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 | 4,045,824,738 | I_kwDOQb6kR87xJmLi | 41,100 | https://github.com/openclaw/openclaw/issues/41100 | https://api.github.com/repos/openclaw/openclaw/issues/41100 | Compaction fails silently with 'Compaction cancelled' — safeguard returns no user-friendly reason | ## Bug Description
When compaction triggers on a fresh session (or after /reset), it fails with the message 'Compaction cancelled' and no actionable reason. The underlying cause is the compaction safeguard hook returning `{ cancel: true }` without a `reason` field.
## Root Cause
In `dist/compact-D3emcZgv.js`, the `com... | closed | completed | true | 1 | [] | [] | 2026-03-09T14:08:12Z | 2026-04-27T04:56:52Z | 2026-04-24T21:04:34Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | starbuck100 | 25,417,736 | MDQ6VXNlcjI1NDE3NzM2 | User | false |
openclaw/openclaw | 4,045,829,384 | I_kwDOQb6kR87xJnUI | 41,102 | https://github.com/openclaw/openclaw/issues/41102 | https://api.github.com/repos/openclaw/openclaw/issues/41102 | [Bug]: 审批请求不能正确返回 | ### Bug type
Regression (worked before, now fails)
### Summary
审批请求(waitDecision)已发送成功,但审批结果(resolve)从未到达 Gateway。
在 agent 请求 exec 相关权限,在 web ui 同意后,agent 并没有继续往下进行
背景:云服务器,本地电脑 ssh 。审批 exec 权限,允许一次,会话没有继续往下进行
### Steps to reproduce
1. 让会话执行:执行一下pwd命令看下
2. 会弹出审核请求
3. 点击 Allowed (once)
### Expected behavior
点击 ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T14:09:07Z | 2026-04-27T04:56:51Z | 2026-04-24T21:04:37Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | LiShiwen0206 | 26,299,226 | MDQ6VXNlcjI2Mjk5MjI2 | User | false |
openclaw/openclaw | 4,045,881,025 | I_kwDOQb6kR87xJz7B | 41,116 | https://github.com/openclaw/openclaw/issues/41116 | https://api.github.com/repos/openclaw/openclaw/issues/41116 | Discord DM thread-bound ACP spawn fails with 'Session binding adapter failed to bind target conversation' | ## Summary
Discord DM `thread:true` ACP spawns fail with `Session binding adapter failed to bind target conversation`, while plain ACP runs succeed.
## Environment
- OpenClaw version: 2026.3.7 / 2026.3.8 observed during testing
- Runtime: Linux, Node 22.22.0
- Channel: Discord DM (`chat_type: direct`)
- ACP backend: `... | closed | completed | true | 3 | [] | [] | 2026-03-09T14:18:40Z | 2026-04-27T04:56:50Z | 2026-04-24T21:04:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bbbl-xiangzi | 195,702,006 | U_kgDOC6os9g | User | false |
openclaw/openclaw | 4,045,941,198 | I_kwDOQb6kR87xKCnO | 41,128 | https://github.com/openclaw/openclaw/issues/41128 | https://api.github.com/repos/openclaw/openclaw/issues/41128 | [Bug]: Isolated cron + announce delivery regression in 2026.3.8 - sessions not created, no messages delivered | # [Bug]: Isolated cron + announce delivery regression in 2026.3.8 - sessions not created, no messages delivered
## Summary
OpenClaw 2026.3.8 introduced a regression in isolated cron jobs with announce delivery. Cron jobs report `delivered: true` but no session is created and no messages are actually sent to Telegram.... | closed | completed | true | 3 | [] | [] | 2026-03-09T14:29:15Z | 2026-04-27T04:56:49Z | 2026-04-24T21:04:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Louise-Qiuqiu | 253,923,225 | U_kgDODyKPmQ | User | false |
openclaw/openclaw | 4,045,953,135 | I_kwDOQb6kR87xKFhv | 41,129 | https://github.com/openclaw/openclaw/issues/41129 | https://api.github.com/repos/openclaw/openclaw/issues/41129 | [Bug]: Cron announce delivery reports success but messages never arrive (Telegram) | ### Bug type
Regression (worked before, now fails)
### Summary
Isolated cron jobs with delivery.mode: announce report delivered: true and deliveryStatus: delivered in run logs, but messages never actually arrive in Telegram. Direct messages via the message tool work fine.
### Steps to reproduce
Create cron job:
op... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-09T14:31:39Z | 2026-04-27T04:56:49Z | 2026-04-24T21:04:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wirelessjoe | 36,973 | MDQ6VXNlcjM2OTcz | User | false |
openclaw/openclaw | 4,045,965,731 | I_kwDOQb6kR87xKImj | 41,130 | https://github.com/openclaw/openclaw/issues/41130 | https://api.github.com/repos/openclaw/openclaw/issues/41130 | Feature Request: Native Streaming Response Support for webchat Channel (SSE/WebSocket) | ### Summary
# Feature Request: Native Streaming Response Support for webchat Channel (SSE/WebSocket)
## 📌 Summary
**Current Issue**: OpenClaw's `webchat` channel does not support **native streaming responses**. Users must wait for the entire AI response to be generated before seeing any output, resulting in poor us... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-03-09T14:34:09Z | 2026-04-27T04:56:48Z | 2026-04-24T21:04:51Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | qingyunpor | 220,163,817 | U_kgDODR9u6Q | User | false |
openclaw/openclaw | 4,045,970,749 | I_kwDOQb6kR87xKJ09 | 41,132 | https://github.com/openclaw/openclaw/issues/41132 | https://api.github.com/repos/openclaw/openclaw/issues/41132 | Model Capability Declaration and Message Compatibility Check | ## Problem Description
When a model receives an unsupported message type (e.g., images), it throws an error `BadRequestError: Image content is not supported by this model`. This error message remains in the context, causing all subsequent messages to fail with the same error. The only solution is to use `/reset` to cl... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T14:34:55Z | 2026-04-27T04:56:47Z | 2026-04-24T21:04:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wrbing728 | 8,429,360 | MDQ6VXNlcjg0MjkzNjA= | User | false |
openclaw/openclaw | 4,045,989,458 | I_kwDOQb6kR87xKOZS | 41,137 | https://github.com/openclaw/openclaw/issues/41137 | https://api.github.com/repos/openclaw/openclaw/issues/41137 | [Bug]: generating QR code fails in web UI | ### Bug type
Regression (worked before, now fails)
### Summary
version
2026.3.7
Trying to connect via web ui
http://127.0.0.1:18789/chat?session=agent%3Amain%3Amain
it says it has generated a QR code
but
no QR code or very badly formatted
<img width="3451" height="1346" alt="Image" src="https://github.com/user-at... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T14:38:18Z | 2026-04-27T04:56:46Z | 2026-04-24T21:04:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pleabargain | 640,846 | MDQ6VXNlcjY0MDg0Ng== | User | false |
openclaw/openclaw | 4,046,044,993 | I_kwDOQb6kR87xKb9B | 41,148 | https://github.com/openclaw/openclaw/issues/41148 | https://api.github.com/repos/openclaw/openclaw/issues/41148 | Kimi 搜索集成的时间感知错误 | ## 问题描述
Kimi 搜索集成的时间感知错误
## 问题现象
1. 系统时间:2026年3月9日
2. 询问 Kimi "今天伊朗局势"
3. Kimi 回复说"2026年是未来日期,知识截止于2024年4月"
4. 即使明确告知"今天是2026年3月9日",模型仍然坚持错误认知
## 根本原因
Kimi API 的系统提示词中注入了错误的日期(2024年12月),导致模型时间认知错乱
## 期望行为
模型应该能正确识别当前日期,并使用搜索工具获取实时信息
## 测试环境
- OpenClaw 版本:2026.3.8
- 模型:kimi-k2.5
- 搜索配置:provider: "kimi"
## 附加信... | closed | completed | true | 1 | [] | [] | 2026-03-09T14:47:44Z | 2026-04-27T04:56:46Z | 2026-04-24T21:05:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | coobee0620 | 20,234,643 | MDQ6VXNlcjIwMjM0NjQz | User | false |
openclaw/openclaw | 4,046,112,174 | I_kwDOQb6kR87xKsWu | 41,162 | https://github.com/openclaw/openclaw/issues/41162 | https://api.github.com/repos/openclaw/openclaw/issues/41162 | Bug: isolated session with announce delivery hijacks primary session route | # Bug: isolated session with announce delivery hijacks primary session route
## Summary
When a subagent/cron task runs with `sessionTarget=isolated` and `delivery.mode=announce`, and the user replies to the announcement message, OpenClCore incorrectly rebinds the primary session route to the isolated session.
This c... | closed | completed | true | 1 | [] | [] | 2026-03-09T14:59:31Z | 2026-04-27T04:56:45Z | 2026-04-24T21:05:05Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pen364692088 | 19,620,358 | MDQ6VXNlcjE5NjIwMzU4 | User | false |
openclaw/openclaw | 4,046,208,530 | I_kwDOQb6kR87xLD4S | 41,174 | https://github.com/openclaw/openclaw/issues/41174 | https://api.github.com/repos/openclaw/openclaw/issues/41174 | [Feature]: when API limit is reached suggest a different API | ### Summary
<img width="3840" height="545" alt="Image" src="https://github.com/user-attachments/assets/aa628795-dbea-4104-95dd-40fd6906a7ed" />
I have many API options but work doesn't stop just because I have reached API limit on one API.
suggestion: inform user that API limit has been reached
THEN
suggest other... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T15:18:07Z | 2026-04-27T04:56:44Z | 2026-04-24T21:05:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pleabargain | 640,846 | MDQ6VXNlcjY0MDg0Ng== | User | false |
openclaw/openclaw | 4,046,223,145 | I_kwDOQb6kR87xLHcp | 41,175 | https://github.com/openclaw/openclaw/issues/41175 | https://api.github.com/repos/openclaw/openclaw/issues/41175 | Discord: large image attachments cause inbound worker timeout or silent drop | ## Summary
When a Discord message includes large image attachments (15-17MB PNGs), the inbound worker either:
1. Times out after 1800 seconds without delivering the message to the agent
2. Silently drops the message entirely (no log entry for the inbound message)
## Environment
- OpenClaw version: latest (npm, instal... | closed | completed | true | 4 | [] | [] | 2026-03-09T15:20:54Z | 2026-04-27T04:56:44Z | 2026-04-24T21:05:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | guitarfreak26 | 149,803,471 | U_kgDOCO3Rzw | User | false |
openclaw/openclaw | 4,046,256,884 | I_kwDOQb6kR87xLPr0 | 41,181 | https://github.com/openclaw/openclaw/issues/41181 | https://api.github.com/repos/openclaw/openclaw/issues/41181 | [Bug] Himalaya Skill returns raw function_call/exec commands instead of natural language response | ### Bug type
Regression (worked before, now fails)
### Summary
When interacting with the `himalaya` skill via the OpenClaw chat interface (e.g., configuring email accounts, checking version, deleting the skill), the system fails to return human-readable natural language results (e.g., "Configuration successful", "F... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T15:26:45Z | 2026-04-27T04:56:43Z | 2026-04-24T21:05:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ylgkger | 16,173,716 | MDQ6VXNlcjE2MTczNzE2 | User | false |
openclaw/openclaw | 4,046,282,951 | I_kwDOQb6kR87xLWDH | 41,186 | https://github.com/openclaw/openclaw/issues/41186 | https://api.github.com/repos/openclaw/openclaw/issues/41186 | Feature Request: Add message deduplication configuration for webchat channel | ## Problem
The webchat channel in OpenClaw exhibits duplicate message output due to the `delivery-mirror` mechanism. Each message appears twice in the chat history:
1. Original message from the AI model
2. Duplicate message from `delivery-mirror` model
This creates a poor user experience in the webchat control UI.
