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 | 4,284,070,734 | I_kwDOQb6kR87_WbtO | 68,204 | https://github.com/openclaw/openclaw/issues/68204 | https://api.github.com/repos/openclaw/openclaw/issues/68204 | Unified run trace schema across agent, ACP, subagent, and task flows | **Labels:** `architecture`, `observability`, `tracing`, `P0`
### Summary
Create one canonical run trace schema that works across main agent runs, ACP sessions, subagents, and task flows.
### Why
Tracing exists, but it is fragmented and too low-level. When a run goes wrong, it is hard to quickly answer:
- what happene... | open | null | false | 1 | [] | [] | 2026-04-17T16:45:14Z | 2026-05-03T04:59:37Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | ahmednasr999 | 5,443,403 | MDQ6VXNlcjU0NDM0MDM= | User | false |
openclaw/openclaw | 4,284,070,889 | I_kwDOQb6kR87_Wbvp | 68,205 | https://github.com/openclaw/openclaw/issues/68205 | https://api.github.com/repos/openclaw/openclaw/issues/68205 | Add a core eval harness for memory, tool routing, delegation, and proactive behavior | **Labels:** `quality`, `evals`, `regression`, `P0`
### Summary
Add a small, fixture-driven eval harness to catch core behavior regressions before they hit production use.
### Why
OpenClaw is now powerful enough that regressions are more dangerous than missing features. Changes can silently break:
- memory continuity
... | closed | completed | true | 1 | [] | [] | 2026-04-17T16:45:16Z | 2026-04-28T04:51:35Z | 2026-04-25T18:26:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ahmednasr999 | 5,443,403 | MDQ6VXNlcjU0NDM0MDM= | User | false |
openclaw/openclaw | 4,284,070,972 | I_kwDOQb6kR87_Wbw8 | 68,206 | https://github.com/openclaw/openclaw/issues/68206 | https://api.github.com/repos/openclaw/openclaw/issues/68206 | Introduce typed handoff contracts for expert, worker, and agent-as-tool delegation | **Labels:** `architecture`, `multi-agent`, `delegation`, `P0`
### Summary
Introduce explicit handoff types and required contracts so delegation is predictable, traceable, and easier to evaluate.
### Why
Several delegation patterns are currently mixed together:
- expert delegation
- bounded worker execution
- agent-as... | closed | completed | true | 1 | [] | [] | 2026-04-17T16:45:17Z | 2026-04-27T04:35:22Z | 2026-04-25T03:45:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ahmednasr999 | 5,443,403 | MDQ6VXNlcjU0NDM0MDM= | User | false |
openclaw/openclaw | 4,284,076,304 | I_kwDOQb6kR87_WdEQ | 68,208 | https://github.com/openclaw/openclaw/issues/68208 | https://api.github.com/repos/openclaw/openclaw/issues/68208 | [Bug] Reasoning stream 嵌套循环显示错误 - 思考过程中又触发思考 | ## Bug 描述
在开启 Reasoning Stream 模式时，当模型需要执行多步操作（如查询网页或读取文件）时，会出现嵌套的思考循环。界面显示为递归式的多重思考框，而不是正常的单一思考流程。
## 复现步骤
1. 开启 Reasoning 模式（）
2. 询问需要多步查询的问题，例如 "OpenClaw 最近有哪些大更新？"
3. 当模型尝试使用工具（web_fetch、exec等）获取信息时，思考过程会触发嵌套循环
## 截图
（已附上截图）
## 环境信息
- OpenClaw 版本：当前最新
- 渠道：Feishu
- 模型：MiniMax-M2.7
## 期望行为
思考过程应该正常显示，不应该出现 ... | closed | completed | true | 1 | [] | [] | 2026-04-17T16:46:18Z | 2026-04-28T04:51:34Z | 2026-04-25T18:26:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xuylong | 4,427,070 | MDQ6VXNlcjQ0MjcwNzA= | User | false |
openclaw/openclaw | 4,208,484,853 | I_kwDOQb6kR8762GH1 | 61,502 | https://github.com/openclaw/openclaw/issues/61502 | https://api.github.com/repos/openclaw/openclaw/issues/61502 | Slack interactive button clicks should reliably reach the target session and thread | ## Original Request
> 这些修改会break其他东西吧，暂时不要改了，把刚才学到的东西记录下来，作为一个issue，首先feature reques是让slack可以接收到button。然后在feature reequest description写上我们尝试过的方法和对架构可能提出的影响。现在先去提issue吧。
>
> **English Translation:** These changes may break other things, so don't keep changing code for now. Please record what we learned as an issue. Fir... | open | null | false | 2 | [] | [] | 2026-04-05T20:13:20Z | 2026-05-02T06:31:06Z | null | NONE | null | 20260502T090030Z | 2026-05-02T09:00:30Z | xinbenlv | 640,325 | MDQ6VXNlcjY0MDMyNQ== | User | false |
openclaw/openclaw | 4,284,109,634 | I_kwDOQb6kR87_WlNC | 68,211 | https://github.com/openclaw/openclaw/issues/68211 | https://api.github.com/repos/openclaw/openclaw/issues/68211 | TTS file generated in temp but QQ Bot requires media storage path | ### Bug Description
When using MiniMax TTS provider with QQ Bot channel in OpenClaw v4.15, TTS audio files are generated to the system temp directory (AppData\Local\Temp\openclaw\tts-*\voice-xxx.mp3), but the QQ Bot adapter validates that media files must be inside the QQ Bot media storage directory before sending. Thi... | closed | completed | false | 1 | [] | [
"sliverp"
] | 2026-04-17T16:53:18Z | 2026-04-25T07:56:00Z | 2026-04-25T07:56:00Z | NONE | null | 20260425T090015Z | 2026-04-25T09:00:15Z | lian-koishi | 105,502,146 | U_kgDOBknVwg | User | false |
openclaw/openclaw | 4,284,086,721 | I_kwDOQb6kR87_WfnB | 68,209 | https://github.com/openclaw/openclaw/issues/68209 | https://api.github.com/repos/openclaw/openclaw/issues/68209 | [Bug]: Switching from openai-codex/gpt-5.4 to codex/gpt-5.4 can trigger runaway context growth and off-task workspace contamination | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Switching an OpenClaw agent from `openai-codex/gpt-5.4` to `codex/gpt-5.4` made the session unstable and unusable in my environment.
The stable baseline was:
- `openai-codex/gpt-5.4`
- `plugins.entries.codex.enabled: ... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-04-17T16:48:06Z | 2026-05-02T21:25:43Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | jorgeszabo | 10,103,289 | MDQ6VXNlcjEwMTAzMjg5 | User | false |
openclaw/openclaw | 3,972,081,714 | I_kwDOQb6kR87swSgy | 22,676 | https://github.com/openclaw/openclaw/issues/22676 | https://api.github.com/repos/openclaw/openclaw/issues/22676 | [Bug]: Signal daemon stop() race condition on SIGUSR1 restart — orphaned processes and send failures | ## Summary
During SIGUSR1 gateway restarts (triggered by `config.patch`, `config.apply`, etc.), the signal daemon sends SIGTERM to signal-cli but does not wait for the process to exit. The new signal-cli instance spawns before the old one releases its HTTP port and config file lock, resulting in orphaned processes and... | open | null | false | 15 | [] | [
"shakkernerd"
] | 2026-02-21T13:59:59Z | 2026-05-05T03:10:16Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | UberKitten | 566,438 | MDQ6VXNlcjU2NjQzOA== | User | false |
openclaw/openclaw | 4,284,158,267 | I_kwDOQb6kR87_WxE7 | 68,215 | https://github.com/openclaw/openclaw/issues/68215 | https://api.github.com/repos/openclaw/openclaw/issues/68215 | [Bug]: Update process re-injects anonymousTelemetryId into openclaw-mem0 config, breaking schema validation on v1.0.6+ | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After running openclaw update (gateway update.run), the update process re-injects the anonymousTelemetryId field into the openclaw-mem0 plugin config. Since v1.0.6 added additionalProperties: false to the plugin schema, this... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-17T17:03:48Z | 2026-04-28T04:51:33Z | 2026-04-25T18:26:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rocky-eridian | 264,716,984 | U_kgDOD8dCuA | User | false |
openclaw/openclaw | 4,284,183,329 | I_kwDOQb6kR87_W3Mh | 68,216 | https://github.com/openclaw/openclaw/issues/68216 | https://api.github.com/repos/openclaw/openclaw/issues/68216 | [Bug]: Openclaw with Gemini CLI model provider fails to write to Identity.md and other workspace files. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Using OpenClaw with `google-gemini-cli` via OAuth on Linux, the agent works with injected workspace files but cannot write/edit files like `IDENTITY.md`.
Using OpenClaw with `google-gemini-cli` via OAuth on Li... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-04-17T17:09:25Z | 2026-05-02T21:50:03Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | vshanbha | 14,871,482 | MDQ6VXNlcjE0ODcxNDgy | User | false |
openclaw/openclaw | 4,284,120,255 | I_kwDOQb6kR87_Wny_ | 68,214 | https://github.com/openclaw/openclaw/issues/68214 | https://api.github.com/repos/openclaw/openclaw/issues/68214 | [feature] Add on-unauthorized hook event and suppress-pairing-reply config flag | # [feature] Add on-unauthorized hook event and suppress-pairing-reply config flag
## Context
OpenClaw currently replies to every unauthorized inbound DM (not on `allowFrom`) with a pairing-code message that includes the bot owner's phone number and a ready-to-paste approval command. In `dmPolicy: "pairing"` mode this... | open | null | false | 3 | [] | [] | 2026-04-17T16:55:32Z | 2026-05-02T14:22:10Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | Trupe-Rs | 214,037,382 | U_kgDODMHzhg | User | false |
openclaw/openclaw | 4,163,940,727 | I_kwDOQb6kR874MLF3 | 56,982 | https://github.com/openclaw/openclaw/issues/56982 | https://api.github.com/repos/openclaw/openclaw/issues/56982 | [Bug]: openclaw doctor reporting gateway error when trusted proxy auth mode is enabled | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When you run `openclaw doctor`, if you have trusted proxy auth mode configured for your OpenClaw, you will see the following message for the doctor's evaluation of the gateway:
```
◇ Gateway auth ────────────... | closed | completed | true | 4 | [
"bug"
] | [] | 2026-03-29T11:46:39Z | 2026-04-28T04:30:32Z | 2026-04-25T11:54:26Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nickytonline | 833,231 | MDQ6VXNlcjgzMzIzMQ== | User | false |
openclaw/openclaw | 4,284,212,940 | I_kwDOQb6kR87_W-bM | 68,218 | https://github.com/openclaw/openclaw/issues/68218 | https://api.github.com/repos/openclaw/openclaw/issues/68218 | TTS audio base64 stored inline in session history causes immediate context overflow | TTS audio base64 stored inline in session history causes immediate context overflow
## Environment
- OpenClaw: v2026.4.15
- Model: zai/glm-5.1 (203K context window)
- TTS Provider: local-tts (custom plugin, Qwen3-TTS)
- Channel: Feishu
- OS: macOS (ARM64)
## Description
When TTS is enabled (`messages.tts.enabled: t... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:14:46Z | 2026-04-28T04:51:33Z | 2026-04-25T18:26:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | lhtpluto | 132,045,861 | U_kgDOB97cJQ | User | false |
openclaw/openclaw | 4,284,250,714 | I_kwDOQb6kR87_XHpa | 68,220 | https://github.com/openclaw/openclaw/issues/68220 | https://api.github.com/repos/openclaw/openclaw/issues/68220 | Feature request: allowBots filter for Telegram channel (mirrors Discord behavior) | ## Summary
The Discord and Slack channel configs support `allowBots: false` to prevent bot-authored messages from triggering agent activation. The Telegram channel has no equivalent filter.
