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 | 4,090,349,077 | I_kwDOQb6kR87zzcYV | 49,183 | https://github.com/openclaw/openclaw/issues/49183 | https://api.github.com/repos/openclaw/openclaw/issues/49183 | RFC: Per-session tool policies — bridging operator scopes and agent tool access | # RFC: Per-session tool policies
## Summary
Add session-level tool filtering that bridges the existing operator scope system (gateway auth) with the agent tool policy engine, enabling multi-user deployments where different sessions have different tool access.
This is a focused subset of #43673 (org/team deployment) ... | open | null | false | 2 | [] | [] | 2026-03-17T18:22:03Z | 2026-04-30T22:39:24Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | antonio-mello-ai | 186,876,424 | U_kgDOCyOCCA | User | false |
openclaw/openclaw | 4,090,364,776 | I_kwDOQb6kR87zzgNo | 49,185 | https://github.com/openclaw/openclaw/issues/49185 | https://api.github.com/repos/openclaw/openclaw/issues/49185 | [Feature]: Configurable maxRetries per provider for model API calls | ### Summary
Add a configurable maxRetriesPerCandidate (or per-provider maxRetries) option to control how many times a provider SDK retries before cascading to the next fallback model in runWithModelFallback.
### Problem to solve
When a model provider returns rate_limit (429) or overloaded (529) errors, the underlyin... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-17T18:25:18Z | 2026-04-30T22:39:26Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jasonneilwalters | 124,079,376 | U_kgDOB2VNEA | User | false |
openclaw/openclaw | 4,090,492,251 | I_kwDOQb6kR87zz_Vb | 49,190 | https://github.com/openclaw/openclaw/issues/49190 | https://api.github.com/repos/openclaw/openclaw/issues/49190 | [Feature]: Cron should distinguish execution success from delivery failure and show warning instead of error | ### Summary
Add a cron run status model that distinguishes successful execution with failed delivery from true execution failure, so Control UI can show `warning` instead of `error` in these cases.
### Problem to solve
Currently, isolated cron jobs can be marked as `error` even when the task itself executes successf... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-17T18:47:36Z | 2026-04-30T21:58:35Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | NorthernDream | 48,941,387 | MDQ6VXNlcjQ4OTQxMzg3 | User | false |
openclaw/openclaw | 4,090,671,195 | I_kwDOQb6kR87z0rBb | 49,205 | https://github.com/openclaw/openclaw/issues/49205 | https://api.github.com/repos/openclaw/openclaw/issues/49205 | [Bug]: Control UI messages can reach shared context but still not appear in Open WebUI visible chat history | **Bug type**
Behavior bug (incorrect output/state without crash)
**Summary**
Messages sent from OpenClaw Control UI into an Open WebUI-backed chat can correctly reach the shared OpenClaw transcript and model context, but still fail to appear in Open WebUI visible chat history as normal message bubbles.
**Steps to rep... | open | null | false | 3 | [] | [] | 2026-03-17T19:20:57Z | 2026-04-30T22:39:27Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | SHAREN | 6,128,858 | MDQ6VXNlcjYxMjg4NTg= | User | false |
openclaw/openclaw | 4,090,948,674 | I_kwDOQb6kR87z1uxC | 49,223 | https://github.com/openclaw/openclaw/issues/49223 | https://api.github.com/repos/openclaw/openclaw/issues/49223 | WhatsApp inter_session delivery requests can be wrongly suppressed as REPLY_SKIP / ANNOUNCE_SKIP / NO_REPLY | ## Summary
When a WhatsApp group session receives an `inter_session` turn that is actually asking it to post a normal human message into the group, the session can incorrectly choose `REPLY_SKIP`, `ANNOUNCE_SKIP`, or `NO_REPLY` instead of emitting the requested message.
This appears to be a precedence bug between:
- ... | open | null | false | 2 | [] | [] | 2026-03-17T20:14:49Z | 2026-05-04T20:14:08Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | TinyTb | 5,957,298 | MDQ6VXNlcjU5NTcyOTg= | User | false |
openclaw/openclaw | 4,090,958,823 | I_kwDOQb6kR87z1xPn | 49,225 | https://github.com/openclaw/openclaw/issues/49225 | https://api.github.com/repos/openclaw/openclaw/issues/49225 | WhatsApp: reactions can succeed while group text/reply sends fail, and transcript text can look like delivery | ## Summary
In WhatsApp group sessions, the reaction/ack path can work while actual outbound text messages or replies do not land.
This suggests a split-path bug where:
- WhatsApp reactions/acks are successfully dispatched
- but text-message or reply-message sends are not actually delivered
- meanwhile the session tra... | closed | completed | true | 3 | [] | [] | 2026-03-17T20:16:40Z | 2026-05-02T04:26:34Z | 2026-04-30T02:38:57Z | CONTRIBUTOR | null | 20260502T060142Z | 2026-05-02T06:01:42Z | TinyTb | 5,957,298 | MDQ6VXNlcjU5NTcyOTg= | User | false |
openclaw/openclaw | 4,091,297,552 | I_kwDOQb6kR87z3D8Q | 49,250 | https://github.com/openclaw/openclaw/issues/49250 | https://api.github.com/repos/openclaw/openclaw/issues/49250 | [Feature] Handle prompt/heartbeat race conflicts without losing the prompt | When a prompt races with a heartbeat in the same channel, the prompt currently disappears and the heartbeat is not shown immediately until a page refresh. After refresh, the prompt can appear to have been lost, even if it may eventually execute after the heartbeat. This is especially bad for long or complex prompts tha... | open | null | false | 2 | [] | [] | 2026-03-17T21:28:55Z | 2026-05-04T16:25:47Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,091,297,570 | I_kwDOQb6kR87z3D8i | 49,251 | https://github.com/openclaw/openclaw/issues/49251 | https://api.github.com/repos/openclaw/openclaw/issues/49251 | [Feature] Queue prompts when API limits prevent an immediate response | If the API limit is reached and the current prompt cannot be handled by either the current model or the fallback models, the prompt should be pushed to the queue instead of being left orphaned.
Desired behavior:
- Preserve the prompt in the queue when no model can respond because of API limits.
- Make it visible to th... | open | null | false | 2 | [] | [] | 2026-03-17T21:28:55Z | 2026-05-04T16:34:51Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,091,351,176 | I_kwDOQb6kR87z3RCI | 49,257 | https://github.com/openclaw/openclaw/issues/49257 | https://api.github.com/repos/openclaw/openclaw/issues/49257 | feat: Control UI debug mode — ephemeral connection logs and error details | ## Summary
Add a debug/diagnostic mode to the Control UI that surfaces connection logs and errors in the browser, without persisting anything to device storage after the tab/app is closed.
## Motivation
When the gateway WS connection fails (e.g. auth error, device identity issue, network problem), the UI currently s... | open | null | false | 2 | [] | [] | 2026-03-17T21:40:48Z | 2026-05-04T16:34:52Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | LiranUdi | 154,533,573 | U_kgDOCTX-xQ | User | false |
openclaw/openclaw | 4,091,386,237 | I_kwDOQb6kR87z3Zl9 | 49,259 | https://github.com/openclaw/openclaw/issues/49259 | https://api.github.com/repos/openclaw/openclaw/issues/49259 | [Feature] Prune stale orphaned sessions from Dashboard Sessions | In the Dashboard Sessions screen, add a way to delete stale orphaned sessions that are older than a configurable age and whose Telegram channel/topic or Discord channel has already been deleted.
Desired behavior:
- Detect sessions that are older than X hours and orphaned because their Telegram channel/topic or Discord... | open | null | false | 4 | [] | [] | 2026-03-17T21:50:46Z | 2026-05-04T18:21:36Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,091,457,524 | I_kwDOQb6kR87z3q_0 | 49,273 | https://github.com/openclaw/openclaw/issues/49273 | https://api.github.com/repos/openclaw/openclaw/issues/49273 | [Feature]: Change default: nativeSkills: "auto" → "none" to avoid command quota exhaustion | ### Summary
nativeSkills: "auto" registers all skills as platform commands, exhausting Telegram/Discord 100-command limit on normal installs.
### Problem to solve
With 42+ ready skills (standard bundled + a few ClawHub installs), the gateway fails to start because setMyCommands exceeds Telegram's 100-command hard li... | open | null | false | 5 | [
"enhancement"
] | [] | 2026-03-17T22:10:03Z | 2026-05-04T18:21:38Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | witness1993x | 70,125,430 | MDQ6VXNlcjcwMTI1NDMw | User | false |
openclaw/openclaw | 4,091,909,998 | I_kwDOQb6kR87z5Zdu | 49,317 | https://github.com/openclaw/openclaw/issues/49317 | https://api.github.com/repos/openclaw/openclaw/issues/49317 | [Bug]: WhatsApp group @mentions not detected when LID mentions are present (wasMentioned=false even though normalizedMentionedJids matches selfE164) | ### Bug type
Regression (worked before, now fails)
### Summary
Mention detection bug — a real bug in OpenClaw where wasMentioned always returns false because the bot's own number being in allowFrom triggers a isSelfChat code path that skips the E164 comparison. This should be reported to the OpenClaw team
ENVIRONM... | closed | completed | true | 5 | [
"bug",
"regression"
] | [] | 2026-03-18T00:20:04Z | 2026-05-01T04:40:29Z | 2026-04-29T03:39:39Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | descobarg98 | 25,942,247 | MDQ6VXNlcjI1OTQyMjQ3 | User | false |
openclaw/openclaw | 4,092,181,913 | I_kwDOQb6kR87z6b2Z | 49,350 | https://github.com/openclaw/openclaw/issues/49350 | https://api.github.com/repos/openclaw/openclaw/issues/49350 | Google Chat: markdown renders as raw text; typing indicator says 'OpenClaw is typing' instead of app name | ## Issues with Google Chat integration
Two UX issues in the Google Chat channel:
**1. Markdown formatting not stripped**
Responses containing markdown (bold, headers, etc.) render as raw asterisks/symbols in Google Chat since it doesn't support markdown. The channel adapter should strip or convert markdown to plain t... | open | null | false | 2 | [] | [] | 2026-03-18T01:46:40Z | 2026-05-04T16:35:13Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | alvin-p-stock[bot] | 268,088,027 | BOT_kgDOD_qy2w | Bot | false |
openclaw/openclaw | 4,092,295,121 | I_kwDOQb6kR87z63fR | 49,376 | https://github.com/openclaw/openclaw/issues/49376 | https://api.github.com/repos/openclaw/openclaw/issues/49376 | Feature: configurable LLM retry with backoff on transient errors (overloaded/529) | ## Summary
When Anthropic returns a transient error (HTTP 529 overloaded, 503, or 500), the gateway currently classifies it and surfaces the error to the user. There's no configurable retry-with-backoff for the main LLM request path.