#... | closed | completed | true | 1 | [] | [] | 2026-03-09T15:31:21Z | 2026-04-27T04:56:42Z | 2026-04-24T21:05:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | EsonGitHub | 12,593,267 | MDQ6VXNlcjEyNTkzMjY3 | User | false |
openclaw/openclaw | 4,046,289,755 | I_kwDOQb6kR87xLXtb | 41,188 | https://github.com/openclaw/openclaw/issues/41188 | https://api.github.com/repos/openclaw/openclaw/issues/41188 | [Feature]: Option to hide tool outputs from chat (silent mode) | ### Summary
Feature Request: Option to hide tool outputs from chat (silent mode)
<img width="1080" height="2460" alt="Image" src="https://github.com/user-attachments/assets/332986e4-75cf-4557-840a-a092ca952e38" />
Description:
When AI runs commands, raw outputs show in chat — makes it messy. Want something like nebu... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T15:32:39Z | 2026-04-27T04:56:41Z | 2026-04-24T21:05:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | shail699 | 215,453,664 | U_kgDODNeP4A | User | false |
openclaw/openclaw | 4,046,356,051 | I_kwDOQb6kR87xLn5T | 41,202 | https://github.com/openclaw/openclaw/issues/41202 | https://api.github.com/repos/openclaw/openclaw/issues/41202 | [Optimization] 优化重复调用: exec(['command']) | ### Describe the problem
在日常使用中,Agent 极高频地通过原始 `exec` 调用来进行基础的文件搜索和修改操作(如 `grep`, `cat`, `sed`)。这不仅消耗大量 Token 维持复杂的终端会话状态,还容易出现 Shell 语法错误(如特殊字符转义失败)。
### Proposal
提供更高阶、自带防错机制的文件处理工具(如专用的 Search & Replace 语义化指令),或者在 `exec` 层面封装常用组合动作,减少模型直接拼写 Shell 命令的频率。这能显著提升执行速度、降低成本并增加稳定性。 | closed | completed | true | 1 | [] | [] | 2026-03-09T15:45:18Z | 2026-04-27T04:56:41Z | 2026-04-24T21:05:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tianxiao1430-jpg | 255,869,001 | U_kgDOD0BASQ | User | false |
openclaw/openclaw | 4,046,356,645 | I_kwDOQb6kR87xLoCl | 41,203 | https://github.com/openclaw/openclaw/issues/41203 | https://api.github.com/repos/openclaw/openclaw/issues/41203 | [Optimization] 优化重复调用: browser(['action', 'profile']) | ### Describe the problem
目前 `browser` 工具在处理复杂的表单填写、多页面跳转提取数据时,模型必须进行大量 `action` 步骤(navigate -> locate element -> click -> wait -> extract)。这些极高频的多步重复调用带来了极大的延迟、高额 Token 开销,并很容易因中途的元素加载延迟而中断整个会话循环。
### Proposal
引入更高级的 `browser` 工作流机制或宏(Macro)能力:
1. **聚合型动作指令**:例如 `fill_and_submit_form` 或 `search_and_extract_results`,将多... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T15:45:26Z | 2026-04-27T04:56:40Z | 2026-04-24T21:05:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tianxiao1430-jpg | 255,869,001 | U_kgDOD0BASQ | User | false |
openclaw/openclaw | 4,046,362,588 | I_kwDOQb6kR87xLpfc | 41,205 | https://github.com/openclaw/openclaw/issues/41205 | https://api.github.com/repos/openclaw/openclaw/issues/41205 | [Bug]: Broken favicon URL in Control UI (old Railway template URL) | ### Bug type
Regression (worked before, now fails)
### Summary
The Control UI is trying to load favicon from an old Railway template URL that no longer works.
### Steps to reproduce
**Description:**
The Control UI is trying to load favicon from an old Railway template URL that no longer works.
**URL:** https://[R... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T15:46:35Z | 2026-04-27T04:56:39Z | 2026-04-24T21:05:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jsingal | 33,892,090 | MDQ6VXNlcjMzODkyMDkw | User | false |
openclaw/openclaw | 4,046,402,800 | I_kwDOQb6kR87xLzTw | 41,211 | https://github.com/openclaw/openclaw/issues/41211 | https://api.github.com/repos/openclaw/openclaw/issues/41211 | AGENTS.md startup sequence should note that workspace files are already in Project Context | ### Summary
Add a note in the generated AGENTS.md startup sequence clarifying that SOUL.md, USER.md, and AGENTS.md are already injected into the system prompt by OpenClaw — agents should only explicitly read files not in Project Context.
### Problem to solve
The default AGENTS.md generated by OpenClaw contains this ... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-09T15:54:22Z | 2026-04-27T04:56:38Z | 2026-04-24T21:05:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | huangyoje | 25,878,791 | MDQ6VXNlcjI1ODc4Nzkx | User | false |
openclaw/openclaw | 4,046,407,887 | I_kwDOQb6kR87xL0jP | 41,212 | https://github.com/openclaw/openclaw/issues/41212 | https://api.github.com/repos/openclaw/openclaw/issues/41212 | [Feature]: The Shadow Clone | ### Summary
Instead of Openclaw just being a tool you "use," it becomes a version of you that never sleeps.
The "Shadow Clone" feature would watch your workflow for a few days—how you edit, the specific way you structure your code, and even your "unprofessional and free" way of talking to people. Once it’s synced, yo... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-09T15:55:13Z | 2026-04-27T04:56:37Z | 2026-04-24T21:05:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | aryamthecodebreaker | 130,867,585 | U_kgDOB8zhgQ | User | false |
openclaw/openclaw | 4,046,414,843 | I_kwDOQb6kR87xL2P7 | 41,214 | https://github.com/openclaw/openclaw/issues/41214 | https://api.github.com/repos/openclaw/openclaw/issues/41214 | [Bug]: 2026.3.8 remote mode: browser node routing can prevent local Chrome relay (127.0.0.1:18792) from starting | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading from 2026.3.2 to 2026.3.8 in remote/tailnet mode (Mac has both npm CLI node-host and macOS app node), Chrome extension relay can fail with: "Relay not reachable/authenticated at http://127.0.0.1:18792/". Pinning browser routing to CLI nod... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-09T15:56:21Z | 2026-04-27T04:56:37Z | 2026-04-24T21:05:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kongwy | 9,990,631 | MDQ6VXNlcjk5OTA2MzE= | User | false |
openclaw/openclaw | 4,046,428,148 | I_kwDOQb6kR87xL5f0 | 41,215 | https://github.com/openclaw/openclaw/issues/41215 | https://api.github.com/repos/openclaw/openclaw/issues/41215 | [Bug]: Outbound reply parsing flattens indentation in final payload text | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
OpenClaw can flatten indentation in final assistant payload text during outbound reply parsing.
This is not limited to JSON. It affects any indentation-sensitive multiline output, including:
- plain indented text
- fenced code blocks
- struc... | closed | completed | true | 1 | [] | [] | 2026-03-09T15:58:19Z | 2026-04-27T04:56:35Z | 2026-04-24T21:05:51Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | seilk | 88,271,769 | MDQ6VXNlcjg4MjcxNzY5 | User | false |
openclaw/openclaw | 4,046,434,345 | I_kwDOQb6kR87xL7Ap | 41,216 | https://github.com/openclaw/openclaw/issues/41216 | https://api.github.com/repos/openclaw/openclaw/issues/41216 | Feature request: memory flush before /new and /restart session reset | ## Problem
`/new` and `/restart` immediately destroy the session without any opportunity to persist context. The pre-compaction memory flush (`compaction.memoryFlush`) works great for auto-compaction, but session resets bypass it entirely.
This means any work, decisions, or context accumulated since the last memory w... | closed | completed | true | 3 | [] | [] | 2026-03-09T15:59:15Z | 2026-04-27T04:29:30Z | 2026-04-24T21:05:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | SubAllStar | 152,678,168 | U_kgDOCRmvGA | User | false |
openclaw/openclaw | 4,046,532,902 | I_kwDOQb6kR87xMTEm | 41,234 | https://github.com/openclaw/openclaw/issues/41234 | https://api.github.com/repos/openclaw/openclaw/issues/41234 | [Bug]: '--disable-blink-features=AutomationControlled' flag causes Browser Control Service Unavailability for Agent Web Scraping | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When attempting to perform web scraping tasks using the OpenClaw browser tool (coordinated by an Agent), the browser control service repeatedly becomes unavailable, leading to scraping failures. A persistent warning message appears during b... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T16:16:21Z | 2026-04-27T04:56:34Z | 2026-04-24T21:05:59Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | csumichael | 266,318,054 | U_kgDOD9-w5g | User | false |
openclaw/openclaw | 4,046,542,191 | I_kwDOQb6kR87xMVVv | 41,237 | https://github.com/openclaw/openclaw/issues/41237 | https://api.github.com/repos/openclaw/openclaw/issues/41237 | [Feature]: ElevenLabs support for understanding audio and voice notes | ### Summary
Add ElevenLabs as a provider for media understanding.
### Problem to solve
It looks as of today, openclaw doesn't support ElevenLabs for voice notes yet.