## Problem
In a multi-agent setup using per-topic routing, when a cron job posts output to a Telegram group topic, the same age... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:21:32Z | 2026-04-28T04:51:32Z | 2026-04-25T18:26:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Winston9520 | 143,687,718 | U_kgDOCJCAJg | User | false |
openclaw/openclaw | 4,284,277,628 | I_kwDOQb6kR87_XON8 | 68,221 | https://github.com/openclaw/openclaw/issues/68221 | https://api.github.com/repos/openclaw/openclaw/issues/68221 | Cron job manual execution fails with TypeError: Cannot read properties of undefined (reading 'startsWith') | ## OpenClaw Cron Job Issue Report
### Issue Title
Cron job manual execution fails with `TypeError: Cannot read properties of undefined (reading 'startsWith')`
### Issue Body
**Describe the bug**
When manually executing a cron job from the OpenClaw Control UI, the job fails with the error:
```
TypeError: Cannot read ... | closed | not_planned | true | 2 | [
"close:invalid"
] | [
"vincentkoc"
] | 2026-04-17T17:27:06Z | 2026-04-28T05:28:09Z | 2026-04-25T09:43:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | easonfu23 | 196,026,231 | U_kgDOC68fdw | User | false |
openclaw/openclaw | 4,284,279,096 | I_kwDOQb6kR87_XOk4 | 68,222 | https://github.com/openclaw/openclaw/issues/68222 | https://api.github.com/repos/openclaw/openclaw/issues/68222 | [Bug]: Kimi Code model frequent sessions_yield tool call/output interrupts long-running tasks, requires manual intervention to resume | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
The Kimi Code model in OpenClaw frequently generates tool call: sessions_yield and tool output: sessions_yield during long-running task conversations, which interrupts task and dialogue flow and requires manual input to resume ... | open | null | false | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-17T17:27:24Z | 2026-05-03T03:37:19Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | huangjk1103 | 32,481,141 | MDQ6VXNlcjMyNDgxMTQx | User | false |
openclaw/openclaw | 4,284,111,267 | I_kwDOQb6kR87_Wlmj | 68,212 | https://github.com/openclaw/openclaw/issues/68212 | https://api.github.com/repos/openclaw/openclaw/issues/68212 | [Bug]: Google Gemini TTS voiceName changes do not affect output voice in Discord TTS replies | ## Summary
When using the new Google Gemini TTS provider in OpenClaw `v2026.4.15`, changing `messages.tts.providers.google.voiceName` does not appear to change the actual output voice in Discord-delivered TTS replies.
I tested multiple documented Gemini voices, including `Kore` and `Puck`, with gateway restarts betwee... | closed | completed | true | 1 | [] | [] | 2026-04-17T16:53:40Z | 2026-04-27T04:32:48Z | 2026-04-25T04:22:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mogglemoss | 10,966,195 | MDQ6VXNlcjEwOTY2MTk1 | User | false |
openclaw/openclaw | 4,284,303,666 | I_kwDOQb6kR87_XUky | 68,223 | https://github.com/openclaw/openclaw/issues/68223 | https://api.github.com/repos/openclaw/openclaw/issues/68223 | Multi Telegram Bot WS Handshake Timeout Cascade on Gateway Start | ## Summary
When running 4 Telegram bot agents, gateway restart triggers a cascade of WebSocket handshake timeouts. All bots fail to connect on the first attempt, taking 3-5 minutes before auto-recovery succeeds.
## Environment
- **OpenClaw**: 2026.4.15 (041266a)
- **OS**: Windows 10 (10.0.19045, x64)
- **Node**: v24... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:32:26Z | 2026-04-28T04:51:31Z | 2026-04-25T18:26:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Snufkin200 | 276,913,742 | U_kgDOEIFeTg | User | false |
openclaw/openclaw | 4,090,232,568 | I_kwDOQb6kR87zy_74 | 49,178 | https://github.com/openclaw/openclaw/issues/49178 | https://api.github.com/repos/openclaw/openclaw/issues/49178 | [Feature]: Reusable gateway WebSocket client SDK package | ### Summary
Extract a universal, platform-agnostic WebSocket client SDK (`@openclaw/gateway-client`) for communicating with the OpenClaw gateway.
### Problem to solve
Both the CLI (`src/gateway/client.ts`) and control-ui (`ui/src/ui/gateway.ts`) implement the gateway WebSocket protocol independently — duplicate hand... | open | null | false | 5 | [] | [] | 2026-03-17T18:01:10Z | 2026-04-30T21:58:34Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | DeJeune | 67,425,183 | MDQ6VXNlcjY3NDI1MTgz | User | false |
openclaw/openclaw | 4,124,187,908 | I_kwDOQb6kR8710h0E | 53,214 | https://github.com/openclaw/openclaw/issues/53214 | https://api.github.com/repos/openclaw/openclaw/issues/53214 | TUI: toggle visibility of tool calls (ctrl+o equivalent) | When the agent makes tool calls (edit, exec, read, gateway, etc.), they execute silently — the operator only sees text responses. There is no way to see what the agent is doing behind the scenes.
Request: a toggle (like Claude Code's ctrl+o) to show/hide tool call summaries in the TUI. Default collapsed, expandable on... | closed | completed | true | 2 | [] | [] | 2026-03-23T23:00:16Z | 2026-04-30T04:46:06Z | 2026-04-25T06:16:12Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mividtim | 15,264,802 | MDQ6VXNlcjE1MjY0ODAy | User | false |
openclaw/openclaw | 4,284,340,824 | I_kwDOQb6kR87_XdpY | 68,226 | https://github.com/openclaw/openclaw/issues/68226 | https://api.github.com/repos/openclaw/openclaw/issues/68226 | Feature request: Add session duration (elapsed time) to status footer | ## Summary
Add the current session's elapsed time (duration) to the `🧵 Session:` line in the status footer.
## Current Behaviour
```
🧵 Session: agent:fraser:slack:channel:c0asptb5dmj • updated 1m ago
```
The status footer shows when the session was last updated but not how long the session has been active.
## Re... | open | null | false | 3 | [] | [] | 2026-04-17T17:40:53Z | 2026-05-02T23:12:39Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | marshall-gordfam | 271,781,056 | U_kgDOEDMMwA | User | false |
openclaw/openclaw | 4,284,382,147 | I_kwDOQb6kR87_XnvD | 68,228 | https://github.com/openclaw/openclaw/issues/68228 | https://api.github.com/repos/openclaw/openclaw/issues/68228 | Image paste in webchat silently dropped when a custom-provider model (e.g. claude-opus-4-7 via models.json) is active | **OpenClaw version:** 2026.4.15
**Channel:** webchat (control-ui)
**Provider:** anthropic (direct, via agent-level `models.json`)
**Model:** `anthropic/claude-opus-4-7`
## Summary
When a user pastes an image into the webchat compose area, the Control UI correctly captures the image and submits it as an `attachments` ... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:48:49Z | 2026-04-28T04:51:30Z | 2026-04-25T18:27:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | amarsmission | 31,492,306 | MDQ6VXNlcjMxNDkyMzA2 | User | false |
openclaw/openclaw | 4,284,400,069 | I_kwDOQb6kR87_XsHF | 68,229 | https://github.com/openclaw/openclaw/issues/68229 | https://api.github.com/repos/openclaw/openclaw/issues/68229 | [Bug] Session file replaced on cache-ttl expiry — all conversation history lost | ## Priority: Critical
This is the primary cause of complete context loss in group channels.
## Version
OpenClaw 2026.4.2 (d74a122)
## Description
When a session's cache-ttl expires (or a gateway restart invalidates the cache), OpenClaw creates a **new session file** rather than continuing from the existing one. The p... | closed | completed | true | 2 | [] | [] | 2026-04-17T17:52:31Z | 2026-04-27T04:40:47Z | 2026-04-25T02:21:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | chrismclarnon-ep | 275,793,450 | U_kgDOEHBGKg | User | false |
openclaw/openclaw | 4,284,400,386 | I_kwDOQb6kR87_XsMC | 68,230 | https://github.com/openclaw/openclaw/issues/68230 | https://api.github.com/repos/openclaw/openclaw/issues/68230 | [Feature] Add persistent context pruning mode that survives gateway restarts | ## Summary
`contextPruning.mode` only supports `"off"` and `"cache-ttl"`. `cache-ttl` relies on Anthropic's server-side prompt caching, which is invalidated on every gateway restart. There is no disk-backed persistent mode.
## Version
OpenClaw 2026.4.2 (d74a122)
## Problem
With `cache-ttl`, the effective TTL is deter... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:52:34Z | 2026-04-30T04:39:26Z | 2026-04-26T05:19:13Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chrismclarnon-ep | 275,793,450 | U_kgDOEHBGKg | User | false |
openclaw/openclaw | 4,284,400,567 | I_kwDOQb6kR87_XsO3 | 68,231 | https://github.com/openclaw/openclaw/issues/68231 | https://api.github.com/repos/openclaw/openclaw/issues/68231 | [Feature] Add per-group agentId/model override for Telegram channels | ## Summary
The Telegram channel schema supports `agentId` at the **topic** level (`channels.telegram.groups.<id>.topics.<id>.agentId`) but NOT at the **group** level. This means different groups cannot use different models without converting groups to forum mode and using topics.