## Current behavior
- Cron jobs have `cron.retry` with configurable `maxAttempts`, ... | open | null | false | 2 | [] | [] | 2026-03-18T02:25:14Z | 2026-05-04T16:35:17Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | zenfish | 2,144,880 | MDQ6VXNlcjIxNDQ4ODA= | User | false |
openclaw/openclaw | 4,092,303,603 | I_kwDOQb6kR87z65jz | 49,380 | https://github.com/openclaw/openclaw/issues/49380 | https://api.github.com/repos/openclaw/openclaw/issues/49380 | Plugin context engines cannot increment /status compaction count | ## Problem
`/status` shows `🧹 Compactions: 0` for agents using plugin context engines (e.g. mem-engine, lossless-claw), even after hundreds of successful compactions.
## Root Cause
The `compactionCount` displayed by `/status` can only be incremented through two paths:
1. **Mid-stream auto-compaction** — when the b... | open | null | false | 2 | [] | [] | 2026-03-18T02:28:20Z | 2026-05-04T16:35:19Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | petersykim | 35,867,020 | MDQ6VXNlcjM1ODY3MDIw | User | false |
openclaw/openclaw | 4,092,308,896 | I_kwDOQb6kR87z662g | 49,381 | https://github.com/openclaw/openclaw/issues/49381 | https://api.github.com/repos/openclaw/openclaw/issues/49381 | Feishu: duplicate final replies can occur after model failover from rate-limited primary | ## Summary
In Feishu direct messages, a single user message can sometimes result in **two final assistant replies** when the primary model fails with rate limiting and OpenClaw falls back to a secondary model.
This appears to be **more than a Feishu rendering problem**. Based on logs, the likely issue is in the **fai... | open | null | false | 4 | [] | [] | 2026-03-18T02:30:28Z | 2026-05-04T16:26:09Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | qwertysc | 27,398,513 | MDQ6VXNlcjI3Mzk4NTEz | User | false |
openclaw/openclaw | 4,092,462,577 | I_kwDOQb6kR87z7gXx | 49,408 | https://github.com/openclaw/openclaw/issues/49408 | https://api.github.com/repos/openclaw/openclaw/issues/49408 | [Bug]: Symbolic links for skill directories under ~/.openclaw/skills/ are not consistently resolved | ## [Description]
When adding an OpenClaw Skill by creating a symbolic link (e.g., `ln -s`) linking an external folder to the `~/.openclaw/skills/` directory, the Agent fails to recognize or load the skill during session initialization.
## [Steps to Reproduce]
1. **Prepare a physical skill directory**: Create a valid ... | open | null | false | 3 | [] | [] | 2026-03-18T03:27:47Z | 2026-05-04T16:26:12Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | RayJiang4S | 13,723,616 | MDQ6VXNlcjEzNzIzNjE2 | User | false |
openclaw/openclaw | 4,092,897,198 | I_kwDOQb6kR87z9Keu | 49,507 | https://github.com/openclaw/openclaw/issues/49507 | https://api.github.com/repos/openclaw/openclaw/issues/49507 | [Bug] Cron Job 使用错误的飞书账号发送消息 - 缺少 accountId 自动推断 | ## 问题描述
当创建 Cron Job 时，如果只设置了 `agentId` 但没有设置 `accountId`，系统会使用错误的飞书账号发送消息，导致身份混乱。
### 复现场景
1. 创建 Yaya 每日提醒 Cron Job，设置 `agentId: "yaya-claw"` 但未设置 `accountId`
2. 定时任务触发时，系统使用默认账号（可能是其他用户的账号）发送消息
3. 结果：Yaya 收到了用其他用户身份发送的消息
### 影响
- Cron Job 使用错误的飞书账号发送消息
- Cron Job 执行时使用错误的 agent 身份，导致读取数据权限错乱
## 解决方案建议
### 方案：自... | open | null | false | 3 | [] | [] | 2026-03-18T05:45:46Z | 2026-05-04T18:21:41Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ikennychoi | 21,104,264 | MDQ6VXNlcjIxMTA0MjY0 | User | false |
openclaw/openclaw | 4,092,949,645 | I_kwDOQb6kR87z9XSN | 49,517 | https://github.com/openclaw/openclaw/issues/49517 | https://api.github.com/repos/openclaw/openclaw/issues/49517 | [Bug]: /new /reset | ### Bug type
Regression (worked before, now fails)
### Summary
/new should spawn a new session without resetting the old one; /reset should be the only command that resets
### Steps to reproduce
/new should spawn a new session without resetting the old one; /reset should be the only command that resets
### Expect... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-03-18T06:00:09Z | 2026-05-04T16:26:31Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | KaysonYeh | 45,199,150 | MDQ6VXNlcjQ1MTk5MTUw | User | false |
openclaw/openclaw | 4,092,988,015 | I_kwDOQb6kR87z9gpv | 49,523 | https://github.com/openclaw/openclaw/issues/49523 | https://api.github.com/repos/openclaw/openclaw/issues/49523 | [Bug]: Session transcript mirroring writes header cwd from process.cwd(), allowing cross-agent workspace contamination | ### Bug type
Regression / workspace-isolation bug
### Summary
`src/config/sessions/transcript.ts` writes the session header `cwd` from `process.cwd()` when `appendAssistantMessageToSessionTranscript()` creates a new transcript file.
In the same runtime, other embedded-agent paths temporarily call `process.chdir(eff... | open | null | false | 3 | [] | [] | 2026-03-18T06:10:28Z | 2026-05-04T16:35:27Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | eggyrooch-blip | 226,265,806 | U_kgDODXyKzg | User | false |
openclaw/openclaw | 4,092,988,627 | I_kwDOQb6kR87z9gzT | 49,524 | https://github.com/openclaw/openclaw/issues/49524 | https://api.github.com/repos/openclaw/openclaw/issues/49524 | memory_search can stall a live session when the memory manager/tool-result path does not fail open | ## Bug type
Reliability / session-liveness bug
## Summary
A failed or stalled `memory_search` can paralyze a live session instead of degrading gracefully.
In our case, the assistant reached a `memory_search` tool call during a live Telegram turn, and the turn never completed. From the user perspective, the bot looked... | open | null | false | 2 | [] | [] | 2026-03-18T06:10:38Z | 2026-05-04T16:26:34Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | liaosvcaf | 51,533,973 | MDQ6VXNlcjUxNTMzOTcz | User | false |
openclaw/openclaw | 4,093,289,901 | I_kwDOQb6kR87z-qWt | 49,567 | https://github.com/openclaw/openclaw/issues/49567 | https://api.github.com/repos/openclaw/openclaw/issues/49567 | [Feature] Allow per-binding tool restrictions (tools.deny) in route/ACP bindings | ## Summary
There is currently no way to restrict specific tools on a per-channel or per-group basis through bindings. `RouteBindingSchema` and `AcpBindingSchema` use Zod `.strict()`, which rejects any fields beyond `agentId`, `comment`, and `match`. Attempting to add `tools.deny` or similar fields causes Gateway start... | open | null | false | 1 | [] | [] | 2026-03-18T07:26:39Z | 2026-05-04T20:22:31Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | semiok | 1,728,693 | MDQ6VXNlcjE3Mjg2OTM= | User | false |
openclaw/openclaw | 4,093,339,126 | I_kwDOQb6kR87z-2X2 | 49,577 | https://github.com/openclaw/openclaw/issues/49577 | https://api.github.com/repos/openclaw/openclaw/issues/49577 | [Bug]: Gateway/channel can appear healthy while inbound delivery is broken | ## Summary
OpenClaw can report the gateway/channel as healthy enough to trust while actual inbound message handling is effectively broken.
## Environment
- OpenClaw 2026.3.13
- macOS 26.3.1
- local gateway via LaunchAgent
- token auth
- channel in use: Feishu
## Observed behavior
I observed a state where:
- channel s... | open | null | false | 2 | [] | [] | 2026-03-18T07:38:11Z | 2026-05-04T16:26:41Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | liaoandi | 31,417,269 | MDQ6VXNlcjMxNDE3MjY5 | User | false |
openclaw/openclaw | 4,093,378,996 | I_kwDOQb6kR87z_AG0 | 49,591 | https://github.com/openclaw/openclaw/issues/49591 | https://api.github.com/repos/openclaw/openclaw/issues/49591 | [Feature]: 使用本地模型报错：400 no body | ### Summary
手动配置和把模型信息发给openclaw进行配置都报错embedded run agent end: ... isError=true error=400 status code (no body)
### Problem to solve
手动配置和把模型信息发给openclaw进行配置都报错embedded run agent end: ... isError=true error=400 status code (no body)
### Proposed solution
手动配置和把模型信息发给openclaw进行配置都报错embedded run agent end: .... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-18T07:48:18Z | 2026-05-04T16:35:35Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | cwcw0502 | 66,961,209 | MDQ6VXNlcjY2OTYxMjA5 | User | false |
openclaw/openclaw | 4,093,460,022 | I_kwDOQb6kR87z_T42 | 49,599 | https://github.com/openclaw/openclaw/issues/49599 | https://api.github.com/repos/openclaw/openclaw/issues/49599 | [Feature]: Support custom HTTP headers on CLI WebSocket connections | ### Summary
Allow passing custom HTTP headers on CLI WebSocket connections to remote gateways via `--header` flag, `OPENCLAW_GATEWAY_HEADERS` env var, and/or `gateway.remote.headers` config.