- the [doc page](https://docs.openclaw.ai/nodes/audio) only lists `OpenAI → Groq → Deepgram → Google`
- the [provider directory](https://github.com/op... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T16:18:12Z | 2026-04-27T04:56:33Z | 2026-04-24T21:06:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wlnirvana | 3,822,785 | MDQ6VXNlcjM4MjI3ODU= | User | false |
openclaw/openclaw | 4,325,595,492 | I_kwDOQb6kR88AAAABAdNZZA | 71,262 | https://github.com/openclaw/openclaw/issues/71262 | https://api.github.com/repos/openclaw/openclaw/issues/71262 | [Feature]: voice-call realtime mode should expose gateway agent tools (bridge tool calls back to agent runtime) | ## Summary
The `voice-call` plugin's `realtime` mode (OpenAI / Gemini Live end-to-end audio) cannot call into the gateway's agent tools, skills, or memory. The realtime voice provider runs as a self-contained session that talks directly to the caller — the plugin has the scaffolding to pass `tools` to the provider (`r... | closed | completed | true | 1 | [] | [] | 2026-04-24T21:07:10Z | 2026-04-27T04:41:53Z | 2026-04-25T01:51:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | danrasmuson | 4,013,767 | MDQ6VXNlcjQwMTM3Njc= | User | false |
openclaw/openclaw | 4,046,626,929 | I_kwDOQb6kR87xMqBx | 41,258 | https://github.com/openclaw/openclaw/issues/41258 | https://api.github.com/repos/openclaw/openclaw/issues/41258 | openclaw backup create (state/full) silently exits 0 with empty 10-byte gzip archive | ## Description
`openclaw backup create` exits with code 0 and produces no output, leaving behind a 10-byte `.tmp` file containing only a bare gzip header. No `.tar.gz` archive is ever finalized.
`--only-config` works correctly and produces a valid archive.
## Version
```
openclaw 2026.3.8 (3caab92)
```
macOS Darwi... | closed | completed | true | 3 | [] | [] | 2026-03-09T16:34:46Z | 2026-04-27T04:56:33Z | 2026-04-24T21:09:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tjropenclaw | 265,708,139 | U_kgDOD9Ziaw | User | false |
openclaw/openclaw | 4,046,736,650 | I_kwDOQb6kR87xNE0K | 41,280 | https://github.com/openclaw/openclaw/issues/41280 | https://api.github.com/repos/openclaw/openclaw/issues/41280 | Rusty (computer-use) services not running on srv1325349 | ## Issue
Rusty computer-use services are not running on srv1325349. All three ports (8420, 8421, 8422) are unreachable.
## Discovery
- **Date:** 2026-03-09 16:54 UTC
- **Method:** Overnight infrastructure audit (cron)
- **Agent:** sysadmin
## Evidence
- `curl localhost:8420/health` → Connection refused
- `curl localh... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T16:55:06Z | 2026-04-27T04:56:32Z | 2026-04-24T21:09:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,046,917,342 | I_kwDOQb6kR87xNw7e | 41,301 | https://github.com/openclaw/openclaw/issues/41301 | https://api.github.com/repos/openclaw/openclaw/issues/41301 | Feature request: Include model name in agent/embedded error logs | ## Feature Request
When an API error occurs (e.g., rate limit, overload), the error log currently only shows:
**Problem:** It's difficult to quickly identify which model/provider failed without grepping backwards through logs for the runId.
**Requested Enhancement:** Include the model name and provider in the erro... | closed | completed | true | 1 | [] | [] | 2026-03-09T17:28:42Z | 2026-04-27T04:56:31Z | 2026-04-24T21:09:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | luiachan-svg | 257,427,697 | U_kgDOD1gI8Q | User | false |
openclaw/openclaw | 4,046,963,760 | I_kwDOQb6kR87xN8Qw | 41,310 | https://github.com/openclaw/openclaw/issues/41310 | https://api.github.com/repos/openclaw/openclaw/issues/41310 | Messages duplicated when streaming mode is enabled (Telegram & Feishu) | When streaming mode is enabled, messages are being split or duplicated when sent to Telegram or Feishu channels.
Environment:
• OpenClaw 2026.3.8
• macOS
• Node.js v22.14.0
Steps to Reproduce:
1. Set "streaming": true or "streaming": "partial" in channels.telegram or channels.feishu
2. Send a message (especially lo... | closed | completed | true | 1 | [] | [] | 2026-03-09T17:38:18Z | 2026-04-27T04:56:31Z | 2026-04-24T21:09:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | xiangyingchang | 10,992,221 | MDQ6VXNlcjEwOTkyMjIx | User | false |
openclaw/openclaw | 4,047,010,857 | I_kwDOQb6kR87xOHwp | 41,315 | https://github.com/openclaw/openclaw/issues/41315 | https://api.github.com/repos/openclaw/openclaw/issues/41315 | [Bug]: macOS LaunchAgent restart instability: gateway restart often leaves service not loaded; install+restart required | ## Summary
On macOS LaunchAgent setups, `openclaw gateway restart` frequently fails to bring the service back and the gateway ends in a `service not loaded` state. Manual recovery with `openclaw gateway install` then `openclaw gateway restart` is repeatedly required.
## Environment
- OpenClaw: 2026.3.7
- OS: macOS 26.... | closed | completed | true | 11 | [] | [] | 2026-03-09T17:48:03Z | 2026-04-27T04:56:30Z | 2026-04-24T21:09:46Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | blexibot | 263,048,527 | U_kgDOD63NTw | User | false |
openclaw/openclaw | 4,047,073,679 | I_kwDOQb6kR87xOXGP | 41,326 | https://github.com/openclaw/openclaw/issues/41326 | https://api.github.com/repos/openclaw/openclaw/issues/41326 | Slack DM sessions do not refresh available_skills after workspace custom skill discovery updates | ## Summary
On macOS with OpenClaw 2026.3.8, workspace custom skills are discovered correctly, but an existing Slack DM session does not see newly added skills in its prompt-visible `available_skills`.
In short:
- skill discovery is working
- Dashboard / `openclaw skills list` reflect the new skill
- but the active S... | closed | completed | true | 1 | [] | [] | 2026-03-09T18:00:27Z | 2026-04-27T04:56:29Z | 2026-04-24T21:09:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kazuyuki-eguchi | 201,872,039 | U_kgDODAhSpw | User | false |
openclaw/openclaw | 4,047,163,102 | I_kwDOQb6kR87xOs7e | 41,348 | https://github.com/openclaw/openclaw/issues/41348 | https://api.github.com/repos/openclaw/openclaw/issues/41348 | iMessage delivery: garbled prefix bytes in message content | ## Description
Messages delivered via iMessage announce (cron job delivery) have garbled UTF-8 replacement characters (`U+FFFD`) followed by length-like bytes prepended to the message content.
## Example
```
hex prefix: efbfbd18022a2ae29c85204b337320436c757374
text: '�\x18\x02**✅ K3s Cluster: 6 noder Re'
```
The firs... | closed | completed | true | 2 | [] | [] | 2026-03-09T18:19:39Z | 2026-04-27T04:56:28Z | 2026-04-24T21:09:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Hovhas | 25,573,180 | MDQ6VXNlcjI1NTczMTgw | User | false |
openclaw/openclaw | 4,047,186,291 | I_kwDOQb6kR87xOylz | 41,352 | https://github.com/openclaw/openclaw/issues/41352 | https://api.github.com/repos/openclaw/openclaw/issues/41352 | Discord session stops receiving messages after Compaction | ## Bug Description
After a Compaction event is triggered in a Discord session, the session stops receiving new messages. The only way to recover is to restart the gateway.
## Root Cause Analysis
In \`compact-B247y5Qt.js\`, the \`isCompacting()\` function checks:
\`\`\`javascript
isCompacting: () => state.compaction... | closed | completed | true | 3 | [] | [] | 2026-03-09T18:24:46Z | 2026-04-27T04:56:28Z | 2026-04-24T21:10:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | meareabc | 74,286,633 | MDQ6VXNlcjc0Mjg2NjMz | User | false |
openclaw/openclaw | 4,047,223,501 | I_kwDOQb6kR87xO7rN | 41,356 | https://github.com/openclaw/openclaw/issues/41356 | https://api.github.com/repos/openclaw/openclaw/issues/41356 | MemoryFlush not triggering - tokens exceed threshold but flush never fires | ## Bug Description
Memory flush is configured but never triggers even when tokens far exceed threshold.
## Config
```json
"memoryFlush": {
"enabled": true,
"softThresholdTokens": 4000
}
```
## Reproduction
1. Config loaded at 11:11 UTC (confirmed in logs)
2. Session has ~355,000 tokens (89x over threshold)
3. Flu... | closed | completed | true | 3 | [] | [] | 2026-03-09T18:33:22Z | 2026-04-27T04:56:27Z | 2026-04-24T21:10:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vinaykotha433-del | 262,733,006 | U_kgDOD6j8zg | User | false |
openclaw/openclaw | 4,047,263,544 | I_kwDOQb6kR87xPFc4 | 41,358 | https://github.com/openclaw/openclaw/issues/41358 | https://api.github.com/repos/openclaw/openclaw/issues/41358 | [Feature Request] WebChat建议增加隐藏工具执行过程的配置选项 | ### Summary
希望WebChat或Gateway配置中增加一个选项,用于隐藏工具执行过程的详细输出(如 [non-text content: thinking, toolCall]),只显示最终的回复内容,提升历史对话回溯效率。
### Problem to solve
使用WebChat时,对话历史中会显示很多工具执行过程,如:
- [non-text content: thinking, toolCall]
- Tool result (exec): ...
- Tool result (read): ...
这些过程内容严重影响回溯历史对话的效率,用户需要从大量技术输出中筛选有效信息。
### Propos... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-09T18:42:01Z | 2026-04-27T04:56:26Z | 2026-04-24T21:10:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jisicheng | 36,007,820 | MDQ6VXNlcjM2MDA3ODIw | User | false |
openclaw/openclaw | 4,047,315,000 | I_kwDOQb6kR87xPSA4 | 41,363 | https://github.com/openclaw/openclaw/issues/41363 | https://api.github.com/repos/openclaw/openclaw/issues/41363 | Native Voice Input for Control UI (ChatGPT-style) | # Title (copy this):
Native Voice Input for Control UI (ChatGPT-style)
# Body (copy everything below):
## Summary
Add real-time voice input capability to the OpenClaw Control UI, similar to ChatGPT's Advanced Voice Mode, using Web Audio API and local/transcription services.
## Motivation
Voice input is becoming a ... | closed | completed | true | 4 | [] | [] | 2026-03-09T18:53:53Z | 2026-04-27T04:56:25Z | 2026-04-24T21:10:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | goomsdd21-coder | 264,220,160 | U_kgDOD7-uAA | User | false |
openclaw/openclaw | 4,047,361,831 | I_kwDOQb6kR87xPdcn | 41,371 | https://github.com/openclaw/openclaw/issues/41371 | https://api.github.com/repos/openclaw/openclaw/issues/41371 | [Feature]: Add connection Timeout | ### Summary
Add a configurable provider connection / first-response timeout for LLM requests.
