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 | 4,139,074,269 | I_kwDOQb6kR872tULd | 54,743 | https://github.com/openclaw/openclaw/issues/54743 | https://api.github.com/repos/openclaw/openclaw/issues/54743 | config set throws Maximum call stack size exceeded | ## Summary
`openclaw config set` crashes with `RangeError: Maximum call stack size exceeded` when writing to `openclaw.json`. The config file is not modified. Read operations (`openclaw config get`) work fine.
## Environment
- OpenClaw 2026.3.24 (cff6dc9)
- macOS (Apple Silicon)
- Node.js via Homebrew (`/opt/homebre... | closed | completed | true | 1 | [] | [] | 2026-03-25T22:45:16Z | 2026-04-29T04:32:48Z | 2026-04-27T02:01:24Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | thomaswillner | 204,787,723 | U_kgDODDTQCw | User | false |
openclaw/openclaw | 4,140,223,507 | I_kwDOQb6kR872xswT | 54,854 | https://github.com/openclaw/openclaw/issues/54854 | https://api.github.com/repos/openclaw/openclaw/issues/54854 | Control UI regression: drag-and-drop file to chat no longer works / old 'drop to local file path' workflow is gone | ## Summary
After updating to OpenClaw 3.24, the Web UI / Control UI chat no longer supports the old drag-and-drop file workflow.
Previously, I could drag a local file directly into the chat input, and it would turn into a usable local file path/reference in the message. That workflow was much faster and more practica... | closed | completed | true | 1 | [] | [] | 2026-03-26T03:31:27Z | 2026-04-29T04:32:47Z | 2026-04-27T02:01:30Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | XinGaoSH | 104,917,561 | U_kgDOBkDqOQ | User | false |
openclaw/openclaw | 4,142,094,135 | I_kwDOQb6kR87241c3 | 55,063 | https://github.com/openclaw/openclaw/issues/55063 | https://api.github.com/repos/openclaw/openclaw/issues/55063 | [Bug]: /new and /reset preserve session model override; reset banner default can differ from agents.defaults.model.primary | ### What happened?
In a WhatsApp direct session, `/new` does not clear the session-level model override.
It starts a fresh session ID, but the new session still inherits the old `providerOverride` / `modelOverride`, so the chat remains pinned to the previous model instead of falling back to the configured default.
A... | closed | not_planned | true | 1 | [] | [] | 2026-03-26T09:17:45Z | 2026-04-29T04:32:46Z | 2026-04-27T02:01:43Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | gshlomi | 4,084,415 | MDQ6VXNlcjQwODQ0MTU= | User | false |
openclaw/openclaw | 4,142,279,347 | I_kwDOQb6kR8725iqz | 55,074 | https://github.com/openclaw/openclaw/issues/55074 | https://api.github.com/repos/openclaw/openclaw/issues/55074 | Make MEDIA_MAX_BYTES configurable instead of hardcoded 5MB | ## Summary
`MEDIA_MAX_BYTES` is hardcoded to `5 * 1024 * 1024` (5MB) in `src/media/store.ts:14` with no configuration option to override it. This causes file uploads from messaging channels (e.g. WeCom) to silently fail when the file is slightly over 5MB.
## Problem
When a user sends a 6.3MB Excel file via WeCom, th... | closed | not_planned | true | 1 | [] | [] | 2026-03-26T09:43:57Z | 2026-04-29T04:32:45Z | 2026-04-27T02:01:50Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | JinluZhang | 47,548,345 | MDQ6VXNlcjQ3NTQ4MzQ1 | User | false |
openclaw/openclaw | 4,143,696,510 | I_kwDOQb6kR872-8p- | 55,163 | https://github.com/openclaw/openclaw/issues/55163 | https://api.github.com/repos/openclaw/openclaw/issues/55163 | [Bug]: Doctor auto-enable creates config reload loop for local extensions in ~/.openclaw/extensions/ | ## Bug Type
Regression (config reload loop)
## Summary
When local extensions are placed in `~/.openclaw/extensions/`, `openclaw doctor` / `applyPluginAutoEnable` auto-adds them to `plugins.entries.<id>`, which triggers a config change, which triggers a reload, which re-runs the doctor — infinite loop. This prevents ... | closed | completed | true | 2 | [] | [] | 2026-03-26T12:59:51Z | 2026-04-29T04:32:44Z | 2026-04-27T02:01:59Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | austinmao | 6,487,975 | MDQ6VXNlcjY0ODc5NzU= | User | false |
openclaw/openclaw | 4,221,899,298 | I_kwDOQb6kR877pRIi | 62,890 | https://github.com/openclaw/openclaw/issues/62890 | https://api.github.com/repos/openclaw/openclaw/issues/62890 | [Bug]: Missing @img/colour module breaks inline screenshot rendering | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The sharp package is missing the @img/colour dependency, causing browser screenshots to save to disk but not render inline in chat.
