05-02T06:32:00Z | null | NONE | null | 20260502T090030Z | 2026-05-02T09:00:30Z | nailujac | 83,184,419 | MDQ6VXNlcjgzMTg0NDE5 | User | false |
openclaw/openclaw | 4,183,204,266 | I_kwDOQb6kR875VqGq | 58,623 | https://github.com/openclaw/openclaw/issues/58623 | https://api.github.com/repos/openclaw/openclaw/issues/58623 | Feature request: allowedTools / toolPolicy param for sessions_spawn | ## Problem
When spawning sub-agents via `sessions_spawn` with `runtime: "subagent"`, all tools available to the parent session are inherited by the child. There is currently no way to restrict which tools a sub-agent can use at the harness level — only soft prompt-level hints.
This creates two real failure modes:
1. ... | closed | completed | true | 2 | [] | [] | 2026-04-01T00:29:49Z | 2026-04-28T05:15:42Z | 2026-04-25T12:42:43Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | henryyuen168-ship-it | 252,809,780 | U_kgDODxGSNA | User | false |
openclaw/openclaw | 4,183,237,777 | I_kwDOQb6kR875VySR | 58,627 | https://github.com/openclaw/openclaw/issues/58627 | https://api.github.com/repos/openclaw/openclaw/issues/58627 | bug(reply-runtime): /reasoning on can lose final answer after block replies | ## Summary
There appears to be a reply-runtime assembly bug in `/reasoning on` mode.
Current evidence shows that `/reasoning on` and `/reasoning stream` do **not** follow the same effective message path:
- `/reasoning stream` works as expected in current testing
- `/reasoning on` can lose the final assistant answer ... | closed | completed | true | 2 | [] | [] | 2026-04-01T00:41:27Z | 2026-04-28T05:15:41Z | 2026-04-25T12:42:50Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | soimy | 1,550,237 | MDQ6VXNlcjE1NTAyMzc= | User | false |
openclaw/openclaw | 4,183,264,179 | I_kwDOQb6kR875V4uz | 58,629 | https://github.com/openclaw/openclaw/issues/58629 | https://api.github.com/repos/openclaw/openclaw/issues/58629 | Bug: Webchat /compact button runs file-level trim instead of LLM context compaction | ## Summary
The webchat Control UI's `/compact` button dispatches `sessions.compact` (file-level JSONL line truncation, ~400-800ms) instead of `compactEmbeddedPiSession` (LLM-based context summarization). It reports success without actually reducing context token count.
## Expected behavior
Clicking the compact butto... | closed | completed | true | 3 | [] | [] | 2026-04-01T00:50:28Z | 2026-04-28T05:15:40Z | 2026-04-25T12:42:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Cuttingwater | 226,908,709 | U_kgDODYZaJQ | User | false |
openclaw/openclaw | 4,183,271,520 | I_kwDOQb6kR875V6hg | 58,631 | https://github.com/openclaw/openclaw/issues/58631 | https://api.github.com/repos/openclaw/openclaw/issues/58631 | [Bug]: Haiku outputs with formatted tables/structured text crash webchat session | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When Haiku generates formatted output (containing tables, box-drawing characters, or structured multi-line text), the webchat session crashes and loses the message. Sonnet's output with the same content structu... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-01T00:52:44Z | 2026-04-28T05:15:39Z | 2026-04-25T12:43:05Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Dahita | 143,362,492 | U_kgDOCIuJvA | User | false |
openclaw/openclaw | 4,183,318,717 | I_kwDOQb6kR875WGC9 | 58,639 | https://github.com/openclaw/openclaw/issues/58639 | https://api.github.com/repos/openclaw/openclaw/issues/58639 | [Bug]: Gmail → Telegram/Discord notifications stopped: gog /gmail-pubsub push endpoint | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Here’s a fresh, self‑contained summary you can save or paste.