Currently OpenClaw aborts provider requests if the model does not produce an HTTP response or first streaming token quickly enough. This is problematic for local or large models that require longer initialization or prefill ... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-03-09T19:04:42Z | 2026-04-27T04:56:25Z | 2026-04-24T21:10:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | filipgrzalja | 254,735,035 | U_kgDODy7yuw | User | false |
openclaw/openclaw | 4,047,422,696 | I_kwDOQb6kR87xPsTo | 41,379 | https://github.com/openclaw/openclaw/issues/41379 | https://api.github.com/repos/openclaw/openclaw/issues/41379 | Bug: 421 Error and mismatch error in container logs | The user reported a `421 error` and a `mismatch error` in the container logs. Due to the limited access in the container environment (Azure Container Apps), the exact stack trace could not be extracted automatically. Please investigate occurrences of 421 errors and mismatch errors in the gateway. | closed | completed | true | 6 | [] | [] | 2026-03-09T19:18:44Z | 2026-04-27T04:56:24Z | 2026-04-24T21:10:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | serdarsemen | 3,353,767 | MDQ6VXNlcjMzNTM3Njc= | User | false |
openclaw/openclaw | 4,047,598,658 | I_kwDOQb6kR87xQXRC | 41,395 | https://github.com/openclaw/openclaw/issues/41395 | https://api.github.com/repos/openclaw/openclaw/issues/41395 | Hostinger gateway 1090 commits behind main - urgent update needed | ## Overnight Audit Finding (2026-03-09)
**Server:** srv1325349 (Hostinger VPS)
**Current Version:** 2026.3.3
**Available Version:** 2026.3.8
**Git Status:** ahead 2, behind 1090 commits
### Issue
The production gateway is significantly behind the main branch (1090 commits). This gap may include critical bug fixes, se... | closed | not_planned | true | 2 | [] | [] | 2026-03-09T19:55:46Z | 2026-04-27T04:56:23Z | 2026-04-24T21:10:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,047,652,649 | I_kwDOQb6kR87xQkcp | 41,397 | https://github.com/openclaw/openclaw/issues/41397 | https://api.github.com/repos/openclaw/openclaw/issues/41397 | voice-call: initial message uses Twilio <Say> fallback instead of ElevenLabs when media stream not yet active | ## Description
In conversation mode, `speakInitialMessage` is called when the media stream WebSocket connects. However, the ElevenLabs telephony TTS path requires the media stream to be fully active before it can stream audio. If the initial message fires before the stream is confirmed active, it falls back to Twilio'... | closed | completed | true | 2 | [] | [] | 2026-03-09T20:07:33Z | 2026-04-27T04:56:22Z | 2026-04-24T21:10:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tjropenclaw | 265,708,139 | U_kgDOD9Ziaw | User | false |
openclaw/openclaw | 4,047,850,736 | I_kwDOQb6kR87xRUzw | 41,432 | https://github.com/openclaw/openclaw/issues/41432 | https://api.github.com/repos/openclaw/openclaw/issues/41432 | Manual /compact can fail with "no real conversation messages to summarize" when real messages are only in turnPrefixMessages | ## Summary
Manual `/compact` can fail in the compaction safeguard even when the session contains real conversation messages.
The safeguard currently only checks:
```ts
preparation.messagesToSummarize.some(isRealConversationMessage)
```
But in at least one real/manual compaction path, the relevant user/assistant/too... | closed | completed | true | 1 | [] | [] | 2026-03-09T20:51:56Z | 2026-04-27T04:56:21Z | 2026-04-24T21:10:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | thomas-kuehn | 9,082,176 | MDQ6VXNlcjkwODIxNzY= | User | false |
openclaw/openclaw | 4,047,905,917 | I_kwDOQb6kR87xRiR9 | 41,438 | https://github.com/openclaw/openclaw/issues/41438 | https://api.github.com/repos/openclaw/openclaw/issues/41438 | [Bug]: openai-completions provider: assistant content serialized as array instead of string, breaking multi-turn tool calls | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using a custom provider with api: openai-completions, the assistant message content field is serialized as a JSON array (Anthropic internal format) instead of a string, causing all multi-turn tool call sessions to loop indefinitely.
#... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T21:04:46Z | 2026-04-27T04:56:20Z | 2026-04-24T21:10:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pnck | 6,532,324 | MDQ6VXNlcjY1MzIzMjQ= | User | false |
openclaw/openclaw | 4,047,991,471 | I_kwDOQb6kR87xR3Kv | 41,447 | https://github.com/openclaw/openclaw/issues/41447 | https://api.github.com/repos/openclaw/openclaw/issues/41447 | [Feature]: Compatibility with Agent Skills with plugin.json root | ### Summary
Recently noticed that the dotnet team released some skill files for Claude code here: https://github.com/dotnet/skills?tab=readme-ov-file
They're in the plugins folder.
Was wondering if there is either:
1. A way to use them on the current version of OpenClaw is a very plug-and-play manor
2. Plans to allo... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T21:25:35Z | 2026-04-27T04:56:19Z | 2026-04-24T21:10:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Eric1243123 | 142,035,508 | U_kgDOCHdKNA | User | false |
openclaw/openclaw | 4,048,041,983 | I_kwDOQb6kR87xSDf_ | 41,458 | https://github.com/openclaw/openclaw/issues/41458 | https://api.github.com/repos/openclaw/openclaw/issues/41458 | [Bug]: TTS voice message fails on Feishu channel - audio file not sent correctly | ### Bug type
Regression (worked before, now fails)
### Summary
TTS generated audio files are not sent correctly to Feishu. Short messages work occasionally, but longer messages result in 0-byte files and API errors.
### Steps to reproduce
1. Configure Feishu channel with OpenClaw
2. Use tts tool to generate voice:... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T21:38:16Z | 2026-04-27T04:56:18Z | 2026-04-24T21:10:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Joselolun | 257,445,322 | U_kgDOD1hNyg | User | false |
openclaw/openclaw | 4,048,097,067 | I_kwDOQb6kR87xSQ8r | 41,463 | https://github.com/openclaw/openclaw/issues/41463 | https://api.github.com/repos/openclaw/openclaw/issues/41463 | [Bug]: Custom subagent not listed by /agents and rejected by sessions_spawn (allowed: none) on 2026.3.8 and 2026.2.26 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Custom subagents defined in agents.list are not exposed by /agents and cannot be spawned via sessions_spawn, even with allowAgents: ["*"]. Reproduced on latest release (2026.3.8), then re-tested on 2026.2.26 with a clean minimal config and ... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T21:52:39Z | 2026-04-27T04:56:18Z | 2026-04-24T21:10:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | nouchy | 33,549,554 | MDQ6VXNlcjMzNTQ5NTU0 | User | false |
openclaw/openclaw | 4,048,271,737 | I_kwDOQb6kR87xS7l5 | 41,486 | https://github.com/openclaw/openclaw/issues/41486 | https://api.github.com/repos/openclaw/openclaw/issues/41486 | Ollama cron jobs: lightContext mode still sends large system prompt, causing timeouts on small local models | ## Problem
When using Ollama models (e.g., `qwen3:8b`, `mistral:7b`) for cron jobs with `lightContext: true`, the effective context sent to the model is still ~15-20K tokens. This causes timeouts on models with 32K context windows, especially when thinking tokens are involved.
## Root Cause Analysis
`lightContext: t... | closed | completed | true | 1 | [] | [] | 2026-03-09T22:34:39Z | 2026-04-27T04:56:17Z | 2026-04-24T21:10:55Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 27103highlandcrest-dev | 261,647,582 | U_kgDOD5hs3g | User | false |
openclaw/openclaw | 4,048,286,869 | I_kwDOQb6kR87xS_SV | 41,491 | https://github.com/openclaw/openclaw/issues/41491 | https://api.github.com/repos/openclaw/openclaw/issues/41491 | [Bug]: WhatsApp audio pipeline does not invoke STT transcription (whisper-large-v3-turbo / groq provider) | ### Bug type
Regression (worked before, now fails)
### Summary
WhatsApp audio messages received but STT transcription pipeline never invoked — audio silently dropped
[openclaw-bug-7573.md](https://github.com/user-attachments/files/25854797/openclaw-bug-7573.md)
### Steps to reproduce
1. Configure tools.media.audi... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-09T22:38:37Z | 2026-04-27T04:56:16Z | 2026-04-24T21:10:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dougdooley | 29,562,640 | MDQ6VXNlcjI5NTYyNjQw | User | false |
openclaw/openclaw | 4,048,289,021 | I_kwDOQb6kR87xS_z9 | 41,492 | https://github.com/openclaw/openclaw/issues/41492 | https://api.github.com/repos/openclaw/openclaw/issues/41492 | Regression: Google 429 still kills fallback chain on v2026.3.8 (originally #13623) | ## Bug
When multiple models from the same provider (e.g. 3 Google models) are in the failover chain and all return rate-limit errors, the chain does NOT fall through to models from different providers (e.g. DeepSeek at position 4). Instead, it retries the same rate-limited provider and gives up with `FailoverError`.
... | closed | completed | true | 4 | [] | [] | 2026-03-09T22:39:12Z | 2026-04-27T04:56:15Z | 2026-04-24T21:11:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | moeedahmed | 5,780,040 | MDQ6VXNlcjU3ODAwNDA= | User | false |
openclaw/openclaw | 4,048,289,108 | I_kwDOQb6kR87xS_1U | 41,493 | https://github.com/openclaw/openclaw/issues/41493 | https://api.github.com/repos/openclaw/openclaw/issues/41493 | Feature: Surface model errors to user when fallback chain fails | ## Problem
When a model in the fallback chain fails (e.g., with `errorMessage: "This operation was aborted"`), the user sees nothing. The response is silently lost, leaving users wondering why Kai doesn't respond ~40% of the time.