### Problem to solve
The CLI cannot connect to remote gateways behind reverse proxies that require authentication headers (e.g... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-18T08:07:53Z | 2026-05-04T16:35:37Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | alexminza | 149,919 | MDQ6VXNlcjE0OTkxOQ== | User | false |
openclaw/openclaw | 4,093,489,842 | I_kwDOQb6kR87z_bKy | 49,603 | https://github.com/openclaw/openclaw/issues/49603 | https://api.github.com/repos/openclaw/openclaw/issues/49603 | Orphaned lock files not cleared on gateway restart when PID matches current process | staleMs and isAlive(pid) checks work correctly for external crashes, but don't handle the case where the gateway process itself survives an API failure while losing in-memory HELD_LOCKS state. On restart, lock files owned by the current PID are not swept because the PID appears alive.
**Suggested fix:** on gateway boo... | open | null | false | 4 | [] | [] | 2026-03-18T08:13:32Z | 2026-05-02T01:14:36Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | NueloColab | 236,769,917 | U_kgDODhzSfQ | User | false |
openclaw/openclaw | 4,093,511,204 | I_kwDOQb6kR87z_gYk | 49,609 | https://github.com/openclaw/openclaw/issues/49609 | https://api.github.com/repos/openclaw/openclaw/issues/49609 | [Bug]: sandbox-browser noVNC clipboard garbles non-Latin-1 characters (CJK, Cyrillic, etc.) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Non-Latin-1 characters (CJK, Cyrillic, etc.) pasted via the noVNC clipboard panel in the sandbox-browser container are garbled, because x11vnc does not support the Extended Clipboard pseudo-encoding required for UTF-8 transfer over the VNC ... | open | null | false | 1 | [] | [] | 2026-03-18T08:17:14Z | 2026-05-04T16:35:40Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,093,652,949 | I_kwDOQb6kR870AC_V | 49,665 | https://github.com/openclaw/openclaw/issues/49665 | https://api.github.com/repos/openclaw/openclaw/issues/49665 | Control UI: Channels page renders empty (silent lazy-load failure) | ## Bug
The Control UI Channels page renders completely empty — no channel cards, no "Channel health" section, nothing. Only the page header ("Channels" / "Channels and settings.") is visible.
The `channels.status` WebSocket RPC returns successfully (confirmed in gateway logs), so the data exists. The issue is purely ... | open | null | false | 2 | [] | [] | 2026-03-18T08:45:54Z | 2026-05-04T18:21:43Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | zlapp | 43,241,560 | MDQ6VXNlcjQzMjQxNTYw | User | false |
openclaw/openclaw | 4,093,944,912 | I_kwDOQb6kR870BKRQ | 49,708 | https://github.com/openclaw/openclaw/issues/49708 | https://api.github.com/repos/openclaw/openclaw/issues/49708 | Session paths resolve through symlinks, causing false-positive doctor orphans | ## Bug
When `~/.openclaw` is a symlink (e.g. `~/.openclaw -> /Volumes/My Shared Files/Documents/.openclaw`), OpenClaw resolves the symlink and stores the **real path** in `sessions.json`:
```json
{
"agent:main:main": {
"sessionFile": "/Volumes/My Shared Files/Documents/.openclaw/agents/main/sessions/74c60953.js... | open | null | false | 3 | [] | [] | 2026-03-18T09:40:37Z | 2026-05-04T16:35:48Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | horry-clawd | 259,232,465 | U_kgDOD3OS0Q | User | false |
openclaw/openclaw | 4,094,270,621 | I_kwDOQb6kR870CZyd | 49,740 | https://github.com/openclaw/openclaw/issues/49740 | https://api.github.com/repos/openclaw/openclaw/issues/49740 | Feature request: cron job auto-retry on failure (--retry-count, --retry-delay) | ## Problem
When a cron job fails (e.g., upstream LLM provider is overloaded/down), the job stays in `error` state and waits for its next scheduled run. For daily crons, this means a failed 07:30 job won't retry until 07:30 the next day.
This morning, 3 daily crons all failed simultaneously because Anthropic Claude Op... | open | null | false | 3 | [] | [] | 2026-03-18T10:44:55Z | 2026-05-04T16:35:50Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | VerdiH | 20,367,733 | MDQ6VXNlcjIwMzY3NzMz | User | false |
openclaw/openclaw | 4,094,338,234 | I_kwDOQb6kR870CqS6 | 49,747 | https://github.com/openclaw/openclaw/issues/49747 | https://api.github.com/repos/openclaw/openclaw/issues/49747 | A2A Slack thread sessions can be normalized to webchat and announce fallback can route to base session | ## Summary
A2A Slack thread sessions can be normalized to `webchat` when the session key is agent-prefixed (for example `agent:builder:slack:channel:C123:thread:...`), and restart/announce fallback can still route through a base session when the current session delivery context is incomplete.
This breaks multi-agent ... | open | null | false | 2 | [] | [] | 2026-03-18T10:58:31Z | 2026-05-04T16:35:52Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | qqchang2nd | 18,712,698 | MDQ6VXNlcjE4NzEyNjk4 | User | false |
openclaw/openclaw | 4,094,457,955 | I_kwDOQb6kR870DHhj | 49,763 | https://github.com/openclaw/openclaw/issues/49763 | https://api.github.com/repos/openclaw/openclaw/issues/49763 | [Bug]: TUI loading spinner freezes when queuing message during active run | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
TUI loading spinner (circling indicator) freezes/stops animating when user sends a new message while Agent is still running. The spinner should continue showing activity but appears stuck, making user think the Agent has hung.
### Environm... | open | null | false | 2 | [] | [] | 2026-03-18T11:21:46Z | 2026-05-04T16:27:03Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | CrimsonDump | 56,749,892 | MDQ6VXNlcjU2NzQ5ODky | User | false |
openclaw/openclaw | 4,094,662,695 | I_kwDOQb6kR870D5gn | 49,776 | https://github.com/openclaw/openclaw/issues/49776 | https://api.github.com/repos/openclaw/openclaw/issues/49776 | [Bug]: WebUI Dashboard page freezes, unresponsive after static resources load | ### Bug type
Regression (worked before, now fails)
### Summary
WebUI Dashboard page freezes after loading (blank/unresponsive), but TUI works perfectly.
### Steps to reproduce
1. Start OpenClaw Gateway:
```bash
openclaw gateway
```
2. Wait for Gateway to be ready (RPC probe: ok)
3. Open browser (Chrome/Edge) a... | closed | not_planned | false | 4 | [
"bug",
"regression"
] | [] | 2026-03-18T11:58:26Z | 2026-05-04T16:45:21Z | 2026-05-04T16:45:21Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | songroad78 | 76,081,914 | MDQ6VXNlcjc2MDgxOTE0 | User | false |
openclaw/openclaw | 4,094,741,434 | I_kwDOQb6kR870EMu6 | 49,788 | https://github.com/openclaw/openclaw/issues/49788 | https://api.github.com/repos/openclaw/openclaw/issues/49788 | Add "Accessible" theme preset (colour-blind safe, high contrast) | ## Summary
Add an "Accessible" theme to the Control UI's theme picker — a high-contrast, colour-vision-deficiency-safe dark theme using only CSS variable overrides, following the same pattern as the Ocean/Spring themes proposed in #28048.
## Problem
~8% of men and ~0.5% of women have some form of colour vision defic... | open | null | false | 1 | [] | [] | 2026-03-18T12:13:43Z | 2026-05-04T16:27:13Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | InfantLab | 497,993 | MDQ6VXNlcjQ5Nzk5Mw== | User | false |
openclaw/openclaw | 4,094,943,449 | I_kwDOQb6kR870E-DZ | 49,804 | https://github.com/openclaw/openclaw/issues/49804 | https://api.github.com/repos/openclaw/openclaw/issues/49804 | Standardize async resource lifecycle management across retry/cleanup paths | ## Problem
Multiple issues (#48183, #48389, #48400) share the same root cause: async operations outlive their cleanup boundaries. Each was fixed individually with ad-hoc guards. The cleanup paths are getting harder to reason about.
## Pattern
1. Async operation starts (server close, retry timer, tool execution)
2. C... | open | null | false | 3 | [] | [] | 2026-03-18T12:53:30Z | 2026-05-04T16:27:21Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Drickon | 41,375,613 | MDQ6VXNlcjQxMzc1NjEz | User | false |
openclaw/openclaw | 4,095,188,807 | I_kwDOQb6kR870F59H | 49,828 | https://github.com/openclaw/openclaw/issues/49828 | https://api.github.com/repos/openclaw/openclaw/issues/49828 | WebChat: Sidebar agent list instead of dropdown selector | ## Feature Request
**Current behavior:**
When multiple agents are configured (e.g. Dave + Trigger), the Control UI webchat uses a dropdown menu to switch between agents.
**Desired behavior:**
A sidebar chat list (similar to WhatsApp/Telegram) where each agent appears as a separate conversation that can be clicked... | open | null | false | 2 | [] | [] | 2026-03-18T13:36:39Z | 2026-05-04T16:36:00Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | binhadher | 258,017,404 | U_kgDOD2EIfA | User | false |
openclaw/openclaw | 4,095,635,032 | I_kwDOQb6kR870Hm5Y | 49,855 | https://github.com/openclaw/openclaw/issues/49855 | https://api.github.com/repos/openclaw/openclaw/issues/49855 | Feishu: 502 Bad Gateway when downloading video files sent as 'file' type (type=file) | ## Description
When a video is sent in a Feishu/Lark group chat via iOS photo library (as opposed to direct camera recording), the message type is `file` rather than `video`. OpenClaw's Feishu channel attempts to download the resource using `type=file` in the API call:
```
GET /open-apis/im/v1/messages/{message_id}/r... | closed | completed | true | 6 | [] | [] | 2026-03-18T14:46:00Z | 2026-05-02T04:26:35Z | 2026-04-30T02:36:19Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | xbanboo | 71,201,906 | MDQ6VXNlcjcxMjAxOTA2 | User | false |
openclaw/openclaw | 4,095,745,250 | I_kwDOQb6kR870IBzi | 49,864 | https://github.com/openclaw/openclaw/issues/49864 | https://api.github.com/repos/openclaw/openclaw/issues/49864 | [Bug]: Per-topic requireMention override is ignored — group-level setting always wins | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When configuring a Telegram forum supergroup with multiple bot accounts, per-topic requireMention: false is ignored when the group-level requireMention is set to true. The expected behavior is that topic-level settings override group-level ... | closed | completed | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-18T14:59:52Z | 2026-05-04T21:53:09Z | 2026-05-04T21:53:09Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | alecjankowski | 213,700,822 | U_kgDODLzQ1g | User | false |
openclaw/openclaw | 4,095,939,871 | I_kwDOQb6kR870IxUf | 49,879 | https://github.com/openclaw/openclaw/issues/49879 | https://api.github.com/repos/openclaw/openclaw/issues/49879 | Feature: agents.defaults.responseUsage config key for global usage footer | ## Problem
`/usage full` (per-response usage footer showing tokens + cost) is per-session only. There's no way to set it as a global default — every new session starts with it off.