### Steps to reproduce
1. Add Playwright MCP as browser tool in OpenClaw 202... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-08T03:35:37Z | 2026-04-29T04:32:44Z | 2026-04-27T02:02:04Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | piro257 | 75,042,356 | MDQ6VXNlcjc1MDQyMzU2 | User | false |
openclaw/openclaw | 4,223,512,499 | I_kwDOQb6kR877va-z | 63,075 | https://github.com/openclaw/openclaw/issues/63075 | https://api.github.com/repos/openclaw/openclaw/issues/63075 | Discord stale slash commands remain visible when channels.discord.commands.native=false | ## Description
When `channels.discord.commands.native` is set to `false`, Discord-native slash commands can remain visible/stale in the Discord UI instead of being fully cleared, which creates a confusing broken-command state.
In practice, this looks like:
- OpenClaw config says native Discord commands are disabled
-... | closed | completed | true | 1 | [] | [] | 2026-04-08T09:31:32Z | 2026-04-29T04:32:43Z | 2026-04-27T02:03:49Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | feoh | 330,070 | MDQ6VXNlcjMzMDA3MA== | User | false |
openclaw/openclaw | 4,224,012,667 | I_kwDOQb6kR877xVF7 | 63,119 | https://github.com/openclaw/openclaw/issues/63119 | https://api.github.com/repos/openclaw/openclaw/issues/63119 | [Feature]: Session compaction: archive transcript files before deletion, with configurable retention | ### Summary
## Problem
When a session undergoes compaction (either automatic due to context limits or manual reset), the transcript file is **permanently deleted** with no recovery mechanism. This causes:
1. **Irreversible data loss**: Hours/days of conversation history vanish instantly
2. **No grace period**: Delet... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-08T11:02:00Z | 2026-04-29T04:32:42Z | 2026-04-27T02:03:54Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | pickmatch | 267,536,034 | U_kgDOD_JGog | User | false |
openclaw/openclaw | 4,224,766,135 | I_kwDOQb6kR8770NC3 | 63,169 | https://github.com/openclaw/openclaw/issues/63169 | https://api.github.com/repos/openclaw/openclaw/issues/63169 | WhatsApp media send reports success but attachment is never delivered (2026.4.8) | ## Summary
WhatsApp media sending appears broken in OpenClaw 2026.4.8 for at least one group chat flow. Text messages send successfully, but video/media sends are logged as `Sent` by the gateway while the recipient never receives the attachment.
## Environment
- OpenClaw: 2026.4.8
- Channel: WhatsApp
- WhatsApp accoun... | closed | completed | true | 4 | [] | [] | 2026-04-08T13:13:28Z | 2026-04-29T04:32:41Z | 2026-04-27T02:03:59Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | claudiobedino | 259,268,385 | U_kgDOD3QfIQ | User | false |
openclaw/openclaw | 4,332,663,653 | I_kwDOQb6kR88AAAABAj8zZQ | 72,504 | https://github.com/openclaw/openclaw/issues/72504 | https://api.github.com/repos/openclaw/openclaw/issues/72504 | Feishu: bot strips its own <at> mention causing NO_REPLY in multi-bot groups | ## Problem
In Feishu group chats where multiple bots are @mentioned together, bots return NO_REPLY even when they were explicitly mentioned. This is caused by normalizeMentions() in extensions/feishu/src/bot-content.ts stripping the receiving bot's own <at> tag when botStripId is set.