## Current Behavior
1. Primary model fails with an error (e.g., Ollama cloud timeout/a... | closed | completed | true | 1 | [] | [] | 2026-03-09T22:39:14Z | 2026-04-27T04:56:15Z | 2026-04-24T21:11:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kai-the-man[bot] | 259,478,402 | BOT_kgDOD3dTgg | Bot | false |
openclaw/openclaw | 4,048,372,348 | I_kwDOQb6kR87xTUJ8 | 41,505 | https://github.com/openclaw/openclaw/issues/41505 | https://api.github.com/repos/openclaw/openclaw/issues/41505 | [Bug]: WhatsApp multi-account ackReaction and allowFrom ignore account-level overrides | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In multi-account WhatsApp deployments, `maybeSendAckReaction` reads `ackReaction` from `cfg.channels.whatsapp.ackReaction` (root level) instead of resolving the account-level override via `resolveWhatsAppAccount()`. Similarly, `buildMention... | closed | completed | true | 1 | [] | [] | 2026-03-09T23:02:49Z | 2026-04-27T04:56:14Z | 2026-04-24T21:11:10Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | asyncjason | 4,665,758 | MDQ6VXNlcjQ2NjU3NTg= | User | false |
openclaw/openclaw | 4,048,400,962 | I_kwDOQb6kR87xTbJC | 41,512 | https://github.com/openclaw/openclaw/issues/41512 | https://api.github.com/repos/openclaw/openclaw/issues/41512 | Test Issue: File Handling with Chinese Characters | ## Description
This is a test issue to verify file handling functionality with Chinese (non-ASCII) characters in filenames and content.
## Test Details
- **Test File**: `中文测试文件.txt` (Chinese Test File.txt)
- **File Location**: `/mnt/e/test/`
- **Content**: Chinese text content for encoding verification
## Purpose
... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T23:11:26Z | 2026-04-27T04:56:13Z | 2026-04-24T21:11:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | liuzongwen | 20,163,084 | MDQ6VXNlcjIwMTYzMDg0 | User | false |
openclaw/openclaw | 4,048,407,212 | I_kwDOQb6kR87xTcqs | 41,514 | https://github.com/openclaw/openclaw/issues/41514 | https://api.github.com/repos/openclaw/openclaw/issues/41514 | Health monitor emits system events for channels removed from config | ## Bug Description
The gateway's built-in health monitor continues to emit system events for channels that have been **completely removed** from `openclaw.json`, including both `channels` and `plugins.entries` sections.
## Steps to Reproduce
1. Configure a BlueBubbles channel in `openclaw.json` (under `channels` and... | closed | completed | true | 1 | [] | [] | 2026-03-09T23:12:57Z | 2026-04-27T04:56:13Z | 2026-04-24T21:11:17Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | AbebaCo | 262,210,827 | U_kgDOD6EFCw | User | false |
openclaw/openclaw | 4,048,407,822 | I_kwDOQb6kR87xTc0O | 41,515 | https://github.com/openclaw/openclaw/issues/41515 | https://api.github.com/repos/openclaw/openclaw/issues/41515 | Embedded runner reports 'API rate limit reached' when provider has full headroom | ## Bug Description
The embedded agent runner reports `API rate limit reached. Please try again later.` and retries with exponential backoff, even when the actual API provider has full rate limit headroom available.
## Steps to Reproduce
1. Set system default model to `anthropic/claude-haiku-4-5-20251001` (API key au... | closed | completed | true | 5 | [] | [] | 2026-03-09T23:13:10Z | 2026-04-27T04:56:12Z | 2026-04-24T21:11:20Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | AbebaCo | 262,210,827 | U_kgDOD6EFCw | User | false |
openclaw/openclaw | 4,072,224,059 | I_kwDOQb6kR87yuTU7 | 45,300 | https://github.com/openclaw/openclaw/issues/45300 | https://api.github.com/repos/openclaw/openclaw/issues/45300 | Bug: requireMention: true broken in multi-account Discord config — all messages return no-mention | ## Summary
`requireMention: true` does not work in a multi-account Discord configuration. Every guild message is dropped at the preflight stage with `reason: "no-mention"`, even when the bot is explicitly @mentioned by a user in the allowlist.
## Environment
- OpenClaw version: 2026.3.12
- OS: macOS 15.6 (arm64)
- D... | closed | completed | false | 3 | [] | [] | 2026-03-13T16:45:30Z | 2026-04-25T08:31:47Z | 2026-04-25T08:31:47Z | NONE | null | 20260425T090015Z | 2026-04-25T09:00:15Z | zachware | 57,385 | MDQ6VXNlcjU3Mzg1 | User | false |
openclaw/openclaw | 4,123,554,799 | I_kwDOQb6kR871yHPv | 53,132 | https://github.com/openclaw/openclaw/issues/53132 | https://api.github.com/repos/openclaw/openclaw/issues/53132 | Discord: multi-account gateway startup hangs at 'awaiting gateway readiness' after Carbon reconcile change (v2026.3.22) | ## Summary
Running 4 Discord bot accounts on a single gateway causes 2–4 bots to hang indefinitely at `client initialized as <id> (<name>); awaiting gateway readiness` on every restart. The issue appeared after upgrading from v2026.3.13 to v2026.3.22 with the same configuration and bot tokens. v2026.3.13 starts all 4 ... | closed | completed | true | 7 | [] | [] | 2026-03-23T20:32:50Z | 2026-04-27T04:56:10Z | 2026-04-24T21:13:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Skeptomenos | 22,348,986 | MDQ6VXNlcjIyMzQ4OTg2 | User | false |
openclaw/openclaw | 4,086,056,271 | I_kwDOQb6kR87zjEVP | 48,712 | https://github.com/openclaw/openclaw/issues/48712 | https://api.github.com/repos/openclaw/openclaw/issues/48712 | [Bug][Critical] LINE/Telegram webhook routes silently unregistered — all inbound messages dropped (404) | ## Summary
**All LINE webhook POST requests return 404**, causing the bot to silently drop every inbound message. The gateway shows LINE providers as "running" and "configured" in status/doctor output, giving no indication that webhooks are completely broken. This has been present since at least `2026.3.12` through `2... | closed | completed | true | 3 | [] | [] | 2026-03-17T03:57:18Z | 2026-04-27T04:56:09Z | 2026-04-24T21:13:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Vigoose | 5,359,336 | MDQ6VXNlcjUzNTkzMzY= | User | false |
openclaw/openclaw | 4,075,146,063 | I_kwDOQb6kR87y5ctP | 45,888 | https://github.com/openclaw/openclaw/issues/45888 | https://api.github.com/repos/openclaw/openclaw/issues/45888 | [Bug] LINE webhook returns 404 - Plugin HTTP routes not registered | ## Description
The LINE channel webhook returns 404 even when properly configured. Investigation reveals that plugin HTTP routes are not being registered correctly in the gateway's HTTP handler.
## Environment
- OpenClaw version: 2026.3.13
- Node.js version: 22.22.0
- OS: Ubuntu 22.04 LTS
## Configuration
```json
... | closed | completed | true | 7 | [] | [] | 2026-03-14T08:31:12Z | 2026-04-27T04:56:09Z | 2026-04-24T21:14:51Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lam7-ddd | 56,620,158 | MDQ6VXNlcjU2NjIwMTU4 | User | false |
openclaw/openclaw | 4,075,727,506 | I_kwDOQb6kR87y7qqS | 46,108 | https://github.com/openclaw/openclaw/issues/46108 | https://api.github.com/repos/openclaw/openclaw/issues/46108 | [Bug]: Synology Chat Plugin: /webhook/synology liefert immer 404, obwohl Route registriert ist (v2026.3.13) Umgebung | ### Bug type
Regression (worked before, now fails)
### Summary
Gateway: OpenClaw 2026.3.13 (Gateway laut systemctl --user status openclaw-gateway)
OS: Linux 6.17.0‑14‑generic (x64)
Node: 25.8.0
Gateway‑Bind: bind=lan, port=18789
Dashboard: http://10.0.0.124:18789/
Synology NAS:
URL: http:... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-14T12:32:32Z | 2026-04-27T04:56:07Z | 2026-04-24T21:14:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DJGIG83 | 78,141,334 | MDQ6VXNlcjc4MTQxMzM0 | User | false |
openclaw/openclaw | 4,089,778,597 | I_kwDOQb6kR87zxRGl | 49,143 | https://github.com/openclaw/openclaw/issues/49143 | https://api.github.com/repos/openclaw/openclaw/issues/49143 | [Bug]: | ### Bug type
Regression (worked before, now fails)
### Summary
# OpenClaw Google Chat route registers but POST /googlechat still returns 404
## Summary
Google Chat channel appears to start successfully and registers `/googlechat`, but real HTTP requests to the webhook still return `404 Not Found`. This causes Googl... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-03-17T16:46:16Z | 2026-04-27T04:56:06Z | 2026-04-24T21:14:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | phkundi | 40,868,562 | MDQ6VXNlcjQwODY4NTYy | User | false |
openclaw/openclaw | 4,226,573,957 | I_kwDOQb6kR8777GaF | 63,285 | https://github.com/openclaw/openclaw/issues/63285 | https://api.github.com/repos/openclaw/openclaw/issues/63285 | [Bug]: Inbound media staged to global ~/.openclaw/media/inbound/ instead of agent workspace path, making files unreachable in sandboxed agents | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When a file is uploaded to Slack and received by a sandboxed agent, the inbound media
staging pipeline writes the file to the **global** `~/.openclaw/media/inbound/` path.
The path injected into the agent's pro... | closed | completed | true | 2 | [] | [] | 2026-04-08T18:00:21Z | 2026-04-27T04:56:05Z | 2026-04-24T21:15:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | brantley-creator | 245,261,988 | U_kgDODp5mpA | User | false |
openclaw/openclaw | 4,190,335,898 | I_kwDOQb6kR875w3Oa | 59,278 | https://github.com/openclaw/openclaw/issues/59278 | https://api.github.com/repos/openclaw/openclaw/issues/59278 | Discord inbound media paths are inaccessible to read tool outside workspace sandbox | Observed in a Discord-agent runtime using the read tool. When a user uploads an image in a Discord channel, the inbound media is exposed in the message envelope as an absolute path under `~/.openclaw/media/inbound/...`, but `read` rejects it because it escapes the workspace sandbox.