The only workaround is putting instructions in AGENTS.md telling the agent to call `session_status` and append a footer manually, but thi... | closed | not_planned | true | 1 | [] | [] | 2026-03-18T15:25:53Z | 2026-05-01T04:40:53Z | 2026-04-29T02:00:47Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dexbot17 | 259,017,668 | U_kgDOD3BLxA | User | false |
openclaw/openclaw | 4,096,024,799 | I_kwDOQb6kR870JGDf | 49,889 | https://github.com/openclaw/openclaw/issues/49889 | https://api.github.com/repos/openclaw/openclaw/issues/49889 | [Bug]: Telegram partial-stream finalization lacks clear delivery-mode observability (sent vs preview-finalized vs canceled) | ### Bug type
Behavior/observability bug
### Summary
With Telegram `streaming = "partial"`, final replies may complete via preview edit paths (`preview-finalized` / `preview-retained`) without clear delivery-mode visibility, making real delivery failures hard to distinguish from successful preview finalization.
### ... | open | null | false | 3 | [] | [] | 2026-03-18T15:39:12Z | 2026-05-04T16:27:34Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,096,202,479 | I_kwDOQb6kR870Jxbv | 49,909 | https://github.com/openclaw/openclaw/issues/49909 | https://api.github.com/repos/openclaw/openclaw/issues/49909 | clawdock-helpers.sh silently ignores docker-compose.override.yml on clawdock-start | ## Problem
\`clawdock-helpers.sh\` uses explicit \`-f\` flags in \`_clawdock_compose()\`, which disables Docker Compose's auto-merge of \`docker-compose.override.yml\`. The wrapper only includes:
- \`docker-compose.yml\`
- \`docker-compose.extra.yml\` (if it exists)
## Tested & Confirmed
Ran \`clawdock-stop\` + \`cl... | open | null | false | 3 | [] | [] | 2026-03-18T16:09:32Z | 2026-05-04T16:36:11Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | EdChavez | 931,547 | MDQ6VXNlcjkzMTU0Nw== | User | false |
openclaw/openclaw | 4,096,293,802 | I_kwDOQb6kR870KHuq | 49,915 | https://github.com/openclaw/openclaw/issues/49915 | https://api.github.com/repos/openclaw/openclaw/issues/49915 | [Bug]: Feishu group chat: slash commands like /new, /status, and /model do not work and may lose message content | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In a Feishu single-bot group chat with group-bound agent routing, slash commands such as /new, /status, and /model do not work.
More importantly, they do not seem to reach the agent as normal text either. From the assistant side, these mess... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-18T16:25:33Z | 2026-05-04T16:27:38Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | hjuyam | 218,579,406 | U_kgDODQdBzg | User | false |
openclaw/openclaw | 4,096,373,804 | I_kwDOQb6kR870KbQs | 49,918 | https://github.com/openclaw/openclaw/issues/49918 | https://api.github.com/repos/openclaw/openclaw/issues/49918 | [Bug]: TUI /new stopped emitting hook-visible new-session behavior after 46008178d | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to 2026.3.7+, TUI /new no longer emits hook-visible new-session behavior because it switched from
resetSession(...) to local setSession(uniqueKey).
### Steps to reproduce
1. Enable a hook that depends on TUI new-session observ... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-03-18T16:38:06Z | 2026-05-04T16:27:39Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | LonExplorer-coder | 240,868,785 | U_kgDODltdsQ | User | false |
openclaw/openclaw | 4,109,796,120 | I_kwDOQb6kR8709oMY | 51,208 | https://github.com/openclaw/openclaw/issues/51208 | https://api.github.com/repos/openclaw/openclaw/issues/51208 | [Bug]: Hook-mapped agent runs never resume codex-cli sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Mapped `/hooks/<name>` agent runs with a stable `sessionKey` persist `cliSessionIds.codex-cli` but still start a fresh `codex exec` on later hook events instead of `codex exec resume <thread_id>`.
### Steps to reproduce
1. Configure a hook m... | open | null | false | 2 | [] | [] | 2026-03-20T19:43:15Z | 2026-05-04T17:49:47Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | tennisonchan | 719,938 | MDQ6VXNlcjcxOTkzOA== | User | false |
openclaw/openclaw | 4,117,275,028 | I_kwDOQb6kR871aKGU | 52,497 | https://github.com/openclaw/openclaw/issues/52497 | https://api.github.com/repos/openclaw/openclaw/issues/52497 | feat: sessions_send timeout response should include target session's model error reason | ## Summary
When `sessions_send` times out because the target session's model is in an error state, the sender receives only `{ status: "timeout" }`. The actual cause (e.g., Gemini `429 Resource Exhausted`) is invisible to the sender, making it impossible to distinguish a routing failure from a model failure.
## Incid... | open | null | false | 3 | [] | [] | 2026-03-22T22:30:22Z | 2026-05-04T19:31:29Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,346,767,837 | I_kwDOQb6kR88AAAABAxZp3Q | 73,850 | https://github.com/openclaw/openclaw/issues/73850 | https://api.github.com/repos/openclaw/openclaw/issues/73850 | Google Meet realtime voice should allow human barge-in during assistant playback | ## Problem
In Google Meet realtime voice sessions, the assistant can be hard to interrupt while it is speaking. In practice, the human speaker may start talking but the OpenClaw Agent keeps playing queued audio, so the conversation becomes assistant-led instead of human-led.
This is especially visible with the Chrome... | closed | completed | true | 2 | [] | [] | 2026-04-28T22:58:27Z | 2026-05-04T04:37:14Z | 2026-05-01T08:01:47Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | huang47 | 157,390 | MDQ6VXNlcjE1NzM5MA== | User | false |
openclaw/openclaw | 4,346,570,757 | I_kwDOQb6kR88AAAABAxNoBQ | 73,835 | https://github.com/openclaw/openclaw/issues/73835 | https://api.github.com/repos/openclaw/openclaw/issues/73835 | Idle gateway high CPU/RSS on VPS; CPU profile points to repeated plugin/model registry + bundled runtime mirror work | ## Summary
My OpenClaw gateway becomes almost unusable on a VPS because `openclaw-gateway` idles at roughly one CPU core and RSS grows to around 1.3–1.6GB. RPC/control operations become slow; for example `node.list` calls were taking ~7.8s.
A CPU profile points at repeated model/plugin registry work:
```text
prewarmC... | closed | completed | true | 4 | [] | [] | 2026-04-28T22:12:39Z | 2026-05-02T04:29:03Z | 2026-04-29T06:52:34Z | CONTRIBUTOR | null | 20260502T060142Z | 2026-05-02T06:01:42Z | brokemac79 | 255,583,030 | U_kgDODzvjNg | User | false |
openclaw/openclaw | 4,346,759,492 | I_kwDOQb6kR88AAAABAxZJRA | 73,848 | https://github.com/openclaw/openclaw/issues/73848 | https://api.github.com/repos/openclaw/openclaw/issues/73848 | Bug: Session reset via Control UI injects blank user message, crashing MiniMax provider | ## Bug: Session reset via Control UI injects blank user message, crashing MiniMax provider
### Summary
When a session is reset via `sessions.patch` from the Control UI, the new session is created with a blank user message (`content: [{"type": "text", "text": ""}]`). This causes MiniMax to reject the request with erro... | closed | not_planned | true | 1 | [] | [] | 2026-04-28T22:56:38Z | 2026-05-01T04:41:15Z | 2026-04-28T23:09:12Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | bobgitmcgrath | 10,177,172 | MDQ6VXNlcjEwMTc3MTcy | User | false |
openclaw/openclaw | 4,346,843,393 | I_kwDOQb6kR88AAAABAxeRAQ | 73,857 | https://github.com/openclaw/openclaw/issues/73857 | https://api.github.com/repos/openclaw/openclaw/issues/73857 | Slack socket-mode disconnects under main-thread load (need worker thread or configurable pong timeout) | # Slack socket-mode disconnects under main-thread load (need worker thread or configurable pong timeout)
## Summary
OpenClaw's Slack socket-mode plugin runs on the main Node.js event loop. When the loop is busy with concurrent agent turns (LLM streaming, JSON parsing of large transcripts, trajectory writes), the Slac... | closed | completed | true | 1 | [] | [] | 2026-04-28T23:14:53Z | 2026-05-01T04:41:14Z | 2026-04-28T23:19:20Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jared-rebel | 147,876,109 | U_kgDOCNBpDQ | User | false |
openclaw/openclaw | 4,346,847,910 | I_kwDOQb6kR88AAAABAxeipg | 73,858 | https://github.com/openclaw/openclaw/issues/73858 | https://api.github.com/repos/openclaw/openclaw/issues/73858 | [Bug] Custom provider models without explicit `provider` field fail `resolveGatewayModelSupportsImages` — images offloaded as text | # [Bug] Custom provider models without explicit `provider` field fail `resolveGatewayModelSupportsImages` — images offloaded as text
## Summary
When a custom model is configured under `models.providers.<name>.models[]` without an explicit `"provider"` field on the model entry, `resolveGatewayModelSupportsImages` fails... | closed | completed | true | 1 | [] | [] | 2026-04-28T23:16:10Z | 2026-05-01T04:41:13Z | 2026-04-28T23:21:03Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | vokasug | 280,127,440 | U_kgDOELJn0A | User | false |
openclaw/openclaw | 4,346,850,652 | I_kwDOQb6kR88AAAABAxetXA | 73,859 | https://github.com/openclaw/openclaw/issues/73859 | https://api.github.com/repos/openclaw/openclaw/issues/73859 | [Bug]: Built-in plugins (minimax, google, talk-voice) fail with RangeError: Maximum call stack size exceeded on Windows | ## Describe the bug
Multiple built-in plugins fail to load at Gateway startup with `RangeError: Maximum call stack size exceeded`. This affects the following plugins:
- `minimax`
- `google`
- `talk-voice`
## Error Log
```
[plugins] minimax failed to load from .../openclaw/dist/extensions/minimax/index.js: RangeErro... | open | null | false | 1 | [] | [] | 2026-04-28T23:16:53Z | 2026-05-03T06:07:48Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | desksk | 26,273,255 | MDQ6VXNlcjI2MjczMjU1 | User | false |
openclaw/openclaw | 4,346,853,840 | I_kwDOQb6kR88AAAABAxe50A | 73,861 | https://github.com/openclaw/openclaw/issues/73861 | https://api.github.com/repos/openclaw/openclaw/issues/73861 | [Bug]: sessions_send intermittent timeout and gateway closed failures despite Gateway running normally | ## Describe the bug
The `sessions_send` mechanism experiences intermittent failures, manifesting as:
- Persistent timeouts when sending messages to other agent sessions
- "gateway closed (1000 normal closure)" errors
- Connection drops mid-transmission
The Gateway itself reports healthy status (listening on port, RPC... | open | null | false | 1 | [] | [] | 2026-04-28T23:17:34Z | 2026-05-03T07:05:28Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | desksk | 26,273,255 | MDQ6VXNlcjI2MjczMjU1 | User | false |
openclaw/openclaw | 4,346,852,455 | I_kwDOQb6kR88AAAABAxe0Zw | 73,860 | https://github.com/openclaw/openclaw/issues/73860 | https://api.github.com/repos/openclaw/openclaw/issues/73860 | [Bug]: message tool returns 400 error for Feishu channel user targets | ## Describe the bug
The `message` tool consistently returns HTTP 400 (Bad Request) errors when sending to specific Feishu user targets, while group chat sends work normally.