## Root Cause
The normalizeMent... | open | null | false | 1 | [] | [] | 2026-04-27T02:04:55Z | 2026-05-03T09:13:05Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | AxelHu | 12,809,280 | MDQ6VXNlcjEyODA5Mjgw | User | false |
openclaw/openclaw | 4,332,669,618 | I_kwDOQb6kR88AAAABAj9Ksg | 72,505 | https://github.com/openclaw/openclaw/issues/72505 | https://api.github.com/repos/openclaw/openclaw/issues/72505 | Gateway infinite loop during initialization on v2026.4.23+ (CPU 100%+, never reaches HTTP bind) | # Bug Report: Gateway Infinite Loop During Initialization on v2026.4.23+
## Summary
OpenClaw gateway enters an infinite loop during startup on versions **v2026.4.23** and **v2026.4.24**, consuming 100%+ CPU and never reaching the HTTP bind phase. The issue is **not plugin-related** and occurs even with a minimal confi... | closed | completed | true | 2 | [] | [] | 2026-04-27T02:07:00Z | 2026-04-30T04:33:47Z | 2026-04-27T10:50:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | csuarezgithubdesa | 97,857,664 | U_kgDOBdUwgA | User | false |
openclaw/openclaw | 4,080,851,654 | I_kwDOQb6kR87zPNrG | 48,010 | https://github.com/openclaw/openclaw/issues/48010 | https://api.github.com/repos/openclaw/openclaw/issues/48010 | [Enhancement] Support for Ollama model-specific parameters like thinking mode control | ## Problem Description
When using Ollama with OpenClaw, certain models like `qwen3.5:9b` default to thinking mode, which significantly increases response times. Currently, OpenClaw lacks the ability to pass model-specific parameters to Ollama's API, preventing users from disabling thinking mode for better performance.... | closed | completed | true | 7 | [] | [] | 2026-03-16T07:58:18Z | 2026-04-29T04:32:39Z | 2026-04-27T02:08:33Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | tangzhi | 3,373,164 | MDQ6VXNlcjMzNzMxNjQ= | User | false |
openclaw/openclaw | 4,332,679,166 | I_kwDOQb6kR88AAAABAj9v_g | 72,507 | https://github.com/openclaw/openclaw/issues/72507 | https://api.github.com/repos/openclaw/openclaw/issues/72507 | Bug: sessions.json grows unbounded due to static schema serialized into dynamic session state | `sessions.json` grows unbounded because `skillsSnapshot` (all skill descriptions + metadata for all 60+ skills) and `systemPromptReport` (full tool schemas for all 150+ tools) are serialized into every session entry. These are static data.
Real example: `stock` agent 12 sessions = 1.8 MB of duplicated static schema da... | closed | not_planned | true | 1 | [] | [] | 2026-04-27T02:10:36Z | 2026-04-30T04:32:03Z | 2026-04-28T03:52:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tur9977 | 18,606,529 | MDQ6VXNlcjE4NjA2NTI5 | User | false |
openclaw/openclaw | 4,332,651,778 | I_kwDOQb6kR88AAAABAj8FAg | 72,500 | https://github.com/openclaw/openclaw/issues/72500 | https://api.github.com/repos/openclaw/openclaw/issues/72500 | [Bug]: Control UI webchat flashes credentials gate on tab re-focus due to immediate connected=false render | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Every time the browser tab loses and regains focus, the webchat briefly flashes the full credentials/login gate before reconnecting and restoring the normal chat UI. On fast connections this lasts ~300–800ms; o... | open | null | false | 1 | [] | [] | 2026-04-27T02:01:11Z | 2026-05-03T08:39:39Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | brandco | 8,084,507 | MDQ6VXNlcjgwODQ1MDc= | User | false |
openclaw/openclaw | 4,206,016,780 | I_kwDOQb6kR876srkM | 61,076 | https://github.com/openclaw/openclaw/issues/61076 | https://api.github.com/repos/openclaw/openclaw/issues/61076 | Bad control character in JSON at fixed position when parsing Anthropic streaming response (embedded agent runtime) | ## Summary
The embedded agent runtime throws `Bad control character in string literal in JSON at position N (line 1 column N+1)` intermittently during Anthropic streaming responses. The error terminates the agent run mid-response.
## Environment
- OpenClaw: 2026.4.1
- OS: macOS Tahoe 26.4 (arm64)
- Node: 24.14.1
- P... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T23:35:18Z | 2026-05-04T04:34:54Z | 2026-05-02T00:06:03Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | dcstahl01 | 273,449,056 | U_kgDOEEyAYA | User | false |
openclaw/openclaw | 4,222,861,894 | I_kwDOQb6kR877s8JG | 63,005 | https://github.com/openclaw/openclaw/issues/63005 | https://api.github.com/repos/openclaw/openclaw/issues/63005 | [Bug]: external raw session keys can split after sessions_send / nested agent canonicalization | ## Summary
Stable external session keys that do not start with agent-prefixed namespaces like `agent:` can split one logical conversation into multiple OpenClaw session records after internal `sessions_send` / nested `agent` activity.