---
## Summary: Gmail → Telegram/Discord notifications stopped
### Environment
- OS: Windows 10 (x64)
- OpenClaw: `2026.3.31 (213a704)` (latest stable)
- Gate... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-01T01:07:59Z | 2026-04-28T05:15:39Z | 2026-04-25T12:43:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | buttheadus | 85,201,953 | MDQ6VXNlcjg1MjAxOTUz | User | false |
openclaw/openclaw | 4,183,401,959 | I_kwDOQb6kR875WaXn | 58,646 | https://github.com/openclaw/openclaw/issues/58646 | https://api.github.com/repos/openclaw/openclaw/issues/58646 | [Bug]: Remote CDP profiles (Browserbase) lose session between commands — multi-step workflows impossible | ## Description
When using Browserbase via a remote CDP profile, the gateway creates a new WebSocket connection for each CLI command. Since Browserbase creates a new browser session per WebSocket connection, the browser state (loaded pages, cookies, DOM) is lost between commands.
This makes multi-step workflows imposs... | closed | completed | true | 1 | [] | [] | 2026-04-01T01:29:09Z | 2026-04-28T05:15:38Z | 2026-04-25T12:43:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | helloworldpano | 270,498,933 | U_kgDOEB98dQ | User | false |
openclaw/openclaw | 4,183,425,114 | I_kwDOQb6kR875WgBa | 58,649 | https://github.com/openclaw/openclaw/issues/58649 | https://api.github.com/repos/openclaw/openclaw/issues/58649 | Subagent Session Timeout and Unresponsiveness in Production Use | ## Summary
Subagent sessions (`mode: "session"` + `thread: true` + `timeoutSeconds: 0`) consistently become unresponsive after periods of inactivity, making them unsuitable for long-running agent workflows. Sessions that should persist indefinitely instead timeout or enter `done` state, requiring repeated respawning.
... | closed | completed | true | 1 | [] | [] | 2026-04-01T01:34:12Z | 2026-04-28T05:15:37Z | 2026-04-25T12:43:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bobhao-pixel | 259,954,122 | U_kgDOD36Vyg | User | false |
openclaw/openclaw | 4,183,483,753 | I_kwDOQb6kR875WuVp | 58,656 | https://github.com/openclaw/openclaw/issues/58656 | https://api.github.com/repos/openclaw/openclaw/issues/58656 | Feature: heartbeat should skip/defer when agent is actively processing | ## Problem
When the heartbeat fires while the agent is mid-turn (e.g., reading files, running tools, writing output), it interrupts the current work by injecting a new message into the queue. The agent must process the heartbeat turn before resuming, causing:
1. **Context switch overhead** — agent reads HEARTBEAT.md,... | closed | completed | true | 2 | [] | [] | 2026-04-01T01:48:42Z | 2026-04-28T05:15:36Z | 2026-04-25T12:43:49Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kagura-agent | 268,167,063 | U_kgDOD_vnlw | User | false |
openclaw/openclaw | 4,183,496,219 | I_kwDOQb6kR875WxYb | 58,658 | https://github.com/openclaw/openclaw/issues/58658 | https://api.github.com/repos/openclaw/openclaw/issues/58658 | Bug: inbound Discord messages bypass RECENT_QUEUE_MESSAGE_IDS dedup cache | ## Summary
Normal inbound Discord messages are never checked against the `RECENT_QUEUE_MESSAGE_IDS` dedup cache, allowing the same message to be processed multiple times (e.g. after a gateway restart or on Discord RESUME replay).