## Error Log
```
[tools] message failed: Request failed with status code 400
raw_params={action:send,channel:feishu,target:user:ou_xxx...,messa... | closed | completed | true | 4 | [] | [] | 2026-04-28T23:17:16Z | 2026-05-05T04:28:50Z | 2026-05-02T04:53:09Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | desksk | 26,273,255 | MDQ6VXNlcjI2MjczMjU1 | User | false |
openclaw/openclaw | 4,346,854,810 | I_kwDOQb6kR88AAAABAxe9mg | 73,862 | https://github.com/openclaw/openclaw/issues/73862 | https://api.github.com/repos/openclaw/openclaw/issues/73862 | [Bug]: Gateway memory and connection accumulation over extended runtime on Windows | ## Describe the bug
OpenClaw Gateway process exhibits gradual memory and resource accumulation over extended runtime, leading to degraded performance and potential instability.
## Observed Behavior
After extended uptime (24+ hours):
- Single Node.js process memory usage grows to ~1GB+
- Accumulation of TIME_WAIT net... | closed | not_planned | true | 1 | [] | [] | 2026-04-28T23:17:52Z | 2026-05-01T04:41:12Z | 2026-04-28T23:22:20Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | desksk | 26,273,255 | MDQ6VXNlcjI2MjczMjU1 | User | false |
openclaw/openclaw | 4,346,855,790 | I_kwDOQb6kR88AAAABAxfBbg | 73,863 | https://github.com/openclaw/openclaw/issues/73863 | https://api.github.com/repos/openclaw/openclaw/issues/73863 | Feature request: Enhanced logging and configurable log rotation | ## Feature Request: Enhanced logging and log rotation
### Problem
Current OpenClaw logging has several limitations that make troubleshooting difficult:
1. **Insufficient detail**: Runtime logs lack sufficient context for diagnosing issues
- Missing connection establishment/termination timestamps
- Limited erro... | closed | completed | true | 1 | [] | [] | 2026-04-28T23:18:06Z | 2026-05-01T04:41:11Z | 2026-04-28T23:22:59Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | desksk | 26,273,255 | MDQ6VXNlcjI2MjczMjU1 | User | false |
openclaw/openclaw | 4,346,872,380 | I_kwDOQb6kR88AAAABAxgCPA | 73,864 | https://github.com/openclaw/openclaw/issues/73864 | https://api.github.com/repos/openclaw/openclaw/issues/73864 | Feature: let openclaw mcp serve request configurable operator scopes | ## Summary
Please add a supported way for `openclaw mcp serve` to request a narrower operator scope set, such as `--scope <scope...>`, a config/env equivalent, or explicit MCP modes like `read`, `write`, and `approvals`.
## Problem
`openclaw mcp serve` is the stdio bridge process used by external MCP clients. Today ... | open | null | false | 2 | [] | [] | 2026-04-28T23:22:25Z | 2026-05-03T03:37:49Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | mstr-six | 169,169,413 | U_kgDOChVSBQ | User | false |
openclaw/openclaw | 4,346,875,416 | I_kwDOQb6kR88AAAABAxgOGA | 73,865 | https://github.com/openclaw/openclaw/issues/73865 | https://api.github.com/repos/openclaw/openclaw/issues/73865 | [Bug]: Gateway startup hangs indefinitely at pre-sidecar readiness path (never reaches 'ready') | ## Description
OpenClaw gateway container hangs during startup and never becomes healthy. Startup trace shows it reaches `runtime.early` but never progresses to `sidecars.*`, `agent model`, or `ready` log events. The container stays hung indefinitely.
## Environment
- **Image**: `ghcr.io/openclaw/openclaw:2026.4.24`... | closed | completed | true | 3 | [] | [] | 2026-04-28T23:23:18Z | 2026-05-02T04:26:37Z | 2026-04-30T02:35:23Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | lpendeavors | 10,328,090 | MDQ6VXNlcjEwMzI4MDkw | User | false |
openclaw/openclaw | 4,068,387,176 | I_kwDOQb6kR87yfqlo | 44,551 | https://github.com/openclaw/openclaw/issues/44551 | https://api.github.com/repos/openclaw/openclaw/issues/44551 | [Feature]: [Feature Request] Gateway UI 支持中文显示 | ### Summary
## Feature Description
希望 Gateway 仪表盘的界面能够支持中文显示，方便中文用户使用。
## Use Case
当前 Gateway UI 的 Audit Log、导航菜单、按钮文字等都是英文，对于中文用户来说阅读理解成本较高。
## Expected Behavior
希望能够提供以下中文支持：
1. **Audit Log（审计日志）** - 将操作名翻译为中文，例如：
- `exec` → "执行命令"
- `read` → "读取文件"
- `cron run` → "定时任务"
- `subagents list` → "子代理列表... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-13T02:27:48Z | 2026-05-04T13:18:39Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ytb000105 | 203,129,756 | U_kgDODBuDnA | User | false |
openclaw/openclaw | 4,074,715,272 | I_kwDOQb6kR87y3ziI | 45,758 | https://github.com/openclaw/openclaw/issues/45758 | https://api.github.com/repos/openclaw/openclaw/issues/45758 | Feature Request: Support YAML as config file format | ## Summary
Add support for YAML as an alternative config file format alongside the current JSON5 format (`~/.openclaw/openclaw.json`).
## Motivation
- **Readability**: YAML is widely used in DevOps/tooling configs (Docker Compose, Kubernetes, Ansible, etc.) and is often more readable than JSON5 for complex nested co... | open | null | false | 6 | [] | [] | 2026-03-14T05:13:17Z | 2026-05-04T00:07:49Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | xuxusheng | 15,248,446 | MDQ6VXNlcjE1MjQ4NDQ2 | User | false |
openclaw/openclaw | 4,083,096,363 | I_kwDOQb6kR87zXxsr | 48,300 | https://github.com/openclaw/openclaw/issues/48300 | https://api.github.com/repos/openclaw/openclaw/issues/48300 | Bug: memory_search hybrid mode not returning FTS matches | ## Summary
When using `memory_search` with hybrid mode enabled, FTS (full-text search) matches are not being returned in results, even when they exist in the SQLite FTS index.
## Steps to Reproduce
1. Have a memory file containing specific keyword (e.g., `memory/persona.md` containing "Clyde")
2. Run `openclaw memor... | open | null | false | 3 | [] | [] | 2026-03-16T15:14:38Z | 2026-05-01T01:22:27Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | sabo961 | 191,984 | MDQ6VXNlcjE5MTk4NA== | User | false |
openclaw/openclaw | 4,099,082,654 | I_kwDOQb6kR870Uwme | 50,178 | https://github.com/openclaw/openclaw/issues/50178 | https://api.github.com/repos/openclaw/openclaw/issues/50178 | [Bug]: 400 thinking is enabled, but reasoning_content is missing in assistant tool call message at index 12 | ### Bug type
Regression (worked before, now fails)
### Summary
When using the Baidu Qianfan `qianfan-code-latest` model, every tool call triggers this error. The model appears to support thinking/reasoning functionality, but OpenClaw fails to handle the `reasoning_content` field properly during tool call processing.... | open | null | false | 3 | [
"bug",
"regression"
] | [] | 2026-03-19T03:39:28Z | 2026-05-04T20:14:47Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | AlexBenjarmin | 134,087,136 | U_kgDOB_4B4A | User | false |
openclaw/openclaw | 4,107,229,667 | I_kwDOQb6kR870z1nj | 50,996 | https://github.com/openclaw/openclaw/issues/50996 | https://api.github.com/repos/openclaw/openclaw/issues/50996 | Feature Request: Interruptible tool calls — agent unresponsive during long process(poll) | ## Title: Feature Request: Interruptible tool calls — agent unresponsive during long process(poll)
### Problem
When an agent calls `process(action: "poll", timeout: 300000)` to wait for a background process, the entire session is blocked for the duration. No incoming messages from the user can be processed until the ... | closed | not_planned | true | 3 | [] | [] | 2026-03-20T11:01:42Z | 2026-05-03T04:34:02Z | 2026-05-01T04:14:57Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | dennis-autosend | 180,860,676 | U_kgDOCse3BA | User | false |
openclaw/openclaw | 4,107,520,160 | I_kwDOQb6kR87008ig | 51,028 | https://github.com/openclaw/openclaw/issues/51028 | https://api.github.com/repos/openclaw/openclaw/issues/51028 | Sessions panel: sort by last meaningful activity, not just last message | ## Summary
The sessions panel should sort sessions by **last meaningful activity**, not last message timestamp. Right now it's opaque — you can't tell why a session is at the top, and heartbeat noise displaces real conversations.