## Version
OpenClaw 2026.4.2 (d74a122)... | closed | completed | true | 2 | [] | [] | 2026-04-17T17:52:36Z | 2026-04-28T04:51:30Z | 2026-04-25T18:27:09Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chrismclarnon-ep | 275,793,450 | U_kgDOEHBGKg | User | false |
openclaw/openclaw | 4,284,400,750 | I_kwDOQb6kR87_XsRu | 68,232 | https://github.com/openclaw/openclaw/issues/68232 | https://api.github.com/repos/openclaw/openclaw/issues/68232 | [Bug] Config edits trigger Telegram channel restart via hot reload — destroys in-memory context | ## Summary
Every edit to `openclaw.json` triggers a hot-reload that restarts the Telegram channel, destroying all in-memory state including `cache-ttl` context and the `groupHistories` message buffer.
## Version
OpenClaw 2026.4.2 (d74a122)
## Steps to Reproduce
1. Have an active conversation in a Telegram group chann... | open | null | false | 2 | [] | [
"vincentkoc"
] | 2026-04-17T17:52:39Z | 2026-05-02T10:36:50Z | null | NONE | null | 20260502T120029Z | 2026-05-02T12:00:29Z | chrismclarnon-ep | 275,793,450 | U_kgDOEHBGKg | User | false |
openclaw/openclaw | 4,284,400,931 | I_kwDOQb6kR87_XsUj | 68,233 | https://github.com/openclaw/openclaw/issues/68233 | https://api.github.com/repos/openclaw/openclaw/issues/68233 | [Bug] bootstrapMaxChars silently truncates workspace files with no user-facing warning | ## Summary
When a workspace file (e.g., SOUL.md) exceeds `bootstrapMaxChars`, it is silently truncated during session bootstrap. No warning is sent to the channel, no log alert is raised, and the user has no way to know that the agent is operating with incomplete instructions.
## Version
OpenClaw 2026.4.2 (d74a122)
#... | closed | completed | true | 1 | [] | [] | 2026-04-17T17:52:41Z | 2026-04-28T04:51:29Z | 2026-04-25T18:27:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chrismclarnon-ep | 275,793,450 | U_kgDOEHBGKg | User | false |
openclaw/openclaw | 4,165,603,646 | I_kwDOQb6kR874ShE- | 57,256 | https://github.com/openclaw/openclaw/issues/57256 | https://api.github.com/repos/openclaw/openclaw/issues/57256 | [Bug]: openclaw status falsely reports openclaw-mem0 memory as unavailable when the live gateway plugin is working | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
`openclaw status --deep` reports:
- `Memory enabled (plugin openclaw-mem0) · unavailable`
However, the live gateway process shows that `openclaw-mem0` is functioning:
- the plugin registers
- the plugin ini... | open | null | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-29T19:30:52Z | 2026-05-03T08:24:37Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | leofaoro | 1,944,817 | MDQ6VXNlcjE5NDQ4MTc= | User | false |
openclaw/openclaw | 4,284,488,036 | I_kwDOQb6kR87_YBlk | 68,237 | https://github.com/openclaw/openclaw/issues/68237 | https://api.github.com/repos/openclaw/openclaw/issues/68237 | [Bug]: SecretRef resolution regression in 2026.4.12+ (socket-mode slack reply fails, filemain-backed plugin secrets read as "unresolved in the active runtime snapshot") | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
On socket-mode Slack with a SecretRef-backed `botToken`, inbound messages are received fine but every final reply throws `channels.slack.accounts.<id>.botToken: unresolved SecretRef … Resolve this command against an active g... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-04-17T18:11:14Z | 2026-04-23T04:22:39Z | 2026-04-20T06:59:10Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | matthewfranglen | 122,936 | MDQ6VXNlcjEyMjkzNg== | User | false |
openclaw/openclaw | 4,054,386,198 | I_kwDOQb6kR87xqQYW | 42,515 | https://github.com/openclaw/openclaw/issues/42515 | https://api.github.com/repos/openclaw/openclaw/issues/42515 | [Bug]: Provider 400 error causes session deadlock — no user notification, no auto-recovery | ## Summary
When a provider returns a **400 error** (e.g. Moonshot API `400 The request was rejected because it was considered high risk`), OpenClaw records an assistant message with `stopReason: "error"` and **empty content**, but sends **no message to the chat**. The user sees nothing.
## Steps to Reproduce
1. Use ... | closed | completed | true | 2 | [] | [] | 2026-03-10T21:44:57Z | 2026-04-28T04:41:43Z | 2026-04-26T02:51:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mvalentsev | 2,305,718 | MDQ6VXNlcjIzMDU3MTg= | User | false |
openclaw/openclaw | 4,284,534,534 | I_kwDOQb6kR87_YM8G | 68,239 | https://github.com/openclaw/openclaw/issues/68239 | https://api.github.com/repos/openclaw/openclaw/issues/68239 | Duplicate message accumulation in webchat session (MiniMax-M2.7) | ## Bug Description
In a webchat session (model: MiniMax-M2.7), assistant replies are accumulating — each new message prepends/contains the full text of all previous messages in the session, creating exponentially growing duplicates.
## Steps to Reproduce
1. Use webchat (Control UI) with MiniMax-M2.7 model
2. Have a m... | closed | completed | true | 1 | [] | [] | 2026-04-17T18:21:04Z | 2026-04-28T04:51:28Z | 2026-04-25T18:27:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KarbonXX | 8,398,417 | MDQ6VXNlcjgzOTg0MTc= | User | false |
openclaw/openclaw | 4,284,557,170 | I_kwDOQb6kR87_YSdy | 68,241 | https://github.com/openclaw/openclaw/issues/68241 | https://api.github.com/repos/openclaw/openclaw/issues/68241 | [Bug]: claude-opus-4-7 rejected as "Unknown model" on 2026.4.15 despite being in codebase and accepted by Anthropic API | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from OpenClaw 2026.4.2 to 2026.4.15, configuring `anthropic/claude-opus-4-7` as agent primary model causes `FailoverError: Unknown model` on every request. The model falls back to claude-opus-4-6. The model s... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-04-17T18:25:10Z | 2026-04-28T04:51:27Z | 2026-04-25T18:27:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | alfredneuworth | 277,086,943 | U_kgDOEIQC3w | User | false |
openclaw/openclaw | 4,284,676,850 | I_kwDOQb6kR87_Yvry | 68,245 | https://github.com/openclaw/openclaw/issues/68245 | https://api.github.com/repos/openclaw/openclaw/issues/68245 | [Bug]: Active Memory blocks reply flow and hangs until embedded-run timeout, provider-independent | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
Active Memory consistently blocks the normal reply path in my setup. The plugin starts correctly, but after active-memory ... start, the chat reply never returns. About 2 minutes later, typing TTL reached appears, followed by e... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-17T18:48:55Z | 2026-04-28T04:51:27Z | 2026-04-25T18:27:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sirrex777 | 69,386,002 | MDQ6VXNlcjY5Mzg2MDAy | User | false |
openclaw/openclaw | 4,138,300,896 | I_kwDOQb6kR872qXXg | 54,689 | https://github.com/openclaw/openclaw/issues/54689 | https://api.github.com/repos/openclaw/openclaw/issues/54689 | [Bug]: Telegram file sending broken after 2026.3.24 update | ### Bug type
Regression (worked before, now fails)
### Summary
**Summary**
Prior to the 2026.3.24 update, the agent was able to send files (PDFs, documents) through the Telegram channel. After the update, this capability appears to have been removed or broken.
**Environment**
- OS: macOS arm64
- OpenClaw version: 2... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-25T20:26:59Z | 2026-04-28T05:25:52Z | 2026-04-25T10:10:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Carlsteamer | 271,066,126 | U_kgDOECgkDg | User | false |
openclaw/openclaw | 4,284,740,812 | I_kwDOQb6kR87_Y_TM | 68,246 | https://github.com/openclaw/openclaw/issues/68246 | https://api.github.com/repos/openclaw/openclaw/issues/68246 | [OpenClaw][Tools] No end-user path to give the 'main' agent access to custom tools — skills are prompt recipes, MCP registry only wires to ACP, plugins are undocumented |
### Summary
A NemoClaw user running the default `main` agent (the always-on assistant that ships with `nemoclaw onboard`) has no supported end-user path to extend the agent's tool set with custom tools. Skills load as prompt guidance only, MCP server registry only wires to ACP (external coding runtimes), and plugins ... | closed | completed | true | 3 | [] | [] | 2026-04-17T19:03:17Z | 2026-04-28T04:51:26Z | 2026-04-25T18:27:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | davidglogan | 31,814,247 | MDQ6VXNlcjMxODE0MjQ3 | User | false |
openclaw/openclaw | 4,078,939,300 | I_kwDOQb6kR87zH6yk | 47,576 | https://github.com/openclaw/openclaw/issues/47576 | https://api.github.com/repos/openclaw/openclaw/issues/47576 | fix(failover): bare leading 402 assistant errors should still enter model fallback | ## Summary
A ZenMux 402 can still bypass model fallback when it reaches the embedded runner as a bare assistant error string like:
```text
402 You have reached your subscription quota limit. Please wait for automatic quota refresh in the rolling time window, upgrade to a higher plan, or use a Pay-As-You-Go API Key fo... | closed | completed | true | 1 | [] | [] | 2026-03-15T18:35:15Z | 2026-04-20T04:26:31Z | 2026-04-17T19:06:57Z | CONTRIBUTOR | null | 20260420T060038Z | 2026-04-20T06:00:38Z | bwjoke | 1,284,814 | MDQ6VXNlcjEyODQ4MTQ= | User | false |
openclaw/openclaw | 4,284,760,329 | I_kwDOQb6kR87_ZEEJ | 68,248 | https://github.com/openclaw/openclaw/issues/68248 | https://api.github.com/repos/openclaw/openclaw/issues/68248 | [Feature Request] Add avatar/favicon customization for Control UI agents | ## Feature Request: Agent Avatar/Favicon Customization in Control UI
### Problem
Currently the OpenClaw Control UI shows a generic star placeholder for all agents/sessions. Users running multiple agents (e.g., Max the main agent, Minion subagents, Hermes, etc.) have no visual way to distinguish which agent they are ta... | open | null | false | 1 | [] | [] | 2026-04-17T19:07:12Z | 2026-05-02T21:25:44Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | KarbonXX | 8,398,417 | MDQ6VXNlcjgzOTg0MTc= | User | false |
openclaw/openclaw | 4,284,771,753 | I_kwDOQb6kR87_ZG2p | 68,249 | https://github.com/openclaw/openclaw/issues/68249 | https://api.github.com/repos/openclaw/openclaw/issues/68249 | [Bug]: Brave plugin enabled in config but not loaded at runtime — web_search tool unavailable | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Brave plugin enabled in `openclaw.json` (`plugins.entries.brave.enabled: true`) but does not appear in the gateway's active plugin list at boot. `web_search` tool is registered in the tool policy but marked "unavailable in t... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-17T19:09:55Z | 2026-04-30T04:39:25Z | 2026-04-26T05:19:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alfredneuworth | 277,086,943 | U_kgDOEIQC3w | User | false |
openclaw/openclaw | 4,284,828,862 | I_kwDOQb6kR87_ZUy- | 68,251 | https://github.com/openclaw/openclaw/issues/68251 | https://api.github.com/repos/openclaw/openclaw/issues/68251 | Memory search embed timeout — model not found for embeddings | ## Description
The memorySearch embed function times out because OpenClaw tries to use an incorrect/inexistent model for vector embeddings.