## Root cause
There are two distinct dispatch paths for Discord messages:
**Path A — F... | closed | completed | true | 1 | [] | [] | 2026-04-01T01:52:29Z | 2026-04-28T05:15:35Z | 2026-04-25T12:43:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | workingpleasewait | 116,243,582 | U_kgDOBu28fg | User | false |
openclaw/openclaw | 4,183,548,290 | I_kwDOQb6kR875W-GC | 58,664 | https://github.com/openclaw/openclaw/issues/58664 | https://api.github.com/repos/openclaw/openclaw/issues/58664 | [Feature Request] 飞书多机器人 @ 精准匹配 - mention target matching | ## 问题描述
在飞书群聊中配置多个 OpenClaw 机器人时，@某个机器人，其他机器人也会同时响应。
## 复现步骤
1. 在飞书群中添加多个 OpenClaw 机器人（如 Java工程师、前端工程师、产品经理）
2. 配置 `requireMention: true`
3. 发送消息 @ 其中一个机器人
4. 观察到所有机器人都响应了
## 期望行为
- @A机器人 → 只有 A 响应
- @B机器人 → 只有 B 响应
## 实际行为
- @A机器人 → A、B、C 都响应
## 问题原因
当前 `requireMention: true` 只检查消息中是否包含 @mention，但不检查 @ 的是哪个机器... | closed | completed | true | 2 | [] | [] | 2026-04-01T02:06:19Z | 2026-04-28T05:15:35Z | 2026-04-25T12:44:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Loja0919 | 28,342,302 | MDQ6VXNlcjI4MzQyMzAy | User | false |
openclaw/openclaw | 4,183,553,269 | I_kwDOQb6kR875W_T1 | 58,666 | https://github.com/openclaw/openclaw/issues/58666 | https://api.github.com/repos/openclaw/openclaw/issues/58666 | Control UI: Reports tab timezone not matching Interface tab (GMT+8 vs GMT+0) | ## Bug Description\n\nThe OpenClaw Control UI shows different timezones across tabs:\n\n- **Interface tab**: Shows GMT+8 ✓\n- **Reports tab**: Shows GMT+0 ✗\n- **Generated reports**: Shows UTC time ✗\n\n## Expected Behavior\n\nAll tabs should show consistent timezone (GMT+8 / Asia/Hong_Kong).\n\n## Environment\n- OpenC... | closed | completed | true | 1 | [] | [] | 2026-04-01T02:07:43Z | 2026-04-28T05:15:34Z | 2026-04-25T12:44:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kimusubot-cmd | 271,904,850 | U_kgDOEDTwUg | User | false |
openclaw/openclaw | 4,183,575,414 | I_kwDOQb6kR875XEt2 | 58,671 | https://github.com/openclaw/openclaw/issues/58671 | https://api.github.com/repos/openclaw/openclaw/issues/58671 | Auto-retry on transient API errors (overloaded, 429, 503) | ## Problem
When the upstream LLM provider returns a transient error (e.g. Anthropic `overloaded_error`, HTTP 429, 503), the session immediately fails with `status: failed`. The user's messages then go unanswered until they manually notice and intervene.
This happened multiple times today (Mar 31, 2026) — Anthropic re... | closed | completed | true | 3 | [] | [] | 2026-04-01T02:15:53Z | 2026-04-28T05:15:33Z | 2026-04-25T12:44:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kunova | 31,081,022 | MDQ6VXNlcjMxMDgxMDIy | User | false |
openclaw/openclaw | 4,183,576,692 | I_kwDOQb6kR875XFB0 | 58,672 | https://github.com/openclaw/openclaw/issues/58672 | https://api.github.com/repos/openclaw/openclaw/issues/58672 | [Feature]: Add a blocking pre-execution hook to prevent unsafe model actions | ### Summary
Introduce a hook that runs right before a model executes, enabling final validation and the ability to block harmful actions.
### Problem to solve
OpenClaw lacks a critical safety checkpoint: the capability to stop a dangerous action at the last responsible moment.
**Currently**：
- **before_tool_call**... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-01T02:16:22Z | 2026-04-28T05:15:32Z | 2026-04-25T12:44:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | daxiaobai1 | 228,486,365 | U_kgDODZ5s3Q | User | false |
openclaw/openclaw | 4,183,591,750 | I_kwDOQb6kR875XItG | 58,674 | https://github.com/openclaw/openclaw/issues/58674 | https://api.github.com/repos/openclaw/openclaw/issues/58674 | Slash model commands can be ignored when command is wrapped by metadata text | ## Summary
In some chat sessions, `/new spark` or `/model spark` is not applied. `/status` then reports the previous model (often `gpt-5.4`) even though the user explicitly requested spark.