Symptom:
- inbound message includes... | closed | completed | true | 1 | [] | [] | 2026-04-01T22:45:31Z | 2026-04-27T04:56:05Z | 2026-04-24T21:15:08Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | hicg-odoo | 169,425,961 | U_kgDOChk8KQ | User | false |
openclaw/openclaw | 4,048,454,604 | I_kwDOQb6kR87xToPM | 41,522 | https://github.com/openclaw/openclaw/issues/41522 | https://api.github.com/repos/openclaw/openclaw/issues/41522 | TUI: session popover preview shows raw security notice wrapper instead of meaningful content | ## Bug Report
### Version
OpenClaw 2026.3.2 (85377a2)
### Surface
TUI session popover (sessions list / `/sessions`)
### Description
The session popover in the TUI displays the **last message snippet** from each session's transcript as the preview text. When a session has run `web_fetch` (or any tool that returns ext... | closed | completed | true | 2 | [] | [] | 2026-03-09T23:24:59Z | 2026-04-27T04:56:02Z | 2026-04-24T21:15:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bryanbaer | 22,436,505 | MDQ6VXNlcjIyNDM2NTA1 | User | false |
openclaw/openclaw | 4,048,472,321 | I_kwDOQb6kR87xTskB | 41,526 | https://github.com/openclaw/openclaw/issues/41526 | https://api.github.com/repos/openclaw/openclaw/issues/41526 | [Bug]: Automatic replies fail silently after the first successful message on a new channel | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Summary
After a new channel (tested with both Telegram and Slack) is configured with dmPolicy: "pairing" and a user is successfully paired, the OpenClaw gateway processes and delivers the reply to the user's very first message. However, al... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T23:29:58Z | 2026-04-27T04:56:02Z | 2026-04-24T21:15:27Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | akcakirahmet | 63,654,377 | MDQ6VXNlcjYzNjU0Mzc3 | User | false |
openclaw/openclaw | 4,048,502,467 | I_kwDOQb6kR87xTz7D | 41,531 | https://github.com/openclaw/openclaw/issues/41531 | https://api.github.com/repos/openclaw/openclaw/issues/41531 | Bug: Telegram Topics misrouting across groups (background jobs mix chat_id and message_thread_id) | ### Summary
Telegram forum topic routing occasionally breaks for background/cron messages when multiple supergroups with Topics are involved. Messages intended for a specific topic/thread in one group sometimes arrive in another group’s root/general.
### Expected
Outbound Telegram sends from skills/cron/background tas... | closed | completed | true | 1 | [] | [] | 2026-03-09T23:39:03Z | 2026-04-27T04:56:01Z | 2026-04-24T21:15:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | zeitgeistlive | 261,338,766 | U_kgDOD5O2jg | User | false |
openclaw/openclaw | 4,048,524,565 | I_kwDOQb6kR87xT5UV | 41,538 | https://github.com/openclaw/openclaw/issues/41538 | https://api.github.com/repos/openclaw/openclaw/issues/41538 | [Bug]: update.run (and gateway restart) should flush open session JSONL transcripts before SIGTERM | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When update.run is triggered, the gateway sends SIGTERM and restarts immediately. Any active session's JSONL transcript that hasn't been fully flushed to disk is lost. The webchat UI caches the conversation client-side (so the user can stil... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T23:46:45Z | 2026-04-27T04:56:00Z | 2026-04-24T21:15:36Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | oliverchoy | 282,752 | MDQ6VXNlcjI4Mjc1Mg== | User | false |
openclaw/openclaw | 4,048,528,322 | I_kwDOQb6kR87xT6PC | 41,539 | https://github.com/openclaw/openclaw/issues/41539 | https://api.github.com/repos/openclaw/openclaw/issues/41539 | Bug: dangerouslyDisableDeviceAuth not applied in WebSocket validation (Control UI) | ## Bug Report
**Version:** OpenClaw 2026.3.8 (3caab92)
**OS:** Linux (Ubuntu)
### Description
The `gateway.controlUi.dangerouslyDisableDeviceAuth` flag is loaded and appears in security warnings, but is not applied during WebSocket connection validation. Control UI connections are still rejected with "device identity... | closed | completed | true | 2 | [] | [] | 2026-03-09T23:48:07Z | 2026-04-27T04:55:59Z | 2026-04-24T21:15:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mlg-crypto | 73,039,385 | MDQ6VXNlcjczMDM5Mzg1 | User | false |
openclaw/openclaw | 4,048,530,014 | I_kwDOQb6kR87xT6pe | 41,540 | https://github.com/openclaw/openclaw/issues/41540 | https://api.github.com/repos/openclaw/openclaw/issues/41540 | Bug: Brave llm-context mode returns empty snippets in web_search | ## Description
When `tools.web.search.brave.mode` is set to `"llm-context"`, the `web_search` tool returns results with an empty `snippets: []` array even though the Brave LLM Context API endpoint is returning content correctly.
## Steps to Reproduce
1. Set config: `tools.web.search.brave.mode = "llm-context"`
2. Ca... | closed | completed | true | 2 | [] | [] | 2026-03-09T23:48:49Z | 2026-04-27T04:55:59Z | 2026-04-24T21:15:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | luiachan-svg | 257,427,697 | U_kgDOD1gI8Q | User | false |
openclaw/openclaw | 4,048,533,163 | I_kwDOQb6kR87xT7ar | 41,541 | https://github.com/openclaw/openclaw/issues/41541 | https://api.github.com/repos/openclaw/openclaw/issues/41541 | Bug: TUI messages display in reverse chronological order / session log overwritten | ## Description
Three related symptoms observed in the TUI (terminal UI), likely same root cause:
**1. Reverse chronological ordering**
Message inputs are displayed newest-first instead of oldest-first. The conversation appears upside-down — most recent messages appear at the top of the display.
**2. Session log over... | closed | completed | true | 1 | [] | [] | 2026-03-09T23:50:01Z | 2026-04-27T04:55:58Z | 2026-04-24T21:15:49Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dgarf29 | 63,170,957 | MDQ6VXNlcjYzMTcwOTU3 | User | false |
openclaw/openclaw | 4,048,535,602 | I_kwDOQb6kR87xT8Ay | 41,542 | https://github.com/openclaw/openclaw/issues/41542 | https://api.github.com/repos/openclaw/openclaw/issues/41542 | bug: local OpenAI-compatible endpoints (e.g. local-dgx-spark) report zero token usage in session transcripts and cron run logs | ## Problem
When using local OpenAI-compatible endpoints (configured as `local-dgx-spark` or `local-macbook-pro` providers, e.g. pointing to `http://192.168.1.152:8002/v1`), **token usage is not recorded** in session transcripts or cron run logs. All usage fields are zero:
```json
{
"input": 0,
"output": 0,
"cac... | closed | completed | true | 5 | [] | [] | 2026-03-09T23:51:03Z | 2026-04-27T04:55:57Z | 2026-04-24T21:15:53Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | JonathanJing | 17,068,507 | MDQ6VXNlcjE3MDY4NTA3 | User | false |
openclaw/openclaw | 4,073,301,513 | I_kwDOQb6kR87yyaYJ | 45,445 | https://github.com/openclaw/openclaw/issues/45445 | https://api.github.com/repos/openclaw/openclaw/issues/45445 | [Bug]: Webhook routes (Google Chat, LINE, etc.) return 404 — httpRoutes lost on plugin registry swap (v2026.3.12) | <!-- ~/.openclaw/GITHUB-ISSUE-BODY.md -->
<!-- Copy this into: https://github.com/openclaw/openclaw/issues/new -->
<!-- Title: [Bug]: Webhook routes (Google Chat, LINE, etc.) return 404 — httpRoutes lost on plugin registry swap (v2026.3.12) -->
### Bug type
Regression (worked before, now fails)
### Summary
Google C... | closed | completed | true | 10 | [] | [] | 2026-03-13T20:38:26Z | 2026-04-27T04:55:56Z | 2026-04-24T21:15:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | George-Claw | 266,208,739 | U_kgDOD94F4w | User | false |
openclaw/openclaw | 4,074,956,003 | I_kwDOQb6kR87y4uTj | 45,822 | https://github.com/openclaw/openclaw/issues/45822 | https://api.github.com/repos/openclaw/openclaw/issues/45822 | [Bug] Synology Chat webhook fails in v2026.3.12+ (works in v2026.3.11) | ## Bug Description
Synology Chat integration stopped working after upgrading from v2026.3.11 to v2026.3.12 and v2026.3.13.
Error: "Unable to send request to bot server. Please contact the bot owner."
## Environment
- OpenClaw: v2026.3.13
- Synology DSM: latest
- Synology Chat webhook: enabled
## Configuration
- dmPo... | closed | completed | true | 7 | [] | [] | 2026-03-14T07:10:00Z | 2026-04-27T04:55:56Z | 2026-04-24T21:15:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | roanpy | 41,194,472 | MDQ6VXNlcjQxMTk0NDcy | User | false |
openclaw/openclaw | 4,072,764,360 | I_kwDOQb6kR87ywXPI | 45,378 | https://github.com/openclaw/openclaw/issues/45378 | https://api.github.com/repos/openclaw/openclaw/issues/45378 | [Bug] Synology Chat webhook route logs as registered but POST /webhook/synology returns 404 after local↔lan switch | ## Summary
After upgrading to **OpenClaw 2026.3.12**, Synology Chat inbound webhooks stopped working in my environment.
The key symptom is:
- OpenClaw logs say the Synology Chat route is registered:
- `Starting Synology Chat channel (account: default, path: /webhook/synology)`
- `Registered HTTP route: /webhook/... | closed | completed | true | 4 | [] | [] | 2026-03-13T18:32:45Z | 2026-04-27T04:55:54Z | 2026-04-24T21:15:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | caox488-a11y | 266,483,029 | U_kgDOD-I1VQ | User | false |
openclaw/openclaw | 4,048,654,673 | I_kwDOQb6kR87xUZFR | 41,562 | https://github.com/openclaw/openclaw/issues/41562 | https://api.github.com/repos/openclaw/openclaw/issues/41562 | Discord: thread session delivery leaks to main session target (#logs) when fallback triggers | ## Description
Thread sessions created in Discord channels occasionally route their responses to the main session's delivery target instead of the originating thread. The leak destination is always the heartbeat `to` channel (in my case, `#logs`).