## Environment
- OpenClaw version: 2026.4.15
- Platform: Docker container on Hostinger VPS
- Model provider: Gemini (gemini-embedding-001)
## What happens
1. Gateway starts norma... | closed | completed | true | 1 | [] | [] | 2026-04-17T19:21:50Z | 2026-04-28T04:51:25Z | 2026-04-25T18:28:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | d-danielsun | 209,839,705 | U_kgDODIHmWQ | User | false |
openclaw/openclaw | 4,284,837,574 | I_kwDOQb6kR87_ZW7G | 68,252 | https://github.com/openclaw/openclaw/issues/68252 | https://api.github.com/repos/openclaw/openclaw/issues/68252 | memory-core: dreaming narrative session cleanup fails with "missing scope: operator.admin" | ## Summary
In OpenClaw 2026.4.14, every dreaming narrative phase (light + REM) fails to clean up its subagent session with `missing scope: operator.admin`. The narrative itself is written to `DREAMS.md` before the cleanup call runs, so user-visible dreaming output is unaffected — but subagent sessions likely accumulat... | closed | duplicate | true | 3 | [] | [] | 2026-04-17T19:23:24Z | 2026-04-30T04:35:20Z | 2026-04-27T09:41:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | omerekin | 36,278,311 | MDQ6VXNlcjM2Mjc4MzEx | User | false |
openclaw/openclaw | 4,284,874,890 | I_kwDOQb6kR87_ZgCK | 68,254 | https://github.com/openclaw/openclaw/issues/68254 | https://api.github.com/repos/openclaw/openclaw/issues/68254 | [Bug]: Internal "Telegram Native Approvals" handler cannot be paired in Docker loopback-only deployment (2026.4.14) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading OpenClaw from **2026.3.28 → 2026.4.14** in a Docker loopback-only deployment, the internal **"Telegram Native Approvals"** handler creates a pending device pairing request with flag `repair` every ~1 second, ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-17T19:30:43Z | 2026-04-28T04:51:24Z | 2026-04-25T18:28:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ex3del | 81,170,265 | MDQ6VXNlcjgxMTcwMjY1 | User | false |
openclaw/openclaw | 4,284,884,897 | I_kwDOQb6kR87_Zieh | 68,256 | https://github.com/openclaw/openclaw/issues/68256 | https://api.github.com/repos/openclaw/openclaw/issues/68256 | Non-default agents with per-agent auth-profiles.json return empty responses via OpenRouter provider | ## Bug Report
### Summary
Non-default agents configured to use `openrouter/*` models return empty responses (`stopReason=stop`, `payloads=0`, `totalTokens=0`) when they have their own `auth-profiles.json` with a valid OpenRouter API key. The same API key works when tested directly via `curl`. Switching the agent to `a... | closed | completed | true | 1 | [] | [] | 2026-04-17T19:32:34Z | 2026-04-30T04:39:25Z | 2026-04-26T05:19:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JaiHealth | 264,775,728 | U_kgDOD8goMA | User | false |
openclaw/openclaw | 4,284,891,511 | I_kwDOQb6kR87_ZkF3 | 68,258 | https://github.com/openclaw/openclaw/issues/68258 | https://api.github.com/repos/openclaw/openclaw/issues/68258 | [Bug]: Gateway self-SIGTERMs on OAuth token refresh (auth.profiles.* unmatched in reload plan) | ## Bug type
Crash (process exits mid-dispatch — graceful SIGTERM + launchd respawn, but all in-flight WebSocket work is dropped).
## Beta release blocker
No.
## Summary
`buildGatewayReloadPlan()` restarts the gateway whenever an OAuth provider writes a token refresh to the persisted config (`auth.profiles.<provide... | open | null | false | 2 | [] | [] | 2026-04-17T19:33:55Z | 2026-05-03T03:22:57Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | mkoslacz | 7,924,595 | MDQ6VXNlcjc5MjQ1OTU= | User | false |
openclaw/openclaw | 4,284,934,036 | I_kwDOQb6kR87_ZueU | 68,259 | https://github.com/openclaw/openclaw/issues/68259 | https://api.github.com/repos/openclaw/openclaw/issues/68259 | Add `openclaw sessions send` CLI command to expose session messaging | ## Problem
- `chain-runner` assumes `openclaw sessions send` exists
- It does not
- Current dispatch silently no-ops
- `team-*` chains therefore fail to wake downstream agents unless manually nudged via MCP `sessions_send`
The Discord self-post fallback some chains use isn't a real dispatch mechanism — it masks the u... | closed | completed | true | 1 | [] | [] | 2026-04-17T19:42:20Z | 2026-05-01T04:44:45Z | 2026-04-28T07:12:42Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | DashLabsDev | 263,981,165 | U_kgDOD7wIbQ | User | false |
openclaw/openclaw | 4,284,968,631 | I_kwDOQb6kR87_Z263 | 68,260 | https://github.com/openclaw/openclaw/issues/68260 | https://api.github.com/repos/openclaw/openclaw/issues/68260 | Bug: ollama plugin configSchema forbids baseUrl but provider code still reads it | # Bug: ollama plugin `configSchema` forbids `baseUrl`, but provider code still reads it
## Environment
- **OpenClaw**: 2026.4.14 (`323493f`)
- **Node**: v22.22.2
- **OS**: macOS 15 (Darwin 25.2.0), Apple Silicon
- **Plugin**: bundled `@openclaw/ollama-provider` (`dist/extensions/ollama`)
## Summary
The bundled Ollam... | closed | completed | true | 1 | [] | [] | 2026-04-17T19:49:23Z | 2026-04-28T04:51:24Z | 2026-04-25T18:28:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | exscape86-lang | 276,645,983 | U_kgDOEH1IXw | User | false |
openclaw/openclaw | 4,284,971,952 | I_kwDOQb6kR87_Z3uw | 68,261 | https://github.com/openclaw/openclaw/issues/68261 | https://api.github.com/repos/openclaw/openclaw/issues/68261 | [Bug]: OpenRouter model runs return blank responses / stopReason=stop payloads=0 on 2026.4.15 Summary | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
On OpenClaw 2026.4.15 (041266a) on macOS, chat responses in the dashboard come back blank. The same failure reproduces outside the dashboard with a direct CLI inference command, so this does not appear to be a stuck session ... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-17T19:50:03Z | 2026-04-28T04:51:23Z | 2026-04-25T18:28:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ericbreitnicholson | 181,784,480 | U_kgDOCtXPoA | User | false |
openclaw/openclaw | 4,284,977,092 | I_kwDOQb6kR87_Z4_E | 68,263 | https://github.com/openclaw/openclaw/issues/68263 | https://api.github.com/repos/openclaw/openclaw/issues/68263 | Possible complementary direction: post-task skill lifecycle management with SkillClaw | Hi, I'm reaching out because `openclaw` sits close to the core OpenClaw workflow.
Project:
https://github.com/AMAP-ML/SkillClaw
This is not meant as a generic wrapper pitch. SkillClaw focuses on a narrower ecosystem gap: after repeated usage, long-lived skill libraries often become noisy, duplicated, stale, and fragm... | closed | not_planned | true | 1 | [] | [] | 2026-04-17T19:51:08Z | 2026-04-28T04:51:22Z | 2026-04-25T18:28:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Upper9527 | 76,113,672 | MDQ6VXNlcjc2MTEzNjcy | User | false |
openclaw/openclaw | 4,285,012,903 | I_kwDOQb6kR87_aBun | 68,264 | https://github.com/openclaw/openclaw/issues/68264 | https://api.github.com/repos/openclaw/openclaw/issues/68264 | [Bug] Canvas/Browser UI Visualization Fails to Render in Chat Environment | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When using the canvas tool for browsing or focusing on the browser view, the visualization of the actual web page content is not reliably visible within the chat interface. The system returns technical errors like "node requ... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-04-17T19:58:42Z | 2026-05-02T22:58:53Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | Skunezero | 200,779,190 | U_kgDOC_eltg | User | false |
openclaw/openclaw | 4,285,027,602 | I_kwDOQb6kR87_aFUS | 68,265 | https://github.com/openclaw/openclaw/issues/68265 | https://api.github.com/repos/openclaw/openclaw/issues/68265 | [NemoClaw][All Platform][Agent] sessions_spawn sub-agents time out at exactly 1m0s with no output and no user-visible recovery |
### Description
When the parent agent spawns a sub-agent via the `sessions_spawn` tool and the sub-agent does not produce a result in time, the agent framework emits an internal completion event with `status: timed out` and `<<<BEGIN_UNTRUSTED_CHILD_RESULT>>>(no output)<<<END_UNTRUSTED_CHILD_RESULT>>>`. The parent th... | closed | completed | true | 1 | [] | [] | 2026-04-17T20:01:42Z | 2026-04-28T04:51:21Z | 2026-04-25T18:28:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | davidglogan | 31,814,247 | MDQ6VXNlcjMxODE0MjQ3 | User | false |
openclaw/openclaw | 4,205,103,274 | I_kwDOQb6kR876pMiq | 60,913 | https://github.com/openclaw/openclaw/issues/60913 | https://api.github.com/repos/openclaw/openclaw/issues/60913 | Support Xiaomi mimo-v2-omni for image understanding (media-understanding provider) | ## Problem
When setting `agents.defaults.imageModel` to `xiaomi/mimo-v2-omni`, the image tool fails with:
- `Error: Model does not support images`
- Or: `No media-understanding provider registered for xiaomi`
Even though `mimo-v2-omni` is configured with `input: ["text", "image"]` in the model catalog.