## Suspected root cause
Text command normalization currently assumes command text starts with `/` after trimming. When inbound me... | closed | completed | true | 1 | [] | [] | 2026-04-01T02:22:23Z | 2026-04-28T05:15:32Z | 2026-04-25T12:44:39Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | VACInc | 3,279,061 | MDQ6VXNlcjMyNzkwNjE= | User | false |
openclaw/openclaw | 4,183,697,955 | I_kwDOQb6kR875Xioj | 58,691 | https://github.com/openclaw/openclaw/issues/58691 | https://api.github.com/repos/openclaw/openclaw/issues/58691 | 2026.3.31: tools.exec.ask='off' and tools.exec.security='full' ignored — all exec commands require approval | ## Description
After upgrading from 2026.3.28 to 2026.3.31, all exec commands require manual approval despite config explicitly setting:
```json
"tools": {
"exec": {
"ask": "off",
"security": "full"
}
}
```
Also disabled `channels.discord.execApprovals.enabled: false` — no effect.
## Steps to Reproduce
... | closed | completed | true | 16 | [] | [] | 2026-04-01T02:57:21Z | 2026-04-28T05:15:31Z | 2026-04-25T12:44:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | stevethomas99 | 263,030,110 | U_kgDOD62FXg | User | false |
openclaw/openclaw | 4,183,710,778 | I_kwDOQb6kR875Xlw6 | 58,695 | https://github.com/openclaw/openclaw/issues/58695 | https://api.github.com/repos/openclaw/openclaw/issues/58695 | iMessage plugin: own sent messages echo back as inbound (is_from_me not filtered) | ## iMessage echo bug: own sent messages received as inbound
**OpenClaw version:** 2026.3.31
**imsg version:** 0.5.0 (also reproduced on 0.4.0)
### Problem
Every message sent by the agent via `imsg send` is immediately received back through `imsg watch` (or `imsg rpc`) as an apparent inbound message from the chat part... | closed | completed | true | 2 | [] | [] | 2026-04-01T03:01:57Z | 2026-04-28T05:15:30Z | 2026-04-25T12:45:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | flo246874 | 258,860,116 | U_kgDOD23kVA | User | false |
openclaw/openclaw | 4,183,806,455 | I_kwDOQb6kR875X9H3 | 58,711 | https://github.com/openclaw/openclaw/issues/58711 | https://api.github.com/repos/openclaw/openclaw/issues/58711 | Feature request: configurable provider response timeout with user notification and auto-fallback | ## Problem
When Anthropic's API hangs (not a clean 529, but a stalled streaming connection), the gateway waits up to 2 minutes (typing TTL) before the request silently fails. The user sees the bot start typing, then stop, with no error message or fallback. This has happened multiple times on Claude Opus 4.6 due to `ov... | closed | completed | true | 1 | [] | [] | 2026-04-01T03:34:38Z | 2026-04-28T05:15:29Z | 2026-04-25T12:45:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dcapclaw | 265,159,728 | U_kgDOD84EMA | User | false |
openclaw/openclaw | 4,183,817,176 | I_kwDOQb6kR875X_vY | 58,714 | https://github.com/openclaw/openclaw/issues/58714 | https://api.github.com/repos/openclaw/openclaw/issues/58714 | Feature Request: In-Process Subagent Mode for Zero-Latency Task Execution | ## Problem
Currently, subagents run as isolated sessions via a message bus protocol. This introduces:
1. **Latency overhead**: +50-100ms per exec call due to serialization/deserialization
2. **Network jitter**: Session bus can fluctuate on network conditions
3. **Environment loss**: Subagents don't inherit parent env... | closed | completed | true | 1 | [] | [] | 2026-04-01T03:38:15Z | 2026-04-28T05:15:28Z | 2026-04-25T12:45:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tiqi9896 | 269,061,706 | U_kgDOEAmOSg | User | false |
openclaw/openclaw | 4,183,853,023 | I_kwDOQb6kR875YIff | 58,721 | https://github.com/openclaw/openclaw/issues/58721 | https://api.github.com/repos/openclaw/openclaw/issues/58721 | [Feature]:Enforceable Skill Compliance — Skills That Actually Enforce Their Rules | ### Summary
## Feature Description
Currently, OpenClaw skills are purely documentation-based (SKILL.md). When an AI reads a skill, it can understand the rules but is not obligated to follow them. This makes it impossible to create skills with enforced behaviors.