## Repro
1. Configure `agents.defaults.heartbeat.to` to point at a sp... | closed | completed | true | 2 | [] | [] | 2026-03-10T00:33:15Z | 2026-04-27T04:55:55Z | 2026-04-24T21:15:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | montrellcruse | 130,179,762 | U_kgDOB8Jisg | User | false |
openclaw/openclaw | 4,076,249,133 | I_kwDOQb6kR87y9qAt | 46,408 | https://github.com/openclaw/openclaw/issues/46408 | https://api.github.com/repos/openclaw/openclaw/issues/46408 | Google Chat webhook route returns 404 despite channel showing as running | ### Bug type
Regression (worked before, now fails)
### Summary
The Google Chat channel reports 'enabled, configured, running, dm:open' via openclaw channels status and the --probe flag says 'works', but all HTTP POST requests to the /googlechat webhook path return 404. No inbound requests from Google Chat ever appea... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-14T17:01:25Z | 2026-04-27T04:55:53Z | 2026-04-24T21:16:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | propello-ventures | 261,936,119 | U_kgDOD5zT9w | User | false |
openclaw/openclaw | 4,048,661,484 | I_kwDOQb6kR87xUavs | 41,564 | https://github.com/openclaw/openclaw/issues/41564 | https://api.github.com/repos/openclaw/openclaw/issues/41564 | [Feature Request] Allow cron wake action in subagent sessions | ## Problem
Subagents cannot call `cron wake` to notify the main session when work completes. The `cron` tool has `ownerOnly: true` in the framework code (`OWNER_ONLY_TOOL_NAME_FALLBACKS` in `pi-tools.policy`), which removes it from the tool list **before** the configurable `alsoAllow` policy can take effect. This mean... | closed | completed | true | 1 | [] | [] | 2026-03-10T00:36:01Z | 2026-04-27T04:55:53Z | 2026-04-24T21:16:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jacobaross | 16,659,674 | MDQ6VXNlcjE2NjU5Njc0 | User | false |
openclaw/openclaw | 4,073,976,528 | I_kwDOQb6kR87y0_LQ | 45,544 | https://github.com/openclaw/openclaw/issues/45544 | https://api.github.com/repos/openclaw/openclaw/issues/45544 | [Bug]: All Google Chat webhooks return 404 after 2026.3.12 update (jiti/ESM registry split) | ## Summary
After updating to **2026.3.12**, all Google Chat webhook paths (`/googlechat-bender`, `/googlechat-eliza`, etc.) return 404. The gateway logs show all accounts starting successfully (`[googlechat] [bender] starting Google Chat webhook`), but every inbound request returns 404.
## Environment
- OpenClaw: 20... | closed | completed | true | 5 | [] | [] | 2026-03-14T00:00:03Z | 2026-04-27T04:55:52Z | 2026-04-24T21:16:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jaybna | 397,804 | MDQ6VXNlcjM5NzgwNA== | User | false |
openclaw/openclaw | 4,048,680,281 | I_kwDOQb6kR87xUfVZ | 41,567 | https://github.com/openclaw/openclaw/issues/41567 | https://api.github.com/repos/openclaw/openclaw/issues/41567 | Silent delivery failure: Telegram onError handler swallows errors without logging | ## Bug
When the Telegram delivery of an assistant reply fails (rate limit, network blip, markdown rendering error, etc.), the error is silently swallowed. No log entry is written anywhere, making these delivery drops completely invisible.
## What happened
1. Agent produced a valid final response (KB ingest confirmat... | closed | completed | true | 1 | [] | [] | 2026-03-10T00:42:48Z | 2026-04-27T04:55:51Z | 2026-04-24T21:16:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mberman84 | 748,450 | MDQ6VXNlcjc0ODQ1MA== | User | false |
openclaw/openclaw | 4,048,698,368 | I_kwDOQb6kR87xUjwA | 41,571 | https://github.com/openclaw/openclaw/issues/41571 | https://api.github.com/repos/openclaw/openclaw/issues/41571 | Bug: stripDanglingAnthropicToolUses checks wrong type ("toolUse" vs "toolCall") and empty assistant messages stored on LLM failure cause session corruption with thinking blocks | ## Summary
Two related bugs discovered after investigating a reproducible `LLM request rejected: messages.3.content.2: thinking or redacted_thinking blocks in the latest assistant message cannot be modified` error on a long-running Telegram DM session.
**Reported by:** @Caml2001 (camlwh013241@gmail.com)
---
## Bug ... | closed | completed | true | 2 | [] | [] | 2026-03-10T00:49:21Z | 2026-04-27T04:55:50Z | 2026-04-24T21:16:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Caml2001 | 96,397,791 | U_kgDOBb7p3w | User | false |
openclaw/openclaw | 4,048,707,321 | I_kwDOQb6kR87xUl75 | 41,574 | https://github.com/openclaw/openclaw/issues/41574 | https://api.github.com/repos/openclaw/openclaw/issues/41574 | [Bug]: TTS audio not playing in WebChat/control-ui | ### Bug type
Regression (worked before, now fails)
### Summary
TTS audio is not played in WebChat/control-ui after reply is generated
### Steps to reproduce
1. Configure TTS with Edge TTS (messages.tts.auto: "always", provider: "edge")
2. Send a message from WebChat (control-ui)
3. AI responds with text, TTS audio... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-10T00:52:46Z | 2026-04-27T04:55:50Z | 2026-04-24T21:16:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | declan2010 | 26,747,598 | MDQ6VXNlcjI2NzQ3NTk4 | User | false |
openclaw/openclaw | 4,086,550,461 | I_kwDOQb6kR87zk8-9 | 48,790 | https://github.com/openclaw/openclaw/issues/48790 | https://api.github.com/repos/openclaw/openclaw/issues/48790 | [Bug]: Channel plugin registrations lost on runtime registry swap — message tool "Unknown channel" for configured channels | ## Description
`getChannelPlugin()` returns `undefined` for properly configured core channels (e.g. Telegram) after the active plugin registry is replaced at runtime. This causes the message tool to fail with `Unknown channel: telegram` even though the channel is in `CHANNEL_IDS` and was fully functional at startup.
... | closed | completed | true | 3 | [] | [] | 2026-03-17T06:41:13Z | 2026-04-27T04:55:49Z | 2026-04-24T21:16:20Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mekenthompson | 8,923,913 | MDQ6VXNlcjg5MjM5MTM= | User | false |
openclaw/openclaw | 4,048,725,088 | I_kwDOQb6kR87xUqRg | 41,575 | https://github.com/openclaw/openclaw/issues/41575 | https://api.github.com/repos/openclaw/openclaw/issues/41575 | Feature request: Simplified API for external services to notify agents (auth callbacks, webhooks) | **Problem:**
External services (like auth servers, webhooks, monitoring systems) cannot easily notify OpenClaw agents to trigger actions without implementing complex WebSocket device authentication.
**Use case:**
When a user completes authentication in an external auth system, we want to automatically notify the agent... | closed | completed | true | 1 | [] | [] | 2026-03-10T00:59:39Z | 2026-04-27T04:55:48Z | 2026-04-24T21:16:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | felipematos | 775,850 | MDQ6VXNlcjc3NTg1MA== | User | false |
openclaw/openclaw | 4,048,738,923 | I_kwDOQb6kR87xUtpr | 41,576 | https://github.com/openclaw/openclaw/issues/41576 | https://api.github.com/repos/openclaw/openclaw/issues/41576 | age channel leaks [[reply_to:ID]] tags into visible message text | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
`sanitizeOutboundText()` in the iMessage channel does not strip `[[reply_to:<id>]]` directive tags before delivery, causing raw metadata to appear in user-visible iMessage replies.
### Steps to reproduce
1. Configure iMessage channel and ... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T01:05:11Z | 2026-04-27T04:55:47Z | 2026-04-24T21:16:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kingmadellc | 257,369,583 | U_kgDOD1cl7w | User | false |
openclaw/openclaw | 4,048,811,100 | I_kwDOQb6kR87xU_Rc | 41,586 | https://github.com/openclaw/openclaw/issues/41586 | https://api.github.com/repos/openclaw/openclaw/issues/41586 | [Feature Request] Allow cron wake action in subagent sessions | ## Problem
Subagents cannot call `cron wake` to notify the main session when work completes. The `cron` tool has `ownerOnly: true` in the framework code (`OWNER_ONLY_TOOL_NAME_FALLBACKS` in `pi-tools.policy`), which removes it from the tool list entirely before the configurable `alsoAllow` policy can take effect.
Thi... | closed | completed | true | 1 | [] | [] | 2026-03-10T01:29:17Z | 2026-04-27T04:55:47Z | 2026-04-24T21:16:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jacobaross | 16,659,674 | MDQ6VXNlcjE2NjU5Njc0 | User | false |
openclaw/openclaw | 4,048,831,455 | I_kwDOQb6kR87xVEPf | 41,595 | https://github.com/openclaw/openclaw/issues/41595 | https://api.github.com/repos/openclaw/openclaw/issues/41595 | Anthropic→Gemini failover: function_response.name is empty, causes INVALID_ARGUMENT | ## Bug Report
**OpenClaw Version:** 2026.3.8 (3caab92)
**Severity:** High — failover silently produces invalid API requests
### What happens
When the primary Anthropic profile times out and OpenClaw fails over to a Google/Gemini profile, the Anthropic→Gemini message format conversion leaves `function_response.name`... | closed | completed | true | 3 | [] | [] | 2026-03-10T01:36:53Z | 2026-04-27T04:55:46Z | 2026-04-24T21:16:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DavidNielsen1031 | 37,390,726 | MDQ6VXNlcjM3MzkwNzI2 | User | false |
openclaw/openclaw | 4,048,890,513 | I_kwDOQb6kR87xVSqR | 41,602 | https://github.com/openclaw/openclaw/issues/41602 | https://api.github.com/repos/openclaw/openclaw/issues/41602 | Rusty workstation unreachable - CU services down (8420/8421/8422) | ## Overnight Audit: 2026-03-10 01:55 UTC
### Issue
Rusty workstation (100.66.108.46) is completely unreachable:
- Ping: 100% packet loss
- SSH: Connection timed out
- CU services (8420/8421/8422): Inaccessible
### Impact
- CU Daemon (8420): Screenshots unavailable
- Task Server (8421): AI feedback loop down
- Workflo... | closed | not_planned | true | 1 | [] | [] | 2026-03-10T01:57:16Z | 2026-04-27T04:55:45Z | 2026-04-24T21:16:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,048,891,673 | I_kwDOQb6kR87xVS8Z | 41,604 | https://github.com/openclaw/openclaw/issues/41604 | https://api.github.com/repos/openclaw/openclaw/issues/41604 | [Feature]: Skip or silence Anthropic credential sync log when not using Anthropic models | ## Summary
Skip or silence the "synced anthropic oauth credentials from claude cli" log message when `agents.defaults.model.primary` is set to a non-Anthropic model (e.g., `kimi-coding/k2p5`).
## Problem to solve
Currently, `syncExternalCliCredentials()` in `src/agents/auth-profiles/external-cli-sync.ts` uncondition... | closed | completed | true | 4 | [] | [] | 2026-03-10T01:57:44Z | 2026-04-27T04:55:45Z | 2026-04-24T21:16:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | yumesha | 237,216,263 | U_kgDODiOiBw | User | false |
openclaw/openclaw | 4,048,900,568 | I_kwDOQb6kR87xVVHY | 41,606 | https://github.com/openclaw/openclaw/issues/41606 | https://api.github.com/repos/openclaw/openclaw/issues/41606 | [Bug]: Memory/skills chokidar watcher can exhaust FDs (FSEventWrap) when watched dirs have many files | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## What happens
When the gateway runs and the workspace has a large number of files under paths watched by the **memory** and **skills** chokidar watchers (e.g. `memory/` or extraPaths with many session/memory files), the process can open 9... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T02:01:07Z | 2026-04-27T04:55:44Z | 2026-04-24T21:16:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | JaneyChat | 260,345,318 | U_kgDOD4SN5g | User | false |
openclaw/openclaw | 4,048,956,729 | I_kwDOQb6kR87xVi05 | 41,634 | https://github.com/openclaw/openclaw/issues/41634 | https://api.github.com/repos/openclaw/openclaw/issues/41634 | [Bug]: Legacy auth profiles accumulate indefinitely and cannot be cleaned up (profile order silently wrong, subagent sessions hang) | ## Summary
Auth profiles accumulate in `auth-profiles.json` over time and can never be cleaned up. Legacy profiles (e.g. `anthropic:manual`, `anthropic:user-me.com`, `anthropic:default`, `anthropic:claude-cli`) pile up silently and corrupt the auth order, causing requests to silently use the wrong account.