## Root Cause... | closed | completed | true | 2 | [] | [] | 2026-04-04T14:51:08Z | 2026-04-28T05:10:12Z | 2026-04-25T13:48:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | AteneaCRodriguez | 258,796,565 | U_kgDOD2zsFQ | User | false |
openclaw/openclaw | 4,285,052,792 | I_kwDOQb6kR87_aLd4 | 68,268 | https://github.com/openclaw/openclaw/issues/68268 | https://api.github.com/repos/openclaw/openclaw/issues/68268 | [Feature]: Friendly Telegram topic names in session list | ### Summary
When using Telegram forum topics, the session dropdown displays raw keys such as agent:main:telegram:group:-123:topic:10 rather than meaningful, human-readable names.
### Problem to solve
When multiple personas share a channel, agents have no way to identify which persona context applies, leading to resp... | open | null | false | 0 | [
"enhancement"
] | [] | 2026-04-17T20:07:04Z | 2026-04-17T20:07:04Z | null | NONE | null | 20260417T202042Z | 2026-04-17T20:20:42Z | lemonheaded | 218,288,988 | U_kgDODQLTXA | User | false |
openclaw/openclaw | 4,285,149,965 | I_kwDOQb6kR87_ajMN | 68,271 | https://github.com/openclaw/openclaw/issues/68271 | https://api.github.com/repos/openclaw/openclaw/issues/68271 | feature: per-agent write-tool path allow/deny list | # Draft: Upstream OpenClaw feature request — per-agent write-tool path policy
**Target repo:** `github.com/openclaw/openclaw`
**Proposed title:** `[feature] per-agent write-tool path allow/deny list`
---
## Motivation
We've observed a class of self-corruption bugs where an agent's `write` tool
clobbers its own oper... | closed | not_planned | true | 1 | [] | [] | 2026-04-17T20:26:52Z | 2026-04-28T04:51:21Z | 2026-04-25T18:28:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rdleclerc | 262,463,626 | U_kgDOD6Tgig | User | false |
openclaw/openclaw | 4,285,160,558 | I_kwDOQb6kR87_alxu | 68,272 | https://github.com/openclaw/openclaw/issues/68272 | https://api.github.com/repos/openclaw/openclaw/issues/68272 | [Bug]: Image attachments dropped even when model supports images Description | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Bug: Image attachments dropped even when model supports images
Environment: OpenClaw 2026.4.14 / 2026.4.15, minimax/MiniMax-M2.7
Problem: When sending images via webchat, the image appears as a thumbnail momentarily then d... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-17T20:28:49Z | 2026-04-28T04:51:20Z | 2026-04-25T18:28:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | openClawcyh | 269,630,542 | U_kgDOEBI8Tg | User | false |
openclaw/openclaw | 4,285,167,751 | I_kwDOQb6kR87_aniH | 68,273 | https://github.com/openclaw/openclaw/issues/68273 | https://api.github.com/repos/openclaw/openclaw/issues/68273 | [OpenClaw][CLI] openclaw agent hides the assistant reply on stdout unless --json is passed |
Running `openclaw agent --agent main -m "<prompt>"` without the `--json` flag prints only `completed` plus a short status banner. The assistant reply text is not written to stdout; it is only retrievable via `--json` and payload traversal.
This is a surprising default for a CLI whose primary purpose is to produce a r... | closed | completed | true | 1 | [] | [] | 2026-04-17T20:30:16Z | 2026-04-28T04:51:19Z | 2026-04-25T18:29:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | davidglogan | 31,814,247 | MDQ6VXNlcjMxODE0MjQ3 | User | false |
openclaw/openclaw | 4,285,244,312 | I_kwDOQb6kR87_a6OY | 68,275 | https://github.com/openclaw/openclaw/issues/68275 | https://api.github.com/repos/openclaw/openclaw/issues/68275 | sessions_spawn auto-injects `streamTo:"parent"` for `runtime="subagent"` and causes hard failure | ## Version
OpenClaw `2026.4.11`
## Summary
When calling `sessions_spawn` with `runtime:"subagent"` and omitting `streamTo`, the transmitted request still contains `streamTo:"parent"`, which triggers validation failure:
`streamTo is only supported for runtime=acp; got runtime=subagent`
This makes subagent spawns fail... | closed | not_planned | true | 2 | [] | [] | 2026-04-17T20:46:44Z | 2026-04-30T04:39:24Z | 2026-04-26T05:19:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | vladzol-dyk | 270,167,415 | U_kgDOEBptdw | User | false |
openclaw/openclaw | 4,190,033,575 | I_kwDOQb6kR875vtan | 59,257 | https://github.com/openclaw/openclaw/issues/59257 | https://api.github.com/repos/openclaw/openclaw/issues/59257 | [Bug]: Isolated cron sessions silently ignore model override — modelApplied: true returned but wrong model runs | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When a cron job specifies a model override (e.g. `"model": "ollama/nemotron-3-super"`), the gateway returns `modelApplied: true` but silently runs the session on the default cloud model instead.
### Steps to reproduce
1. C... | closed | completed | true | 8 | [
"bug",
"regression"
] | [] | 2026-04-01T21:36:53Z | 2026-04-30T04:51:55Z | 2026-04-25T05:15:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | drkyleahuggins | 262,856,883 | U_kgDOD6rgsw | User | false |
openclaw/openclaw | 4,285,319,218 | I_kwDOQb6kR87_bMgy | 68,279 | https://github.com/openclaw/openclaw/issues/68279 | https://api.github.com/repos/openclaw/openclaw/issues/68279 | plugin-sdk/index.js doesn't export stringEnum/optionalStringEnum — breaks external plugins | ## Bug
`stringEnum` and `optionalStringEnum` are defined in `openclaw/plugin-sdk/core` but are **not re-exported** from `openclaw/plugin-sdk` (the main index).
Plugins that do:
```ts
import { stringEnum } from "openclaw/plugin-sdk";
```
fail at runtime with:
```
TypeError: (0 , _pluginSdk.stringEnum) is not a functio... | closed | completed | true | 4 | [] | [] | 2026-04-17T21:02:30Z | 2026-05-01T04:44:36Z | 2026-04-28T07:55:20Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | marzliak | 5,523,806 | MDQ6VXNlcjU1MjM4MDY= | User | false |
openclaw/openclaw | 4,285,421,146 | I_kwDOQb6kR87_blZa | 68,281 | https://github.com/openclaw/openclaw/issues/68281 | https://api.github.com/repos/openclaw/openclaw/issues/68281 | GLM 5.1 / non-frontier models: stopReason=error + output=0 turns silently abort with no retry path | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When ollama/glm-5.1:cloud (and, less commonly, other models) ends a turn with `stopReason="error"`, `usage.output=0`, and empty `content[]`, the embedded runner logs `incomplete turn detected: … stopReason=error payloads=0` and bails. The u... | closed | completed | true | 0 | [] | [] | 2026-04-17T21:26:50Z | 2026-04-23T04:22:15Z | 2026-04-20T23:43:52Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Chased1k | 20,195,034 | MDQ6VXNlcjIwMTk1MDM0 | User | false |
openclaw/openclaw | 4,285,460,418 | I_kwDOQb6kR87_bu_C | 68,282 | https://github.com/openclaw/openclaw/issues/68282 | https://api.github.com/repos/openclaw/openclaw/issues/68282 | [Bug]: status command is slow because telegram doctor contract loading uses slow jiti path | ## Summary
`openclaw status` and related status commands are slow on a fresh process because config validation loads Telegram doctor contract artifacts through the default `jiti` path, which is much slower than native loading for built `dist/extensions/*` JS artifacts.
This does not appear to be a gateway/runtime fai... | closed | completed | true | 2 | [] | [] | 2026-04-17T21:36:12Z | 2026-04-28T04:51:18Z | 2026-04-25T18:29:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ginishuh | 208,320,045 | U_kgDODGq2LQ | User | false |
openclaw/openclaw | 4,285,480,660 | I_kwDOQb6kR87_bz7U | 68,285 | https://github.com/openclaw/openclaw/issues/68285 | https://api.github.com/repos/openclaw/openclaw/issues/68285 | [Bug]: 100% CPU usage, broken tools, and no long-term memory on Node v25.9 / Matrix | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Critical failure of tools and long-term memory persistence, combined with 100% CPU exhaustion on a 2-core VDS after recent updates.
### Steps to reproduce
1. Perform a clean Ubuntu install on a VDS (2 CPU / 4GB RAM).
2. In... | open | null | false | 3 | [
"bug",
"regression"
] | [] | 2026-04-17T21:41:18Z | 2026-05-04T20:35:48Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | vvhyopen | 271,593,057 | U_kgDOEDAuYQ | User | false |
openclaw/openclaw | 4,285,532,256 | I_kwDOQb6kR87_cAhg | 68,287 | https://github.com/openclaw/openclaw/issues/68287 | https://api.github.com/repos/openclaw/openclaw/issues/68287 | [Bug]: openclaw node install fails with 'Unit file openclaw-gateway.service does not exist' on node-only VMs | ## Symptom
On a freshly provisioned node VM (no gateway service installed), running:
```
openclaw node install --host <gw> --port 18789 --runtime node --force
```
fails with:
```
Node install failed: Error: systemctl enable failed: Failed to enable unit:
Unit file openclaw-gateway.service does not exist.