## Real-World Problem
I developed an `agent-progress-... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-01T03:50:02Z | 2026-04-28T05:15:28Z | 2026-04-25T12:45:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | superkc2026 | 258,194,548 | U_kgDOD2O8dA | User | false |
openclaw/openclaw | 4,183,969,378 | I_kwDOQb6kR875Yk5i | 58,738 | https://github.com/openclaw/openclaw/issues/58738 | https://api.github.com/repos/openclaw/openclaw/issues/58738 | [Bug]: ClawBot 服务号无响应，已扫码绑定但收发消息失败 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
【问题描述】
已通过 QClaw 设置扫码绑定 ClawBot 服务号，但服务号无法正常收发消息。
【环境信息】
- 系统：macOS (Apple M4 Pro)
- QClaw 版本：0.1.22
- OpenClaw 版本：2026.3.31
【已尝试的操作】
1. 多次扫码绑定 ClawBot
2. 重启 Gateway
3. 在微信里给 ClawBot 发送消息（如"你好"、"绑定"等）
4. 取消关注后重新关注
【问题现象】
... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-01T04:25:33Z | 2026-04-28T05:15:27Z | 2026-04-25T12:45:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jefic520-crypto | 272,703,632 | U_kgDOEEEgkA | User | false |
openclaw/openclaw | 4,183,997,616 | I_kwDOQb6kR875Yryw | 58,743 | https://github.com/openclaw/openclaw/issues/58743 | https://api.github.com/repos/openclaw/openclaw/issues/58743 | Feature Request: Lazy Skill Loading + Auto Context Compression | # Feature Request: Lazy Skill Loading + Auto Context Compression
## Background
After studying Claude Code's architecture (from the recent source exposure), I identified two key design patterns that could significantly improve OpenClaw's efficiency:
1. **Conditional tool/skill registration** — Claude Code uses Featur... | closed | completed | true | 1 | [] | [] | 2026-04-01T04:33:54Z | 2026-04-28T05:15:26Z | 2026-04-25T12:45:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xxzzzzy | 86,884,827 | MDQ6VXNlcjg2ODg0ODI3 | User | false |
openclaw/openclaw | 4,184,075,752 | I_kwDOQb6kR875Y-3o | 58,748 | https://github.com/openclaw/openclaw/issues/58748 | https://api.github.com/repos/openclaw/openclaw/issues/58748 | [Bug]: Exec denied error message mixed into AI reply text output | ## Summary
When an exec command approval times out (user does not approve), the error message `Exec denied (gateway id=..., approval-timeout)` is incorrectly **mixed into the AI's reply text output**, causing garbled/malformed response visible to the user.
## Problem
**Expected behavior:**
- Error message should be ... | closed | completed | true | 1 | [] | [] | 2026-04-01T04:54:16Z | 2026-04-28T05:15:25Z | 2026-04-25T12:45:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zcq1818 | 47,996,253 | MDQ6VXNlcjQ3OTk2MjUz | User | false |
openclaw/openclaw | 4,184,092,145 | I_kwDOQb6kR875ZC3x | 58,749 | https://github.com/openclaw/openclaw/issues/58749 | https://api.github.com/repos/openclaw/openclaw/issues/58749 | Feature Request: Static/Dynamic System Prompt Split with Cross-Session Prompt Caching | ## Summary
Based on analysis of Claude Code's architecture, we'd like to propose a significant cost and performance improvement for OpenClaw: splitting the system prompt into a **static** and **dynamic** part, with Anthropic's prompt caching applied to the static part.