## OpenCla... | closed | completed | true | 3 | [] | [] | 2026-03-10T02:22:17Z | 2026-04-27T04:55:43Z | 2026-04-24T21:20:20Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | amittell | 1,388,680 | MDQ6VXNlcjEzODg2ODA= | User | false |
openclaw/openclaw | 4,049,072,459 | I_kwDOQb6kR87xV_FL | 41,666 | https://github.com/openclaw/openclaw/issues/41666 | https://api.github.com/repos/openclaw/openclaw/issues/41666 | [Bug] Gateway event loop stalls (79+ min) when Telegram 400 'message too long' error occurs during concurrent cron burst | ## Summary
The OpenClaw gateway event loop can freeze completely for extended periods (observed: 79 minutes) when a Telegram `400 Bad Request: message is too long` error occurs during a concurrent cron job burst. The process stays alive (launchd does not restart it) but stops processing all events — Telegram sends, We... | closed | completed | true | 1 | [] | [] | 2026-03-10T03:03:25Z | 2026-04-27T04:55:42Z | 2026-04-24T21:20:28Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rex05ai | 258,758,601 | U_kgDOD2xXyQ | User | false |
openclaw/openclaw | 4,049,074,052 | I_kwDOQb6kR87xV_eE | 41,667 | https://github.com/openclaw/openclaw/issues/41667 | https://api.github.com/repos/openclaw/openclaw/issues/41667 | [Bug] Cron job delivery routes to wrong topic after gateway restart mid-announce (1012 close) | ## Summary
When the gateway restarts (via launchd KeepAlive after a crash/freeze) while a cron job's announce delivery is in-flight, the re-run of that cron job delivers its output to a different Telegram topic than configured. The `delivery.to` field in `cron/jobs.json` is correct, but the message lands in the wrong ... | closed | completed | true | 1 | [] | [] | 2026-03-10T03:03:51Z | 2026-04-27T04:55:42Z | 2026-04-24T21:20:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rex05ai | 258,758,601 | U_kgDOD2xXyQ | User | false |
openclaw/openclaw | 4,049,120,422 | I_kwDOQb6kR87xWKym | 41,681 | https://github.com/openclaw/openclaw/issues/41681 | https://api.github.com/repos/openclaw/openclaw/issues/41681 | Response Mismatch - Current Question Returns Previous Answer | # Description
When using OpenClaw, there are occasions where the response to a newly submitted question contains the answer from a previous conversation turn instead of answering the current question.
# Reproduction Steps
1. Have a conversation with OpenClaw
2. Submit a new question
3. Sometimes the response is the an... | closed | completed | true | 1 | [] | [] | 2026-03-10T03:21:41Z | 2026-04-27T04:55:41Z | 2026-04-24T21:20:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | xfwangqiang | 14,222,018 | MDQ6VXNlcjE0MjIyMDE4 | User | false |
openclaw/openclaw | 4,049,147,876 | I_kwDOQb6kR87xWRfk | 41,683 | https://github.com/openclaw/openclaw/issues/41683 | https://api.github.com/repos/openclaw/openclaw/issues/41683 | Opus 4.6 1M context: is context1m auto-applied in v2026.3.8, and does it work with Claude Max OAuth? | **Follow-up to #19849** (locked, can't comment)
Following the fix in PR #19789, we've been testing Opus 4.6 1M context on a Claude Max 5x subscription (token auth via Claude Code OAuth, v2026.3.8 on Windows 10).
**What we see:**
\/status\ shows \Context: 0/1.1m\ for \nthropic/claude-opus-4-6\ with \contextTokens: 11... | closed | completed | true | 2 | [] | [] | 2026-03-10T03:29:19Z | 2026-04-27T04:55:40Z | 2026-04-24T21:20:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Stache73 | 7,196,384 | MDQ6VXNlcjcxOTYzODQ= | User | false |
openclaw/openclaw | 4,049,202,500 | I_kwDOQb6kR87xWe1E | 41,699 | https://github.com/openclaw/openclaw/issues/41699 | https://api.github.com/repos/openclaw/openclaw/issues/41699 | [Bug]: Directive whitespace normalization strips indentation from code blocks in outbound replies | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Code blocks in chat replies lose their leading indentation before being sent to channels.
The message still keeps:
• fenced code blocks
• line breaks
But indentation inside the code block is removed, which makes JSON / Python / YAML much h... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T03:46:35Z | 2026-04-27T04:55:39Z | 2026-04-24T21:20:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Boer2333 | 137,675,429 | U_kgDOCDTCpQ | User | false |
openclaw/openclaw | 4,084,510,109 | I_kwDOQb6kR87zdK2d | 48,465 | https://github.com/openclaw/openclaw/issues/48465 | https://api.github.com/repos/openclaw/openclaw/issues/48465 | Plugin channels not discoverable by message tool for proactive sends | ## Problem
The `message` tool (used by agents to send proactive messages) cannot route to plugin-registered channels. Specifically, `telegram-userbot` is registered via `api.registerChannel()` and appears in `listPluginChannelIds()`, but the message tool rejects sends with:
```
Unknown target "-1003812100686" for Tel... | closed | completed | true | 3 | [] | [] | 2026-03-16T19:59:21Z | 2026-04-27T04:55:39Z | 2026-04-24T21:20:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | MaxVidkrytaklishnya | 263,621,344 | U_kgDOD7aK4A | User | false |
openclaw/openclaw | 4,049,204,588 | I_kwDOQb6kR87xWfVs | 41,700 | https://github.com/openclaw/openclaw/issues/41700 | https://api.github.com/repos/openclaw/openclaw/issues/41700 | [Bug]: install ERROR | ### Bug type
Regression (worked before, now fails)
### Summary
when I excute this shell 'sudo openclaw onboard --install-daemon'
It report this error :
[openclaw] Uncaught exception: Error: read EIO
at TTY.onStreamRead (node:internal/stream_base_commons:216:20)
### Steps to reproduce
1. reinstall openclaw
2... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-10T03:47:16Z | 2026-04-27T04:55:38Z | 2026-04-24T21:20:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | skymoon08 | 13,800,048 | MDQ6VXNlcjEzODAwMDQ4 | User | false |
openclaw/openclaw | 4,049,218,719 | I_kwDOQb6kR87xWiyf | 41,703 | https://github.com/openclaw/openclaw/issues/41703 | https://api.github.com/repos/openclaw/openclaw/issues/41703 | [Feature Request] 中文语言支持 - Chinese Language Support for Dashboard | ## 请求功能
希望 OpenClaw 的 Dashboard (Control UI) 支持中文语言。
## 背景
- Dashboard 目前只有英文界面
- 对于中文用户来说,使用门槛较高
- 中文是全球使用人数最多的语言之一
## 建议
1. 添加中文语言包支持
2. 可以参考 i18n 社区方案
3. 允许用户自定义界面语言
---
## Feature Request
Hope OpenClaw Dashboard (Control UI) supports Chinese language.
## Background
- Dashboard is currently English-only
- ... | closed | completed | true | 1 | [] | [] | 2026-03-10T03:51:49Z | 2026-04-27T04:55:37Z | 2026-04-24T21:20:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Colin-Cai0318 | 90,542,683 | MDQ6VXNlcjkwNTQyNjgz | User | false |
openclaw/openclaw | 4,049,220,249 | I_kwDOQb6kR87xWjKZ | 41,704 | https://github.com/openclaw/openclaw/issues/41704 | https://api.github.com/repos/openclaw/openclaw/issues/41704 | Telegram polling stalls indefinitely when proxy TCP connection drops silently | ## Summary
When `channels.telegram.proxy` is configured and the upstream proxy's TCP connection drops silently (no RST/FIN — e.g. NAT timeout, idle connection eviction), the `getUpdates` long-poll fetch hangs indefinitely. The stall detector fires after ~90 s and forces a polling restart, but this cycle repeats every ... | closed | completed | true | 1 | [] | [] | 2026-03-10T03:52:23Z | 2026-04-27T04:55:36Z | 2026-04-24T21:21:00Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sunyifei83 | 4,455,967 | MDQ6VXNlcjQ0NTU5Njc= | User | false |
openclaw/openclaw | 4,049,226,297 | I_kwDOQb6kR87xWko5 | 41,708 | https://github.com/openclaw/openclaw/issues/41708 | https://api.github.com/repos/openclaw/openclaw/issues/41708 | [Bug]: Telegram fails on Sonnet 4.6, related to thinking blocks | ### Bug type
Regression (worked before, now fails)
### Summary
Bug report summary:
• Working: OpenClaw ≤2026.3.1
• Broken: OpenClaw ≥2026.3.2 (confirmed through 2026.3.8)
• Affects: anthropic/claude-sonnet-4-6 on Telegram only
• Works fine: Opus, OpenRouter models, xAI models
• Error: messages.67.content.1: thinkin... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-10T03:54:41Z | 2026-04-27T04:55:36Z | 2026-04-24T21:21:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | CarsonGreenland | 3,248,403 | MDQ6VXNlcjMyNDg0MDM= | User | false |
openclaw/openclaw | 4,049,236,724 | I_kwDOQb6kR87xWnL0 | 41,715 | https://github.com/openclaw/openclaw/issues/41715 | https://api.github.com/repos/openclaw/openclaw/issues/41715 | [Feature]: Task-Based Model Routing(支持任务自适应的模型选择) | ### Summary
Support Task-Based Model Routing
### Problem to solve
Currently OpenClaw uses a fixed single model and cannot automatically select the appropriate model based on task complexity. This causes:
1. Simple tasks (like greetings, confirmations) also invoke large models, wasting resources and slowing responses... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-10T03:58:42Z | 2026-04-27T04:55:35Z | 2026-04-24T21:23:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | luojing9187-coder | 266,912,434 | U_kgDOD-jCsg | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.