```
The i... | closed | completed | true | 2 | [] | [] | 2026-04-17T21:55:19Z | 2026-04-30T04:38:38Z | 2026-04-26T05:36:11Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dlebee-agent | 272,684,230 | U_kgDOEEDUxg | User | false |
openclaw/openclaw | 4,285,552,354 | I_kwDOQb6kR87_cFbi | 68,291 | https://github.com/openclaw/openclaw/issues/68291 | https://api.github.com/repos/openclaw/openclaw/issues/68291 | Post-upgrade 2026.3.13 → 2026.4.15: silent config-schema breakage + Bedrock IAM role regression | ## Summary
After upgrading `openclaw` from **2026.3.13 → 2026.4.15** via `npm install -g`, the gateway enters a systemd restart loop. Two issues surfaced that are not called out in the CHANGELOG:
- **(A)** Pre-existing config files fail schema validation because three legacy keys were silently changed/removed, and th... | closed | completed | true | 2 | [] | [] | 2026-04-17T22:01:25Z | 2026-04-28T04:51:17Z | 2026-04-25T18:29:27Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | niceysam | 256,747,835 | U_kgDOD02pOw | User | false |
openclaw/openclaw | 4,101,019,059 | I_kwDOQb6kR870cJWz | 50,387 | https://github.com/openclaw/openclaw/issues/50387 | https://api.github.com/repos/openclaw/openclaw/issues/50387 | Feature Request: Auto model routing by task type within a single channel | ## Problem
Currently, OpenClaw only supports:
1. Manual model switching via `/model` command
2. Multi-agent routing by channel/account (different Telegram bots → different agents)
3. Model fallbacks on failure
But there's no way to automatically route to different models based on task complexity within a **single cha... | closed | not_planned | true | 2 | [] | [] | 2026-03-19T11:20:27Z | 2026-04-28T04:37:43Z | 2026-04-26T03:43:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rorschachachxd | 265,683,688 | U_kgDOD9YC6A | User | false |
openclaw/openclaw | 4,285,622,564 | I_kwDOQb6kR87_cWkk | 68,294 | https://github.com/openclaw/openclaw/issues/68294 | https://api.github.com/repos/openclaw/openclaw/issues/68294 | OpenAI-compatible audio transcription: HTTP 400 'not multipart/form-data' on Node 24 due to cross-realm FormData/undici dispatcher mismatch | # Bug: OpenAI-compatible audio transcription fails with `request Content-Type isn't multipart/form-data` on Node 24
## TL;DR
On **Node 24.15.0 + OpenClaw 2026.4.15**, any audio transcription request routed through `transcribeOpenAiCompatibleAudio` (Groq, OpenAI, Mistral voxtral, Moonshot, Qwen — anything that end... | closed | completed | true | 1 | [] | [] | 2026-04-17T22:19:18Z | 2026-04-25T04:15:48Z | 2026-04-23T03:56:38Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Jcxu97 | 125,407,588 | U_kgDOB3mRZA | User | false |
openclaw/openclaw | 3,887,833,024 | I_kwDOQb6kR87nu5_A | 7,468 | https://github.com/openclaw/openclaw/issues/7468 | https://api.github.com/repos/openclaw/openclaw/issues/7468 | [Feature]: Dynamic Model Selection Based on Context Size | ## Summary
Currently, OpenClaw uses a fixed model selection strategy (primary + fallback chain) regardless of the input prompt size. This means that small queries use the same expensive/powerful model as large, complex ones, leading to:
- **Unnecessary costs** for simple tasks
- **Suboptimal performance** when a light... | closed | not_planned | true | 4 | [
"enhancement",
"dedupe:parent"
] | [] | 2026-02-02T21:24:04Z | 2026-04-27T05:08:28Z | 2026-04-24T16:38:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | steromano87 | 8,759,452 | MDQ6VXNlcjg3NTk0NTI= | User | false |
openclaw/openclaw | 4,285,637,909 | I_kwDOQb6kR87_caUV | 68,295 | https://github.com/openclaw/openclaw/issues/68295 | https://api.github.com/repos/openclaw/openclaw/issues/68295 | Transcript jsonl stops appending mid-chat for telegram sessions (v2026.4.15) | ## Summary
For a telegram chat session with a stable sessionId, the session transcript jsonl file stops receiving appended entries mid-chat, even though `sessions.json.updatedAt` continues to be bumped for the same sessionId on every turn. Result: the on-disk transcript is missing an arbitrary tail of real turns, whic... | closed | completed | true | 1 | [] | [] | 2026-04-17T22:23:30Z | 2026-04-28T04:51:17Z | 2026-04-25T18:29:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | evilchan666 | 98,579,652 | U_kgDOBeA0xA | User | false |
openclaw/openclaw | 4,285,646,688 | I_kwDOQb6kR87_ccdg | 68,297 | https://github.com/openclaw/openclaw/issues/68297 | https://api.github.com/repos/openclaw/openclaw/issues/68297 | [Feature]: Add --daemon-env flag to openclaw node install | ## Summary
Add a repeatable `--daemon-env KEY=VALUE` flag to `openclaw node install` so that extra `Environment=` lines can be injected into the generated systemd/launchd unit at install time — matching the flag already added to `openclaw onboard --install-daemon` in #68240.
## Motivation
The same problem that motiv... | closed | completed | true | 3 | [] | [] | 2026-04-17T22:26:18Z | 2026-04-30T04:38:29Z | 2026-04-26T05:41:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dlebee-agent | 272,684,230 | U_kgDOEEDUxg | User | false |
openclaw/openclaw | 4,285,647,063 | I_kwDOQb6kR87_ccjX | 68,299 | https://github.com/openclaw/openclaw/issues/68299 | https://api.github.com/repos/openclaw/openclaw/issues/68299 | [Bug]: Matrix plugin ignores dangerouslyAllowPrivateNetwork config key, self-hosted LAN homeservers cannot connect | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
The Matrix plugin's createMatrixClient does not accept or forward any SSRF policy parameter, so the channels.matrix.network.dangerouslyAllowPrivateNetwork config key defined in the schema has no effect and connections to hom... | closed | completed | true | 0 | [
"bug",
"regression"
] | [] | 2026-04-17T22:26:27Z | 2026-04-21T04:22:19Z | 2026-04-18T04:50:52Z | NONE | null | 20260421T060038Z | 2026-04-21T06:00:38Z | nickian | 590,678 | MDQ6VXNlcjU5MDY3OA== | User | false |
openclaw/openclaw | 4,285,656,611 | I_kwDOQb6kR87_ce4j | 68,303 | https://github.com/openclaw/openclaw/issues/68303 | https://api.github.com/repos/openclaw/openclaw/issues/68303 | [OpenClaw][Dashboard] GUI accepts and buffers user messages with no UI indicator when sandbox SSH channel goes stale; flush arrives minutes later |
### Description
When the in-sandbox OpenClaw gateway's SSH channel goes stale (the symptom in the documented `WARN openshell_sandbox::ssh: data on unknown channel ChannelId(N)` logs), the browser dashboard continues to accept user input with no UI indication that the connection is degraded. Messages sent by the user ... | closed | completed | true | 1 | [] | [] | 2026-04-17T22:29:27Z | 2026-04-30T04:35:11Z | 2026-04-27T09:59:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | davidglogan | 31,814,247 | MDQ6VXNlcjMxODE0MjQ3 | User | false |
openclaw/openclaw | 4,285,663,608 | I_kwDOQb6kR87_cgl4 | 68,305 | https://github.com/openclaw/openclaw/issues/68305 | https://api.github.com/repos/openclaw/openclaw/issues/68305 | Reasoning stream warning incorrectly says 'Telegram only' but works in Lark/Feishu | When using the /reasoning command in Lark (Feishu), the bot returns a warning message:
⚙️ Reasoning stream enabled (Telegram only).
However, the reasoning stream actually works fine in Lark/Feishu as well. The warning message is misleading and suggests the feature is not supported in Lark, but it is functional.
**St... | open | null | false | 1 | [] | [] | 2026-04-17T22:31:39Z | 2026-05-02T20:57:29Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | woaiCarlos | 117,679,341 | U_kgDOBwOk7Q | User | false |
openclaw/openclaw | 4,285,709,181 | I_kwDOQb6kR87_crt9 | 68,309 | https://github.com/openclaw/openclaw/issues/68309 | https://api.github.com/repos/openclaw/openclaw/issues/68309 | [Bug]: A problem with mistral small thinking enabled | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
So i get this output that have lots of [object Object][object Object][object Object][object Object] somthing like this and then i see the real normal output that happens only when thinking is enabled for mistra... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-17T22:44:39Z | 2026-05-03T04:59:39Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | elmehdilaalali | 177,358,668 | U_kgDOCpJHTA | User | false |
openclaw/openclaw | 4,285,740,424 | I_kwDOQb6kR87_czWI | 68,315 | https://github.com/openclaw/openclaw/issues/68315 | https://api.github.com/repos/openclaw/openclaw/issues/68315 | [Bug]: Daily session reset stopped working in 4.14+ (regression from 4.5) | ## Description
Daily session reset (mode: "daily", atHour: 4 UTC) stopped working after upgrading from OC 4.5 to 4.14. Sessions that were idle 12+ hours are not resetting, and `/status` never shows 0 context tokens on daily boundary.
## Environment
- OpenClaw 2026.4.15 (previously 4.5, then 4.14)
- Discord multi-agen... | closed | completed | true | 2 | [
"duplicate",
"close:duplicate"
] | [] | 2026-04-17T22:55:03Z | 2026-04-28T04:44:55Z | 2026-04-26T00:28:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | maxatv | 215,872,058 | U_kgDODN3yOg | User | false |
openclaw/openclaw | 4,047,125,957 | I_kwDOQb6kR87xOj3F | 41,342 | https://github.com/openclaw/openclaw/issues/41342 | https://api.github.com/repos/openclaw/openclaw/issues/41342 | [Bug]: TTS audio truncated (last words cut off) when using custom OpenAI-compatible endpoint (Kokoro-FastAPI) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
**Key findings:**
- Kokoro logs show Muxed final packets completing successfully after the 200 OK is returned — meaning Kokoro streams chunks and OpenClaw reads the response before all chunks are written
- The .opus file saved in /tmp/open... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-09T18:11:18Z | 2026-04-30T04:41:07Z | 2026-04-26T04:59:52Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | stuartm1023 | 31,705,693 | MDQ6VXNlcjMxNzA1Njkz | User | false |
openclaw/openclaw | 4,285,751,764 | I_kwDOQb6kR87_c2HU | 68,319 | https://github.com/openclaw/openclaw/issues/68319 | https://api.github.com/repos/openclaw/openclaw/issues/68319 | [Bug]: Task Reconciler drops paused/dormant projects from active task tracking | ## Problem
The Task Reconciler system only reconciles tasks that exist in `tasks/active.yaml` and `tasks/ad-hoc/*.yaml`. Paused, dormant, or "LATER" projects that are documented in `MEMORY.md` and `memory/knowledge/projects/` are invisible to the task system.
This causes:
1. **Silent project disappearance** — projec... | closed | not_planned | true | 1 | [] | [] | 2026-04-17T22:58:34Z | 2026-04-28T04:51:16Z | 2026-04-25T18:29:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | GodotH | 804,151 | MDQ6VXNlcjgwNDE1MQ== | User | false |
openclaw/openclaw | 4,285,736,738 | I_kwDOQb6kR87_cyci | 68,314 | https://github.com/openclaw/openclaw/issues/68314 | https://api.github.com/repos/openclaw/openclaw/issues/68314 | [Bug]: pnpm install --no-optional reports false broken-lockfile error for optional deps | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
`CI=true pnpm install --no-frozen-lockfile --no-optional` fails in a current source checkout with `ERR_PNPM_LOCKFILE_MISSING_DEPENDENCY` for optional packages even though the referenced entries exist in `pnpm-l... | closed | completed | true | 2 | [] | [
"vincentkoc"
] | 2026-04-17T22:53:36Z | 2026-05-02T04:27:29Z | 2026-04-29T19:01:15Z | CONTRIBUTOR | null | 20260502T060142Z | 2026-05-02T06:01:42Z | ai-hpc | 183,861,985 | U_kgDOCvWC4Q | User | false |
openclaw/openclaw | 4,285,780,324 | I_kwDOQb6kR87_c9Fk | 68,323 | https://github.com/openclaw/openclaw/issues/68323 | https://api.github.com/repos/openclaw/openclaw/issues/68323 | feat(tts): add provider-agnostic TTS personas in speech-core | ## Summary
Add a generic TTS persona layer to OpenClaw so spoken replies can use a stable, reusable speech persona across providers, instead of relying on the assistant model to restate voice/style instructions ad hoc in each reply.