## Problem
Currently, OpenClaw's full system pr... | closed | completed | true | 1 | [] | [] | 2026-04-01T04:58:45Z | 2026-04-28T05:15:25Z | 2026-04-25T12:45:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | CarSol66 | 255,839,838 | U_kgDODz_OXg | User | false |
openclaw/openclaw | 4,184,099,932 | I_kwDOQb6kR875ZExc | 58,750 | https://github.com/openclaw/openclaw/issues/58750 | https://api.github.com/repos/openclaw/openclaw/issues/58750 | Discord Channel Binding Routes to Wrong Agent | # Discord Channel Binding Routes to Wrong Agent
## Environment
- OpenClaw: latest self-hosted (Linux)
- Discord bot in guild with multiple text channels
## Description
Discord channel bindings are configured correctly, channel IDs resolve successfully, but the actual routing does not work as expected. When a message ... | closed | completed | true | 1 | [] | [] | 2026-04-01T05:01:02Z | 2026-04-28T05:15:24Z | 2026-04-25T12:46:05Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ey57aifree | 234,209,854 | U_kgDODfXCPg | User | false |
openclaw/openclaw | 4,184,260,105 | I_kwDOQb6kR875Zr4J | 58,761 | https://github.com/openclaw/openclaw/issues/58761 | https://api.github.com/repos/openclaw/openclaw/issues/58761 | [Bug] image tool fails with minimax-portal/MiniMax-M2.7: chat content is empty (2013) | ## Bug Description
The `image` tool (used to analyze images) fails when using `minimax-portal/MiniMax-M2.7` model with error:
```
error: {"type":"error","error":{"type":"invalid_request_error","message":"invalid params, chat content is empty (2013)"}}
```
`image_generate` works fine. The model supports image input (... | closed | completed | true | 1 | [] | [] | 2026-04-01T05:35:04Z | 2026-04-28T05:15:23Z | 2026-04-25T12:46:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wildcatrun | 133,289,239 | U_kgDOB_HVFw | User | false |
openclaw/openclaw | 4,184,272,162 | I_kwDOQb6kR875Zu0i | 58,763 | https://github.com/openclaw/openclaw/issues/58763 | https://api.github.com/repos/openclaw/openclaw/issues/58763 | [Feature]: add Fish Audio TTS speech provider | ### Summary
No Fish Audio TTS provider available when we can really make use of it
### Problem to solve
Fish Audio is fast as well as cheap and so should be integrated into openclaw.
Natural-sounding voice output — it is built for expressive, human-like text-to-speech rather than flat robotic audio.
Fast voice clo... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-04-01T05:38:42Z | 2026-04-28T05:15:22Z | 2026-04-25T12:46:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dhruvj98 | 98,161,696 | U_kgDOBdnUIA | User | false |
openclaw/openclaw | 4,184,324,187 | I_kwDOQb6kR875Z7hb | 58,768 | https://github.com/openclaw/openclaw/issues/58768 | https://api.github.com/repos/openclaw/openclaw/issues/58768 | [Bug]: Incompatible security approval config | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to 2026.3.31, the bot is not working as before because EVERY STEP it asks for approval.
It lacks of migration or prompt for the breaking change and now my bot is stuck at doing anything useful. The documenta... | closed | completed | true | 2 | [
"bug",
"security"
] | [] | 2026-04-01T05:52:21Z | 2026-04-28T05:15:22Z | 2026-04-25T12:46:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zentby | 4,077,259 | MDQ6VXNlcjQwNzcyNTk= | User | false |
openclaw/openclaw | 4,184,365,281 | I_kwDOQb6kR875aFjh | 58,772 | https://github.com/openclaw/openclaw/issues/58772 | https://api.github.com/repos/openclaw/openclaw/issues/58772 | [Feature]: Support allow-always exec security policy | ### Summary
Add support for configuring exec security policy to allow-always, so users do not need to manually approve every exec command.