## Desired Behavior
Sessions bubble to the top based on `lastMeaningfulActivityAt`, whi... | open | null | false | 4 | [] | [] | 2026-03-20T12:08:15Z | 2026-05-04T17:49:02Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | TSCarterJr | 82,057,175 | MDQ6VXNlcjgyMDU3MTc1 | User | false |
openclaw/openclaw | 4,107,671,050 | I_kwDOQb6kR8701hYK | 51,040 | https://github.com/openclaw/openclaw/issues/51040 | https://api.github.com/repos/openclaw/openclaw/issues/51040 | Voice wake mode should respect selected session target in macOS app | ## Description
When using voice wake mode on the macOS node app, responses always route to the main webchat session regardless of which session target is selected in the app's dropdown.
## Expected Behavior
Voice wake responses should route to whatever session the user has selected in the macOS app's session picker ... | closed | completed | true | 3 | [] | [] | 2026-03-20T12:41:48Z | 2026-05-04T04:34:41Z | 2026-05-02T01:55:10Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | carl-jeffrolc | 258,705,842 | U_kgDOD2uJsg | User | false |
openclaw/openclaw | 4,107,685,837 | I_kwDOQb6kR8701k_N | 51,041 | https://github.com/openclaw/openclaw/issues/51041 | https://api.github.com/repos/openclaw/openclaw/issues/51041 | [Feature]: Expose Discord slash interaction response controls to plugin commands | ### Summary
Add stuff like ctx/follow up/editing replies to the api so we can edit command outputs for discord
### Problem to solve
In native OpenClaw plugin commands (`api.registerCommand({ handler(ctx) { ... } })`) on Discord, the handler only receives `PluginCommandContext` and returns `ReplyPayload`.
That means... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-20T12:45:05Z | 2026-05-04T17:49:04Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | TwangyMoney | 87,775,970 | MDQ6VXNlcjg3Nzc1OTcw | User | false |
openclaw/openclaw | 4,107,815,306 | I_kwDOQb6kR8702EmK | 51,054 | https://github.com/openclaw/openclaw/issues/51054 | https://api.github.com/repos/openclaw/openclaw/issues/51054 | Media Understanding config lacks apiKey source documentation, causing setup failure | ## Problem Description
When configuring `tools.media` for audio/video understanding following the [Media Understanding docs](https://docs.openclaw.ai/nodes/media-understanding), the feature fails to work because it cannot find the apiKey. The documentation does not explain where the apiKey should be configured.
## Ste... | open | null | false | 2 | [] | [] | 2026-03-20T13:12:39Z | 2026-05-04T17:49:08Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | shineclaw | 265,269,485 | U_kgDOD8-w7Q | User | false |
openclaw/openclaw | 4,107,855,092 | I_kwDOQb6kR8702OT0 | 51,057 | https://github.com/openclaw/openclaw/issues/51057 | https://api.github.com/repos/openclaw/openclaw/issues/51057 | Timeouts should mark auth profiles as failed to trigger faster model fallback | ## Problem
When the primary model provider times out (e.g. OpenAI Codex outage), the failover loop retries for the entire `agents.defaults.timeoutSeconds` budget (default 600s) before falling back — even though the provider is clearly down.
**Root cause**: In the model selection loop, timeouts are explicitly excluded... | open | null | false | 2 | [] | [] | 2026-03-20T13:19:51Z | 2026-05-04T17:49:10Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | YoanWai | 106,609,173 | U_kgDOBlq6FQ | User | false |
openclaw/openclaw | 4,108,428,334 | I_kwDOQb6kR8704aQu | 51,104 | https://github.com/openclaw/openclaw/issues/51104 | https://api.github.com/repos/openclaw/openclaw/issues/51104 | [Bug]: Mattermost WebSocket add ping/pong keepalive to detect half-dead connections | ## Problem
The Mattermost channel plugin uses a plain WebSocket connection (`monitor-websocket.ts` → `createMattermostConnectOnce`) that does **not** send WebSocket-level ping frames. This means half-dead connections (TCP ESTABLISHED but no application-layer data flow) go undetected until the `staleEventThresholdMs` t... | closed | completed | true | 4 | [] | [] | 2026-03-20T15:06:25Z | 2026-05-02T04:26:37Z | 2026-04-30T02:34:32Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | meredith620 | 1,277,190 | MDQ6VXNlcjEyNzcxOTA= | User | false |
openclaw/openclaw | 4,109,517,294 | I_kwDOQb6kR8708kHu | 51,184 | https://github.com/openclaw/openclaw/issues/51184 | https://api.github.com/repos/openclaw/openclaw/issues/51184 | Feature: Surface cron job name / session label in /status and statusline | ## Problem
When viewing `/status` in the TUI, cron job sessions and their runs display as raw UUIDs:
```
agent:main:cron:b03cbb21-9519-416c-8af0-effe954ae672:run:446baffc-1d47-4208-994b-a2f2b54dd7a5
```
The session store already has a `label` field (e.g., `"Cron: claude-keepalive"`) and cron jobs have a `name` field... | open | null | false | 3 | [] | [] | 2026-03-20T18:40:48Z | 2026-05-04T18:47:09Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Baelson | 7,864,962 | MDQ6VXNlcjc4NjQ5NjI= | User | false |
openclaw/openclaw | 4,109,810,404 | I_kwDOQb6kR8709rrk | 51,245 | https://github.com/openclaw/openclaw/issues/51245 | https://api.github.com/repos/openclaw/openclaw/issues/51245 | [Bug]: Telegram slash sessions still resolve to channel=(unknown), causing elevated allowFrom checks to fail on 2026.3.13 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Telegram slash sessions are still not being recognized correctly for elevated `allowFrom` checks on OpenClaw 2026.3.13.
Even with `tools.elevated.enabled=true` and my Telegram sender ID explicitly allowlisted, `openclaw sandbox explain` sho... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-20T19:46:42Z | 2026-05-04T17:49:49Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Tin55FoilDev | 159,156,864 | U_kgDOCXyKgA | User | false |
openclaw/openclaw | 4,109,845,882 | I_kwDOQb6kR87090V6 | 51,251 | https://github.com/openclaw/openclaw/issues/51251 | https://api.github.com/repos/openclaw/openclaw/issues/51251 | Session modelOverride persists across gateway restarts, silently overrides config default | ## Bug type
Behavior bug (incorrect state without crash)
## Summary
When a session-level model override is set via `/model`, it persists in the session store across gateway restarts. After changing the default model in `openclaw.json` and restarting, existing (and even `/new`-ed) sessions still use the stale overrid... | open | null | false | 2 | [] | [] | 2026-03-20T19:55:33Z | 2026-05-04T20:15:37Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Ike-li | 50,747,621 | MDQ6VXNlcjUwNzQ3NjIx | User | false |
openclaw/openclaw | 4,110,027,374 | I_kwDOQb6kR870-gpu | 51,264 | https://github.com/openclaw/openclaw/issues/51264 | https://api.github.com/repos/openclaw/openclaw/issues/51264 | OOM crash: loadCombinedSessionStoreForGateway loads all agent session stores simultaneously | # OOM crash: loadCombinedSessionStoreForGateway loads all agent session stores simultaneously
## Summary
OpenClaw's gateway crashes with an out-of-memory (OOM) error when `loadCombinedSessionStoreForGateway` is called during hook dispatch. The function eagerly loads **every agent's full `sessions.json`** into memory ... | open | null | false | 3 | [] | [] | 2026-03-20T20:44:44Z | 2026-05-04T17:49:53Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | mband0 | 31,228,859 | MDQ6VXNlcjMxMjI4ODU5 | User | false |
openclaw/openclaw | 4,110,217,957 | I_kwDOQb6kR870_PLl | 51,287 | https://github.com/openclaw/openclaw/issues/51287 | https://api.github.com/repos/openclaw/openclaw/issues/51287 | [Feature]: Configurable Exec Approval timeout | ### Summary
Allow user to override default timeouts for Exec Approvals.
### Problem to solve
The current two minute approval timeout is short enough that it is easy to miss.
### Proposed solution
Expose the DEFAULT_APPROVAL_TIMEOUT_MS and DEVAULT_APPROVAL_REQUEST_TIMEOUT_MS settings (found in /usr/local/lib/node_m... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-20T21:31:13Z | 2026-05-04T20:15:39Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | fyzycyst | 61,153,219 | MDQ6VXNlcjYxMTUzMjE5 | User | false |
openclaw/openclaw | 4,110,283,866 | I_kwDOQb6kR870_fRa | 51,299 | https://github.com/openclaw/openclaw/issues/51299 | https://api.github.com/repos/openclaw/openclaw/issues/51299 | iMessage channel: LaunchAgent can't get FDA/Automation on macOS Sequoia (SIP) | ## Environment
- macOS 26.3.1 (Tahoe/Sequoia)
- SIP enabled
- OpenClaw gateway running as LaunchAgent (`ai.openclaw.gateway`)
- imsg 0.5.0 (Homebrew)
- nix-openclaw flake setup (home-manager)
## Problem
The iMessage channel fails with `permissionDenied` on `chat.db` (FDA) and `-1743` on send (AppleEvents/Automation).... | open | null | false | 3 | [] | [] | 2026-03-20T21:50:01Z | 2026-05-04T18:54:37Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | markthebest12 | 6,416,874 | MDQ6VXNlcjY0MTY4NzQ= | User | false |
openclaw/openclaw | 4,110,649,832 | I_kwDOQb6kR871A4no | 51,336 | https://github.com/openclaw/openclaw/issues/51336 | https://api.github.com/repos/openclaw/openclaw/issues/51336 | Feature Request: Surface API provider name in error/overload messages | ## Problem
When the Anthropic API returns a 529 (overloaded) or similar provider error, OpenClaw forwards a generic message to the user (e.g. "AI services temporarily overloaded"). It is not clear from this message whether the problem is with the upstream AI provider, the OpenClaw gateway, or the local server.