This should be provider-agnostic at the speech-core level, while allowing provider-sp... | closed | completed | true | 2 | [] | [] | 2026-04-17T23:07:19Z | 2026-04-28T04:34:59Z | 2026-04-26T04:12:40Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | barronlroth | 5,567,884 | MDQ6VXNlcjU1Njc4ODQ= | User | false |
openclaw/openclaw | 4,285,796,498 | I_kwDOQb6kR87_dBCS | 68,326 | https://github.com/openclaw/openclaw/issues/68326 | https://api.github.com/repos/openclaw/openclaw/issues/68326 | [Bug]: Sub-agents default to amazon-bedrock instead of custom Anthropic provider, causes credit use; Need default model setup guidance | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Sub-agents use amazon-bedrock as the default provider instead of Anthropic API key, leading to unwanted credit consumption. Main agent works fine, but sub-agents inherit wrong providers from saved sessions.
##... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-17T23:11:42Z | 2026-04-28T04:51:15Z | 2026-04-25T18:29:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kpelts1967 | 210,115,087 | U_kgDODIYaDw | User | false |
openclaw/openclaw | 4,285,909,775 | I_kwDOQb6kR87_dcsP | 68,330 | https://github.com/openclaw/openclaw/issues/68330 | https://api.github.com/repos/openclaw/openclaw/issues/68330 | browser screenshot consistently times out on Windows (CDP works fine) | ## Environment
- **OS:** Windows 10 (10.0.26200, x64)
- **OpenClaw:** 2026.4.15
- **Node:** v24.14.1
- **Browser:** Chrome 147 (managed/openclaw profile)
- **CDP port:** 18800
## Problem
The `browser screenshot` tool consistently times out regardless of timeout value (tested 10s, 15s, 120s). All other browser tools wo... | closed | completed | true | 1 | [] | [] | 2026-04-17T23:43:23Z | 2026-04-27T04:41:56Z | 2026-04-25T01:35:12Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Woodylai24 | 67,989,524 | MDQ6VXNlcjY3OTg5NTI0 | User | false |
openclaw/openclaw | 4,285,899,710 | I_kwDOQb6kR87_daO- | 68,329 | https://github.com/openclaw/openclaw/issues/68329 | https://api.github.com/repos/openclaw/openclaw/issues/68329 | cli-backend never triggers compaction (native or plugin) — context grows unbounded | ## Summary
The `cli-backend` (`provider=claude-cli`) never triggers any form of context compaction — neither native (`agents.defaults.compaction.reserveTokens`) nor plugin-based (lossless-claw). Context grows unbounded past the model window with zero compactions.
## Environment
- OpenClaw 2026.4.15
- Provider: `clau... | closed | completed | true | 4 | [] | [] | 2026-04-17T23:40:43Z | 2026-04-28T04:35:38Z | 2026-04-26T04:04:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jacko2bot | 261,461,124 | U_kgDOD5WUhA | User | false |
openclaw/openclaw | 4,285,980,114 | I_kwDOQb6kR87_dt3S | 68,336 | https://github.com/openclaw/openclaw/issues/68336 | https://api.github.com/repos/openclaw/openclaw/issues/68336 | [Bug]: /btw Side Channel Fails with Ollama Cloud (404 Error) | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
The `/btw` side channel command returns 404 errors when using Ollama
Cloud with `api: "openai-completions"`. The side channel appears to
skip URL path construction that regular API calls perform, hitting the
marketing site inst... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-04-18T00:04:13Z | 2026-04-25T04:15:23Z | 2026-04-23T02:59:12Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | sbmilburn | 11,755,869 | MDQ6VXNlcjExNzU1ODY5 | User | false |
openclaw/openclaw | 4,286,013,205 | I_kwDOQb6kR87_d18V | 68,340 | https://github.com/openclaw/openclaw/issues/68340 | https://api.github.com/repos/openclaw/openclaw/issues/68340 | Gateway config.patch restart kills in-flight agent turn, causing corrupted session history and response loop | ## Bug Description
When `config.patch` triggers a gateway restart while an agent turn is actively executing (e.g., writing to files via tool calls), the restart immediately kills the in-flight turn. This causes two compounding problems:
### Problem 1: No graceful shutdown for active turns
`config.patch` triggers an ... | closed | completed | true | 1 | [] | [] | 2026-04-18T00:14:42Z | 2026-04-28T04:51:14Z | 2026-04-25T18:30:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | OctoClaws | 268,375,381 | U_kgDOD_8VVQ | User | false |
openclaw/openclaw | 4,285,943,928 | I_kwDOQb6kR87_dlB4 | 68,333 | https://github.com/openclaw/openclaw/issues/68333 | https://api.github.com/repos/openclaw/openclaw/issues/68333 | [Bug] Telegram customCommands are unconditionally overwritten by native command sync on startup | ## Bug Description
In recent versions (tested on 2026.4.15), the `customCommands` list defined in `channels.telegram` is ignored or overwritten by the built-in native command set (~17-21 commands) every time the OpenClaw gateway restarts.
## Detailed Scenario
1. User defines a full set of 40 custom commands in `open... | closed | completed | true | 3 | [] | [] | 2026-04-17T23:53:23Z | 2026-04-28T04:51:14Z | 2026-04-25T18:30:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | githubdulong | 59,136,159 | MDQ6VXNlcjU5MTM2MTU5 | User | false |
openclaw/openclaw | 4,286,055,953 | I_kwDOQb6kR87_eAYR | 68,343 | https://github.com/openclaw/openclaw/issues/68343 | https://api.github.com/repos/openclaw/openclaw/issues/68343 | Compaction loop: calculateContextTokens crashes on missing usage.input (Anthropic OAuth path) | ## Bug Description
During compaction, `completeSimple(...)` returns a response whose `usage` field is missing or partial, causing `calculateContextTokens(usage)` to crash with:
```
Cannot read properties of undefined (reading 'input')
```
In the gateway log this surfaces as:
```
[compaction] Full summarization fail... | closed | completed | true | 3 | [] | [] | 2026-04-18T00:26:55Z | 2026-04-28T04:51:12Z | 2026-04-25T18:30:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,286,116,014 | I_kwDOQb6kR87_ePCu | 68,344 | https://github.com/openclaw/openclaw/issues/68344 | https://api.github.com/repos/openclaw/openclaw/issues/68344 | [Bug]: Ollama integration issues: context window ignored + tool support blocking local models | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Hi OpenClaw team,
I wanted to report a set of issues I encountered while trying to use OpenClaw with a local Ollama setup. I’m documenting this step-by-step so it’s easy to reproduce.
Environment
macOS (Apple... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-18T00:45:45Z | 2026-04-29T04:33:03Z | 2026-04-27T01:35:20Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | Stp155906 | 71,754,706 | MDQ6VXNlcjcxNzU0NzA2 | User | false |
openclaw/openclaw | 4,286,199,022 | I_kwDOQb6kR87_ejTu | 68,347 | https://github.com/openclaw/openclaw/issues/68347 | https://api.github.com/repos/openclaw/openclaw/issues/68347 | [Bug]: Schema .strict() rejects `paperclip` property from Paperclip 2026.416.0 wake payload | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## Summary
After upgrading Paperclip to `2026.416.0`, all `openclaw_gateway` adapter runs fail with:
`invalid agent params: at root: unexpected property 'paperclip'`
Paperclip now injects a `paperclip: {...}` object into ... | closed | completed | true | 0 | [
"bug",
"regression"
] | [] | 2026-04-18T01:16:31Z | 2026-04-20T04:26:25Z | 2026-04-18T03:40:07Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | BlueSky330 | 98,229,064 | U_kgDOBdrbSA | User | false |
openclaw/openclaw | 4,286,264,742 | I_kwDOQb6kR87_ezWm | 68,352 | https://github.com/openclaw/openclaw/issues/68352 | https://api.github.com/repos/openclaw/openclaw/issues/68352 | UX: misleading plugin warnings when allowlist entries don't match discovered plugin IDs | ## Problem
When a user configures `plugins.allow` with channel IDs (e.g. `"feishu"`) instead of actual plugin IDs (e.g. `"openclaw-lark"`), two confusing warnings are emitted at startup:
```
[plugins] plugins.allow is empty; discovered non-bundled plugins may auto-load: openclaw-lark (C:\Users\...\extensions\feishu-o... | open | null | false | 3 | [] | [] | 2026-04-18T01:40:12Z | 2026-05-03T03:37:20Z | null | CONTRIBUTOR | null | 20260503T060030Z | 2026-05-03T06:00:30Z | JIRBOY | 49,422,297 | MDQ6VXNlcjQ5NDIyMjk3 | User | false |
openclaw/openclaw | 4,286,277,388 | I_kwDOQb6kR87_e2cM | 68,354 | https://github.com/openclaw/openclaw/issues/68354 | https://api.github.com/repos/openclaw/openclaw/issues/68354 | Dreaming narrative sessions never reuse — session key includes timestamp causing unbounded growth | ## Bug Description
When `memory-core` dreaming is enabled, the Dream Diary narrative subagent creates a **new session on every run** instead of reusing an existing one. The session key includes a millisecond timestamp, making it unique each time:
```js
// dreaming-narrative.ts
function buildNarrativeSessionKey(params... | closed | completed | true | 1 | [] | [] | 2026-04-18T01:44:20Z | 2026-04-28T04:51:11Z | 2026-04-25T18:30:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | happer6012 | 277,163,421 | U_kgDOEIUtnQ | User | false |
openclaw/openclaw | 4,286,269,090 | I_kwDOQb6kR87_e0ai | 68,353 | https://github.com/openclaw/openclaw/issues/68353 | https://api.github.com/repos/openclaw/openclaw/issues/68353 | [Feature]: first-class per-message author attribution on chat.send for multi-user shared sessions | ### Summary
Add an optional, validated `author` object to `ChatSendParamsSchema` so clients can declare per-user identity on a per-message basis, and surface it as the authoritative `senderLabel`/`authorId`/`authorName` on `chat.history`, overriding the install-class `client.id`.