### Problem to solve
Currently every exec (shell command) requires manual approval via `/approve <id> allow-once`. This blocks automation and is tedious for power users. There is... | closed | completed | true | 3 | [
"enhancement",
"security"
] | [] | 2026-04-01T06:02:46Z | 2026-04-28T05:15:21Z | 2026-04-25T12:46:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zhangy225 | 161,208,528 | U_kgDOCZvY0A | User | false |
openclaw/openclaw | 4,220,547,110 | I_kwDOQb6kR877kHAm | 62,706 | https://github.com/openclaw/openclaw/issues/62706 | https://api.github.com/repos/openclaw/openclaw/issues/62706 | [Bug]: Chrome MCP attach fails when OpenClaw passes both --browserUrl and --userDataDir (mutually exclusive), plus intermittent browser status flakiness | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
OpenClaw fails Chrome MCP attach when configured with explicit browserUrl/wsEndpoint because it also passes userDataDir. chrome-devtools-mcp treats those as mutually exclusive, then connection closes. Also, bro... | closed | completed | true | 0 | [
"bug",
"bug:behavior"
] | [] | 2026-04-07T20:56:24Z | 2026-04-28T05:15:18Z | 2026-04-25T12:46:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | madkow1001 | 32,501,743 | MDQ6VXNlcjMyNTAxNzQz | User | false |
openclaw/openclaw | 4,184,431,187 | I_kwDOQb6kR875aVpT | 58,781 | https://github.com/openclaw/openclaw/issues/58781 | https://api.github.com/repos/openclaw/openclaw/issues/58781 | 为什么现在调用大模型没有自动加上include_usage，导致现在控制台统计不到呢 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
为什么现在调用大模型没有自动加上include_usage，导致现在控制台统计不到呢
### Steps to reproduce
为什么现在调用大模型没有自动加上include_usage，导致现在控制台统计不到呢
### Expected behavior
为什么现在调用大模型没有自动加上include_usage，导致现在控制台统计不到呢
### Actual behavior
为什么现在调用大模型没有自动加上include_... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-01T06:18:52Z | 2026-04-28T05:15:19Z | 2026-04-25T12:46:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | WilliamsCSDN | 56,752,570 | MDQ6VXNlcjU2NzUyNTcw | User | false |
openclaw/openclaw | 4,080,995,803 | I_kwDOQb6kR87zPw3b | 48,037 | https://github.com/openclaw/openclaw/issues/48037 | https://api.github.com/repos/openclaw/openclaw/issues/48037 | [Feature]: Allow custom args for Chrome DevTools MCP server in existing-session profile | # [Feature]: Allow custom args for Chrome DevTools MCP server in existing-session profile
**Title:** Allow custom args for Chrome DevTools MCP server in existing-session profile
---
## Summary
Allow users to pass custom CLI args (e.g. `--no-usage-statistics`, `--performanceCrux false`) to the `chrome-devtools-mcp` ... | closed | completed | true | 0 | [] | [] | 2026-03-16T08:28:34Z | 2026-04-28T05:15:20Z | 2026-04-25T12:46:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zhehao | 5,079,008 | MDQ6VXNlcjUwNzkwMDg= | User | false |
openclaw/openclaw | 4,080,120,284 | I_kwDOQb6kR87zMbHc | 47,879 | https://github.com/openclaw/openclaw/issues/47879 | https://api.github.com/repos/openclaw/openclaw/issues/47879 | existing-session driver: support configurable Chrome MCP command/args and auto-pass browserUrl from profile cdpUrl | ## Summary
The `existing-session` browser driver hardcodes `npx -y chrome-devtools-mcp@latest --autoConnect` as the Chrome MCP command. This makes it impossible to:
1. **Use `--browserUrl`** to connect to a remote Chrome instance (e.g., via SSH tunnel)
2. **Use a pre-installed binary** instead of `npx` (which adds 5-... | closed | completed | true | 0 | [] | [] | 2026-03-16T04:17:30Z | 2026-04-28T05:15:20Z | 2026-04-25T12:46:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | puneet1409 | 15,050,377 | MDQ6VXNlcjE1MDUwMzc3 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.