## Req... | open | null | false | 4 | [] | [] | 2026-03-20T23:56:06Z | 2026-05-04T17:50:12Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | wildhuman | 18,494,004 | MDQ6VXNlcjE4NDk0MDA0 | User | false |
openclaw/openclaw | 4,110,922,373 | I_kwDOQb6kR871B7KF | 51,363 | https://github.com/openclaw/openclaw/issues/51363 | https://api.github.com/repos/openclaw/openclaw/issues/51363 | [Bug] Docker sandbox container name collision across multiple instances on same host | ## Environment
- OpenClaw version: v2026.3.13 (61d171a)
- OS: Ubuntu 24.04.3
- Docker: installed, sandbox enabled via `agents.defaults.sandbox`
- Setup: 5+ OpenClaw instances on the same host, isolated via `HOME` env var
## Description
When multiple OpenClaw instances run on the same host with Docker sandbox enabled, ... | open | null | false | 5 | [] | [] | 2026-03-21T01:30:37Z | 2026-05-04T17:50:24Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | wnbhr | 264,616,070 | U_kgDOD8W4hg | User | false |
openclaw/openclaw | 4,111,078,847 | I_kwDOQb6kR871ChW_ | 51,386 | https://github.com/openclaw/openclaw/issues/51386 | https://api.github.com/repos/openclaw/openclaw/issues/51386 | feat(memory): graduate sessionMemory from experimental — solve the multi-session continuity problem | ## Problem
When a user talks to their agent on webchat and switches to Telegram, the agent on Telegram has no knowledge of the webchat conversation. It feels like two separate people — "twin brothers who don't share memory."
The fix already exists: `experimental.sessionMemory`, which automatically indexes all session... | open | null | false | 2 | [] | [] | 2026-03-21T02:40:49Z | 2026-05-04T18:47:17Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | DeekRoumy | 269,637,962 | U_kgDOEBJZSg | User | false |
openclaw/openclaw | 4,111,149,410 | I_kwDOQb6kR871Cyli | 51,393 | https://github.com/openclaw/openclaw/issues/51393 | https://api.github.com/repos/openclaw/openclaw/issues/51393 | Feature: browser tool evaluate should support statements (var/let/const) | ## Feature Request
`browser act kind=evaluate` currently only accepts expressions, not statements. This means `var`, `let`, `const` declarations cause a syntax error:
`Error: Invalid evaluate function: Unexpected token 'var'`
## Current Workaround
Wrap in IIFE:
```javascript
(()=>{ var x = new XMLHttpRequest(); ...... | open | null | false | 2 | [] | [] | 2026-03-21T03:03:58Z | 2026-05-04T18:47:19Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | cpc1986 | 17,445,005 | MDQ6VXNlcjE3NDQ1MDA1 | User | false |
openclaw/openclaw | 4,111,150,332 | I_kwDOQb6kR871Cyz8 | 51,395 | https://github.com/openclaw/openclaw/issues/51395 | https://api.github.com/repos/openclaw/openclaw/issues/51395 | Feature: browser upload tool fallback for non-standard file inputs (ByteUploader) | ## Feature Request
The browser `upload` tool relies on `DOM.setFileInputFiles` CDP method, which only works with standard `<input type="file">` elements.
ByteDance products (抖音, 番茄达人中心, etc.) use a custom `ByteUploader` component that bypasses the standard file input mechanism. When calling `upload`, it returns OK b... | open | null | false | 1 | [] | [] | 2026-03-21T03:04:16Z | 2026-05-04T18:47:21Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | cpc1986 | 17,445,005 | MDQ6VXNlcjE3NDQ1MDA1 | User | false |
openclaw/openclaw | 4,111,152,517 | I_kwDOQb6kR871CzWF | 51,396 | https://github.com/openclaw/openclaw/issues/51396 | https://api.github.com/repos/openclaw/openclaw/issues/51396 | [Bug]: clearUnboundScopes strips operator scopes unconditionally for non-local token-auth clients | ### Bug type
Regression (worked before, now fails)
### Summary
In 2026.3.13, clearUnboundScopes() unconditionally strips all operator scopes for non-local clients without device identity, even when token auth succeeds. This breaks chat.send (and any operator.write-scoped method) for backend clients connecting over a... | open | null | false | 7 | [
"bug",
"regression"
] | [] | 2026-03-21T03:04:54Z | 2026-05-04T17:50:35Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | marccampbell | 173,451 | MDQ6VXNlcjE3MzQ1MQ== | User | false |
openclaw/openclaw | 4,111,466,090 | I_kwDOQb6kR871D_5q | 51,429 | https://github.com/openclaw/openclaw/issues/51429 | https://api.github.com/repos/openclaw/openclaw/issues/51429 | [Bug]: 看起来有人把工作路径hardcode进代码里而且居然被合并发布了 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
20260321，今天刚安装的，最新版，结果openclaw建了一个/Users/wangtao的文件夹，并且把工作区设成了这个目录。
这位wangtao是谁？
Apparently some wangtao hardcode his working space path into the code and somebody merged his code and published, now openclaw just mkdir /Users/wangtao then ... | open | null | false | 11 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T05:17:45Z | 2026-05-04T20:15:44Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | buggiant-coder | 230,094,344 | U_kgDODbb2CA | User | false |
openclaw/openclaw | 4,111,531,110 | I_kwDOQb6kR871EPxm | 51,441 | https://github.com/openclaw/openclaw/issues/51441 | https://api.github.com/repos/openclaw/openclaw/issues/51441 | feat: expose resolved backend model in session_status and agent runtime | ## Problem
When using LiteLLM (or any routing proxy) as the LLM provider, agents only see the **requested model alias** (e.g. `litellm/complex`) — not the **actual backend model** that was used (e.g. `openai/gpt-5.4` or `anthropic/claude-sonnet-4-6`).
This creates a blind spot: agents can't answer "did a fallback hap... | open | null | false | 3 | [] | [] | 2026-03-21T05:48:06Z | 2026-05-04T18:47:22Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Kyzcreig | 9,063,726 | MDQ6VXNlcjkwNjM3MjY= | User | false |
openclaw/openclaw | 4,111,591,995 | I_kwDOQb6kR871Eeo7 | 51,458 | https://github.com/openclaw/openclaw/issues/51458 | https://api.github.com/repos/openclaw/openclaw/issues/51458 | Slack: message action rejects file attachments when interactive elements are present | ## Bug Description
The Slack `send` message action rejects file attachments (`media` parameter) when `interactive` elements (buttons, selects, etc.) are present in the action parameters. The `interactive` parameter is automatically converted to `blocks` via `buildSlackInteractiveBlocks()`, and then the validation at `... | closed | completed | true | 3 | [] | [] | 2026-03-21T06:17:54Z | 2026-05-04T04:33:57Z | 2026-05-02T04:17:32Z | CONTRIBUTOR | null | 20260504T060031Z | 2026-05-04T06:00:31Z | HirokiKobayashi-R | 37,167,840 | MDQ6VXNlcjM3MTY3ODQw | User | false |
openclaw/openclaw | 4,111,766,112 | I_kwDOQb6kR871FJJg | 51,497 | https://github.com/openclaw/openclaw/issues/51497 | https://api.github.com/repos/openclaw/openclaw/issues/51497 | [Bug]: WebChat 会话消息丢失 - Token消耗存在但消息记录消失 | ### Bug type
Regression (worked before, now fails)
### Summary
## 问题描述
用户在 QClaw WebChat 界面中的多条消息在会话中凭空消失，但 Token 消耗记录显示这些消息确实被处理过。此问题已连续两天发生。
## 证据
### 事件 1: 2026-03-21
**会话 ID**: `663ae372-2c69-4417-a607-c45210c35489`
**文件路径**: `C:\Users\daishigong\.qclaw\agents\main\sessions\663ae372-2c69-4417-a607-c45210c354... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-03-21T07:59:02Z | 2026-05-04T17:50:45Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | mydaisg | 164,868,826 | U_kgDOCdOy2g | User | false |
openclaw/openclaw | 4,111,964,079 | I_kwDOQb6kR871F5ev | 51,534 | https://github.com/openclaw/openclaw/issues/51534 | https://api.github.com/repos/openclaw/openclaw/issues/51534 | [Feature]: Discord - auto-inject @mention for guild message replies | ## Summary
In a multi-agent Discord setup with `replyToMode: all` and `requireMention: true`, agents often fail to include `<@userID>` mentions in their replies. This means the original message author does not receive a Discord notification, even though the bot is replying to them.
## Problem
When an agent replies t... | open | null | false | 2 | [] | [] | 2026-03-21T09:50:51Z | 2026-05-04T20:15:48Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ljdljdljdljd | 262,889,315 | U_kgDOD6tfYw | User | false |
openclaw/openclaw | 4,112,020,969 | I_kwDOQb6kR871GHXp | 51,549 | https://github.com/openclaw/openclaw/issues/51549 | https://api.github.com/repos/openclaw/openclaw/issues/51549 | [Bug]: WebChat loses message queue, conversation history, and draft on browser refresh | ### Summary
WebChat UI loses all state on browser refresh (F5): pending message queue, conversation history, and draft text in the input box. This was previously reported in #8331 but auto-closed as stale without a fix. Still reproducible in 2026.3.13.