### Problem to solve
OpenClaw's attr... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-04-18T01:41:26Z | 2026-05-02T22:58:54Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | moogoofoo | 141,287,755 | U_kgDOCGvhSw | User | false |
openclaw/openclaw | 4,286,336,129 | I_kwDOQb6kR87_fEyB | 68,357 | https://github.com/openclaw/openclaw/issues/68357 | https://api.github.com/repos/openclaw/openclaw/issues/68357 | Group chat: Agents automatically @mention other agents when multiple bots are mentioned | ### Problem Description
When a user @mentions multiple agents in a Feishu group chat (e.g., \`@claw-coder @claw-designer @claw-pm @claw-qa\`), each agent's reply automatically includes mentions of the other agents.
The framework injects this System message:
```
[System: Your reply will automatically @mention: claw-des... | closed | not_planned | true | 1 | [] | [] | 2026-04-18T02:04:10Z | 2026-04-30T04:39:23Z | 2026-04-26T05:19:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lidonggg | 33,049,345 | MDQ6VXNlcjMzMDQ5MzQ1 | User | false |
openclaw/openclaw | 4,286,351,743 | I_kwDOQb6kR87_fIl_ | 68,363 | https://github.com/openclaw/openclaw/issues/68363 | https://api.github.com/repos/openclaw/openclaw/issues/68363 | [Bug]: Heartbeat with LM Studio repeatedly loads model multiple times, ends in incomplete turn, occasionally succeeds | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Since the last couple of OpenClaw updates, heartbeat appears to trigger repeated LM Studio model loads for my heartbeat model instead of performing a single run.
The configured heartbeat model is `lmstudio/qwen/qwen3.5-9b`.... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-04-18T02:09:40Z | 2026-04-28T04:51:10Z | 2026-04-25T18:30:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | arvindvenkataramani | 250,267 | MDQ6VXNlcjI1MDI2Nw== | User | false |
openclaw/openclaw | 4,286,401,066 | I_kwDOQb6kR87_fUoq | 68,367 | https://github.com/openclaw/openclaw/issues/68367 | https://api.github.com/repos/openclaw/openclaw/issues/68367 | Light Sleep dreaming phase contaminates unrelated sessions with cross-context memory candidates | ## Bug Description
The **light sleep** dreaming phase writes candidate memories into the daily memory file (`memory/YYYY-MM-DD.md`) as a managed block between `<!-- openclaw:dreaming:light:start -->` and `<!-- openclaw:dreaming:light:end -->` markers. This block is then bootstrapped into **every session** that shares ... | closed | completed | true | 1 | [] | [] | 2026-04-18T02:27:51Z | 2026-04-28T04:51:09Z | 2026-04-25T18:30:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | captivationhub | 224,922,961 | U_kgDODWgNUQ | User | false |
openclaw/openclaw | 4,286,409,537 | I_kwDOQb6kR87_fWtB | 68,369 | https://github.com/openclaw/openclaw/issues/68369 | https://api.github.com/repos/openclaw/openclaw/issues/68369 | Web UI: new chat window stretches full width (regression) | ## Description
The "new chat" welcome state and message thread in the web dashboard stretch to the full width of the content area instead of being constrained to a readable column width.
This is a regression — the fix was previously implemented on branch `issue-57235-web-sessions` (commit `0df41d246c`) but was never ... | closed | not_planned | true | 1 | [] | [] | 2026-04-18T02:32:33Z | 2026-04-30T04:39:22Z | 2026-04-26T05:19:34Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | OpenCodeEngineer | 261,470,075 | U_kgDOD5W3ew | User | false |
openclaw/openclaw | 4,286,418,787 | I_kwDOQb6kR87_fY9j | 68,371 | https://github.com/openclaw/openclaw/issues/68371 | https://api.github.com/repos/openclaw/openclaw/issues/68371 | memory-wiki bridge mode shows 0 exported artifacts without QMD | ## Problem
When using `memory-wiki` in `bridge` mode without QMD as the active memory backend, the bridge reports **0 exported artifacts** even though dream reports and daily notes exist in the active memory system.
```json
{
"vaultMode": "bridge",
"bridge": {
"enabled": true,
"readMemoryArtifacts": true,... | closed | completed | true | 7 | [
"dedupe:child"
] | [
"vincentkoc"
] | 2026-04-18T02:37:41Z | 2026-04-29T04:29:22Z | 2026-04-26T07:09:29Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | guojiongming | 266,004,757 | U_kgDOD9rpFQ | User | false |
openclaw/openclaw | 4,286,472,917 | I_kwDOQb6kR87_fmLV | 68,376 | https://github.com/openclaw/openclaw/issues/68376 | https://api.github.com/repos/openclaw/openclaw/issues/68376 | [Bug]: 在对话框中，当需要调用工具exec时，必会自动重启 | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
当我让AI助手执行任务时，需要调用工具exec时必然会造成openclaw自动重启。调用其他工具的时候没有问题。
### Steps to reproduce
让AI执行任务，需要调用工具exec的时候
### Expected behavior
gateway退出并自行重启
### Actual behavior
在与AI助手对话时，不需要调用工具的时候是正常的，需要调用exec时发生，调用其他工具正常。
### OpenClaw v... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-18T03:03:59Z | 2026-04-28T04:51:08Z | 2026-04-25T18:31:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | lsr376834640 | 24,834,250 | MDQ6VXNlcjI0ODM0MjUw | User | false |
openclaw/openclaw | 4,286,487,419 | I_kwDOQb6kR87_fpt7 | 68,379 | https://github.com/openclaw/openclaw/issues/68379 | https://api.github.com/repos/openclaw/openclaw/issues/68379 | 🐛 memory-core: Dreaming subsystem causes memory bloat (7.8MB → 3.7GB Gateway crash) and session-memory plugin floods memory directory with raw transcripts | ## Problem Summary
The `memory-core` plugin's dreaming feature and session-memory auto-dump mechanism caused severe memory bloat, leading to a Gateway crash with 3.7GB RSS memory usage.
## Root Cause Analysis
### Issue 1: Dreaming writes without cleanup
- The `.dreams/` directory accumulated 7.8MB of data (149,224 l... | closed | completed | true | 1 | [] | [] | 2026-04-18T03:11:12Z | 2026-04-28T04:51:07Z | 2026-04-25T18:31:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xiux1909 | 271,993,296 | U_kgDOEDZJ0A | User | false |
openclaw/openclaw | 4,286,491,224 | I_kwDOQb6kR87_fqpY | 68,380 | https://github.com/openclaw/openclaw/issues/68380 | https://api.github.com/repos/openclaw/openclaw/issues/68380 | WSL2: gateway service runs but openclaw gateway start fails when user D-Bus socket is missing | ## Summary
On WSL2 Ubuntu, OpenClaw Gateway can be running and reachable, but `openclaw gateway start` fails with:
```text
Gateway start failed: Error: systemctl not available; systemd user services are required on Linux.
Tip: openclaw gateway install
Tip: openclaw gateway
Tip: systemctl --user start openclaw-gateway... | closed | completed | true | 1 | [] | [] | 2026-04-18T03:12:51Z | 2026-04-28T04:51:06Z | 2026-04-25T18:31:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | PhanAnh-V | 219,108,224 | U_kgDODQ9TgA | User | false |
openclaw/openclaw | 4,286,485,542 | I_kwDOQb6kR87_fpQm | 68,378 | https://github.com/openclaw/openclaw/issues/68378 | https://api.github.com/repos/openclaw/openclaw/issues/68378 | Bug: Gateway sends Discord session messages to Feishu channel (critical message routing bug) | ## Bug Summary
**Gateway sends Discord session messages to Feishu channel instead of Discord**
When an agent responds in a Discord direct message session, the message is incorrectly delivered to Feishu private chat instead of Discord. This is a critical message routing bug in the Gateway delivery subsystem.
## Envir... | closed | not_planned | true | 3 | [] | [] | 2026-04-18T03:10:19Z | 2026-04-30T04:39:21Z | 2026-04-26T05:19:38Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Suidge | 34,770,064 | MDQ6VXNlcjM0NzcwMDY0 | User | false |
openclaw/openclaw | 4,286,447,249 | I_kwDOQb6kR87_ff6R | 68,374 | https://github.com/openclaw/openclaw/issues/68374 | https://api.github.com/repos/openclaw/openclaw/issues/68374 | [Feature]: Expose claude-cli thinking blocks as reasoning on /v1/responses (and /v1/chat/completions)` | ### Summary
When a claude-cli-backed agent is invoked through the Gateway's HTTP API (`POST /v1/responses` or `POST /v1/chat/completions`), the reasoning/thinking content Claude Code CLI emits internally is stripped before reaching the client. Please forward it on the HTTP surface — the same way `reasoningDefault` alr... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-04-18T02:52:29Z | 2026-05-02T22:58:55Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | jkeen871 | 8,370,294 | MDQ6VXNlcjgzNzAyOTQ= | User | false |
openclaw/openclaw | 4,286,445,284 | I_kwDOQb6kR87_ffbk | 68,373 | https://github.com/openclaw/openclaw/issues/68373 | https://api.github.com/repos/openclaw/openclaw/issues/68373 | [Regression v2026.4.15] Cached plugin restore drops memory capability, breaks wiki bridge imports | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to v2026.4.15, `openclaw wiki status` reports "Bridge: enabled (0 exported artifacts)" and `openclaw wiki bridge import` returns 0 artifacts across 0 workspaces, despite `memory-core` plugin being active and ... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-04-18T02:51:08Z | 2026-04-28T04:51:09Z | 2026-04-25T18:31:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | RoseKongPS | 270,016,663 | U_kgDOEBgglw | User | false |
openclaw/openclaw | 4,217,554,649 | I_kwDOQb6kR877YsbZ | 62,465 | https://github.com/openclaw/openclaw/issues/62465 | https://api.github.com/repos/openclaw/openclaw/issues/62465 | [Bug] edit tool requires exact text match causing memory write failures | ## Bug Description
The edit tool requires exact string matching between oldText and file content, including all whitespace and newlines. When a file is modified by another process or concurrent session, the edit fails silently or with 'Could not find the exact text' error.
## Frequency
7 occurrences observed across 20... | closed | completed | true | 2 | [] | [] | 2026-04-07T12:11:48Z | 2026-04-29T04:38:14Z | 2026-04-26T09:50:30Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jaydenlichong29-create | 268,866,858 | U_kgDOEAaVKg | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.