### Environment
- OpenClaw version: 2026.3.13 (61d171a)
- OS: Wi... | open | null | false | 3 | [] | [] | 2026-03-21T10:28:12Z | 2026-05-04T17:50:56Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ukr-coder | 246,043,930 | U_kgDODqpVGg | User | false |
openclaw/openclaw | 4,112,178,243 | I_kwDOQb6kR871GtxD | 51,572 | https://github.com/openclaw/openclaw/issues/51572 | https://api.github.com/repos/openclaw/openclaw/issues/51572 | Feature: fire session-memory hook on session reset/prune, not just compaction | ## Problem
The `session-memory` hook currently only fires during **auto-compaction** (when context window fills up). It does NOT fire when sessions are:
- Reset by idle timeout (`idleMinutes`)
- Reset by daily reset (`atHour`)
- Pruned by maintenance (`pruneAfter`)
This means session context is silently lost when ses... | open | null | false | 2 | [] | [] | 2026-03-21T11:28:00Z | 2026-05-04T19:11:12Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | martins-oss | 224,328,417 | U_kgDODV764Q | User | false |
openclaw/openclaw | 4,112,246,971 | I_kwDOQb6kR871G-i7 | 51,586 | https://github.com/openclaw/openclaw/issues/51586 | https://api.github.com/repos/openclaw/openclaw/issues/51586 | Feature: allow overriding /new session reset prompt via config | ## Problem
When `/new` or `/reset` is triggered, OpenClaw injects a hardcoded prompt:
> A new session was started via /new or /reset. Run your Session Startup sequence - read the required files before responding to the user. Then greet the user in your configured persona, if one is provided. Be yourself - use your de... | closed | not_planned | true | 2 | [] | [] | 2026-03-21T12:03:10Z | 2026-05-04T04:37:25Z | 2026-05-01T05:27:24Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | david6ca | 45,101,457 | MDQ6VXNlcjQ1MTAxNDU3 | User | false |
openclaw/openclaw | 4,112,254,502 | I_kwDOQb6kR871HAYm | 51,587 | https://github.com/openclaw/openclaw/issues/51587 | https://api.github.com/repos/openclaw/openclaw/issues/51587 | Bug: isRecurringTopOfHourCronExpr fails to match cron expressions with step values | ## Bug Report: `isRecurringTopOfHourCronExpr` fails to match cron expressions with step values
### Summary
The `isRecurringTopOfHourCronExpr` function in `src/cron/stagger.ts` only matches cron expressions where the minute field is exactly `"0"`. This causes it to incorrectly return `false` for valid recurring top-of-... | open | null | false | 2 | [] | [] | 2026-03-21T12:05:51Z | 2026-05-04T21:17:48Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,112,549,437 | I_kwDOQb6kR871IIY9 | 51,628 | https://github.com/openclaw/openclaw/issues/51628 | https://api.github.com/repos/openclaw/openclaw/issues/51628 | Bug: Telegram delivery queue recovery can replay old sendMessage replies and duplicate delivery-mirror transcript entries | ## Bug Description
Telegram outbound replies can be delivered again later after an earlier `sendMessage` network failure, and the replay can also append a duplicate `delivery-mirror` assistant message into the source session transcript.
This appears to be a remaining edge case beyond the existing fixes around Telegra... | open | null | false | 5 | [] | [] | 2026-03-21T13:33:25Z | 2026-05-04T20:23:13Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Pinghuachiu | 9,033,138 | MDQ6VXNlcjkwMzMxMzg= | User | false |
openclaw/openclaw | 4,112,778,524 | I_kwDOQb6kR871JAUc | 51,664 | https://github.com/openclaw/openclaw/issues/51664 | https://api.github.com/repos/openclaw/openclaw/issues/51664 | WebChat copy button doesn't work in Edge browser | In Edge browser, clicking the "Copy" button on code blocks doesn't copy content to clipboard.
User has to manually select text and press Ctrl+C to copy.
Steps to reproduce:
1. Open WebChat in Edge browser
2. Receive a message with code block
3. Click the "Copy" button above the code
4. Content is not copied to clipbo... | open | null | false | 1 | [] | [] | 2026-03-21T14:58:48Z | 2026-05-04T20:59:09Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | cipherzw | 23,001,064 | MDQ6VXNlcjIzMDAxMDY0 | User | false |
openclaw/openclaw | 4,112,787,414 | I_kwDOQb6kR871JCfW | 51,667 | https://github.com/openclaw/openclaw/issues/51667 | https://api.github.com/repos/openclaw/openclaw/issues/51667 | Feature: Native Audio Input for Omni-Modal Models (skip STT transcription) | ## Summary
When using omni-modal models (e.g. `mimo-v2-omni`, Gemini 3 Flash) that natively support audio input, OpenClaw should be able to send voice notes directly to the model instead of routing them through Whisper/transcription first.
## Current Behavior
- User sends a voice note (WhatsApp/Telegram/etc.)
- Open... | open | null | false | 3 | [] | [] | 2026-03-21T15:02:11Z | 2026-05-04T20:59:11Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jojo2a | 268,068,222 | U_kgDOD_plfg | User | false |
openclaw/openclaw | 4,112,874,310 | I_kwDOQb6kR871JXtG | 51,694 | https://github.com/openclaw/openclaw/issues/51694 | https://api.github.com/repos/openclaw/openclaw/issues/51694 | [Bug]: channels.googlechat.accounts is now non-optional, docs indicate optional | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In https://github.com/openclaw/openclaw/pull/46787 we made channels.googlechat.accounts.*.appPrincipal non-optional. the code requires it.
However, the docs show both this field (and channels.googlechats.accounts as a whole) as optional (re... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T15:42:37Z | 2026-05-04T19:11:20Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | donbowman | 5,131,923 | MDQ6VXNlcjUxMzE5MjM= | User | false |
openclaw/openclaw | 4,113,293,703 | I_kwDOQb6kR871K-GH | 51,805 | https://github.com/openclaw/openclaw/issues/51805 | https://api.github.com/repos/openclaw/openclaw/issues/51805 | Feature Request: Shared session context between group chats and DMs | ## Summary
Currently, group chats and DMs always have isolated sessions (`group:<key>` vs `direct:<key>`). There is no way to share conversation context between a group chat and a private chat with the same user.
## Motivation
Users often start a conversation in a group (e.g., Discord server or Telegram group) and t... | open | null | false | 4 | [] | [] | 2026-03-21T18:27:24Z | 2026-05-04T19:30:56Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | amadeusneko | 35,049,527 | MDQ6VXNlcjM1MDQ5NTI3 | User | false |
openclaw/openclaw | 4,113,428,460 | I_kwDOQb6kR871Le_s | 51,825 | https://github.com/openclaw/openclaw/issues/51825 | https://api.github.com/repos/openclaw/openclaw/issues/51825 | TUI shows no activity indicator for system-injected events (bridge-notify, webhooks, cron) | ## Bug Report
### Description
When the agent receives a system-injected event (bridge notification, webhook, cron trigger) and starts working, the TUI shows no activity indicator. The user has no way to know work is happening unless they triggered it themselves with a direct message.
### Steps to Reproduce
1. Set u... | open | null | false | 2 | [] | [] | 2026-03-21T19:00:51Z | 2026-05-04T20:23:16Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | AlethiaQuizForge | 254,334,588 | U_kgDODyjWfA | User | false |
openclaw/openclaw | 4,113,455,905 | I_kwDOQb6kR871Llsh | 51,828 | https://github.com/openclaw/openclaw/issues/51828 | https://api.github.com/repos/openclaw/openclaw/issues/51828 | The Floor is Shaking — Gateway and Session Stability Under Real Workloads | You build on top of the gateway. You assume the floor is solid.
It isn't always.
---
**What the community is running into:**
**1. Cron jobs crashing with `gateway closed 1000`** — The cron scheduler thinks the gateway is open. The gateway has already closed its internal state. They disagree, and the cron job loses.... | closed | not_planned | true | 3 | [] | [] | 2026-03-21T19:07:17Z | 2026-05-03T04:34:01Z | 2026-05-01T04:15:03Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | ocdlmv1 | 266,897,137 | U_kgDOD-iG8Q | User | false |
openclaw/openclaw | 4,113,477,891 | I_kwDOQb6kR871LrED | 51,832 | https://github.com/openclaw/openclaw/issues/51832 | https://api.github.com/repos/openclaw/openclaw/issues/51832 | [Bug]: Multi-agent error cascade when gateway posts error messages to shared Slack channels | # [Bug]: Multi-agent error cascade when gateway posts error messages to shared Slack channels
## Summary
When multiple agents share a Slack channel with `allowBots: true`, gateway-posted error messages (e.g., OAuth failures) trigger a cascade loop. Each agent's error message triggers the other agent to attempt a resp... | open | null | false | 3 | [] | [] | 2026-03-21T19:11:59Z | 2026-05-04T19:11:25Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | scottyallen | 73,208 | MDQ6VXNlcjczMjA4 | User | false |
openclaw/openclaw | 4,113,784,348 | I_kwDOQb6kR871M14c | 51,865 | https://github.com/openclaw/openclaw/issues/51865 | https://api.github.com/repos/openclaw/openclaw/issues/51865 | Security audit: false positive on valid node denyCommands (camera.snap, screen.record, etc.) | ## Summary
`openclaw security audit` reports `gateway.nodes.denyCommands` entries as "unknown command names" even though they are valid, documented node commands.
## Steps to Reproduce
1. Configure `gateway.nodes.denyCommands` with valid node commands:
```json
"denyCommands": [
"camera.snap",
"camera.clip",
"s... | closed | completed | true | 2 | [] | [] | 2026-03-21T20:33:38Z | 2026-05-01T04:40:50Z | 2026-04-29T02:40:42Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | emdash-dot | 265,377,198 | U_kgDOD9FVrg | User | false |
openclaw/openclaw | 4,113,813,690 | I_kwDOQb6kR871M9C6 | 51,871 | https://github.com/openclaw/openclaw/issues/51871 | https://api.github.com/repos/openclaw/openclaw/issues/51871 | [Bug]: Control UI: Cron jobs not displayed in dashboard (2026.3.13) | ### Bug type
Regression (worked before, now fails)
### Summary
**Version:** OpenClaw 2026.3.13 (61d171a)
**OS:** Raspberry Pi 5, Linux (Debian), systemd service
**Install:** npm global
**Bug:**
Cron-Aufgaben werden im Dashboard (Control UI) nicht angezeigt —
die Seite bleibt leer.
**jobs.json ist valid:**
- Date... | open | null | false | 10 | [
"bug",
"regression"
] | [] | 2026-03-21T20:46:39Z | 2026-05-04T21:17:55Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mschmidt739-sys | 255,667,951 | U_kgDODz0u7w | User | false |
openclaw/openclaw | 4,113,818,219 | I_kwDOQb6kR871M-Jr | 51,872 | https://github.com/openclaw/openclaw/issues/51872 | https://api.github.com/repos/openclaw/openclaw/issues/51872 | [Feature]: bearer token authentication for proxy servers | ### Summary
Enable HTTP transport-level authentication for non-web clients (android, ios)
### Problem to solve
It is common to have a reverse proxy in front of web apps, including things like the openclaw gateway.
These in turn might have several means of authentication:
- basic auth
- bearer token
- fixed header k... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-21T20:48:47Z | 2026-05-04T19:11:31Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | donbowman | 5,131,923 | MDQ6VXNlcjUxMzE5MjM= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.