## Steps to reproduce
1. Configure an external integration that delivers inbound mes... | open | null | false | 1 | [] | [] | 2026-04-08T07:32:06Z | 2026-05-03T13:21:16Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | songshikang0111 | 132,348,332 | U_kgDOB-N5rA | User | false |
openclaw/openclaw | 4,224,430,116 | I_kwDOQb6kR877y7Ak | 63,139 | https://github.com/openclaw/openclaw/issues/63139 | https://api.github.com/repos/openclaw/openclaw/issues/63139 | before_model_resolve hook fires once per fallback iteration in runWithModelFallback, defeating runtime failover for routing plugins | ## Summary
The `before_model_resolve` hook is invoked **once per fallback iteration** inside `runWithModelFallback` → `runAgentAttempt` → `runEmbeddedPiAgent`, not once per logical agent run. Routing plugins that unconditionally rewrite `providerOverride` / `modelOverride` therefore defeat the runtime's model fallback... | open | null | false | 2 | [] | [] | 2026-04-08T12:19:24Z | 2026-05-02T19:15:43Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | SmithU64 | 117,943,562 | U_kgDOBwetCg | User | false |
openclaw/openclaw | 4,227,378,131 | I_kwDOQb6kR877-KvT | 63,352 | https://github.com/openclaw/openclaw/issues/63352 | https://api.github.com/repos/openclaw/openclaw/issues/63352 | [Proposal]: Soft repeated file-reread loop detector in exec runtime to reduce quota burn and timeout failures | ## Summary
Proposal: add a soft runtime detector for repeated read-only exec calls against the same file or same file pair within one turn.
Goal:
- reduce quota burn from zero-progress file reread loops
- reduce timeout failures caused by these loops
- keep normal debugging workflows intact by starting with warning-on... | open | null | false | 2 | [] | [] | 2026-04-08T20:18:52Z | 2026-05-03T02:14:46Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | garadun1-cmyk | 251,936,924 | U_kgDODwRAnA | User | false |
openclaw/openclaw | 4,228,008,420 | I_kwDOQb6kR878Aknk | 63,398 | https://github.com/openclaw/openclaw/issues/63398 | https://api.github.com/repos/openclaw/openclaw/issues/63398 | OpenAI rejects @softeria/ms-365-mcp-server tool schemas with 'Validation Error: type' | ## Bug: OpenAI rejects @softeria/ms-365-mcp-server tool schemas with `Validation Error: type`
### Summary
When OpenClaw routes a tool-using request through LiteLLM to OpenAI (`gpt-5`, `gpt-5-mini`) with the `@softeria/ms-365-mcp-server` MCP tools registered, OpenAI returns HTTP 400 with `Validation Error: "type" - In... | open | null | false | 2 | [] | [] | 2026-04-08T22:11:46Z | 2026-05-04T18:47:18Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | tmote | 21,976,980 | MDQ6VXNlcjIxOTc2OTgw | User | false |
openclaw/openclaw | 4,228,209,423 | I_kwDOQb6kR878BVsP | 63,411 | https://github.com/openclaw/openclaw/issues/63411 | https://api.github.com/repos/openclaw/openclaw/issues/63411 | Streaming repetition loops bypass idle timeout; no semantic circuit breaker or context sanitization | Streaming repetition loops bypass idle timeout; no semantic circuit breaker or context sanitization
### Incident Overview
On 2026-04-08, an architectural gap was identified when `google/gemini-3-flash-preview` entered a repetition loop within its `<think>` block. Despite the loop continuing for an extended period, the... | open | null | false | 1 | [] | [] | 2026-04-08T22:54:30Z | 2026-05-03T01:22:33Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | Bruiserhq | 36,858,620 | MDQ6VXNlcjM2ODU4NjIw | User | false |
openclaw/openclaw | 4,228,393,104 | I_kwDOQb6kR878CCiQ | 63,425 | https://github.com/openclaw/openclaw/issues/63425 | https://api.github.com/repos/openclaw/openclaw/issues/63425 | modelByChannel override does not update session's persisted contextTokens — /status shows stale value | ## Description
When `modelByChannel` overrides the runtime model for a session (e.g. from `claude-opus-4-6` to `minimax`), the session's persisted `contextTokens` value is not updated. `/status` continues to show the old model's context window (e.g. 1M instead of minimax's actual 204k).
## Root Cause (source-code tra... | closed | completed | true | 2 | [] | [] | 2026-04-08T23:35:42Z | 2026-05-02T04:27:30Z | 2026-04-29T18:58:23Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,228,443,758 | I_kwDOQb6kR878CO5u | 63,432 | https://github.com/openclaw/openclaw/issues/63432 | https://api.github.com/repos/openclaw/openclaw/issues/63432 | [Bug]: Control UI on iPad/Tailscale fails on image send with `RangeError: Maximum call stack size exceeded` | # [Bug]: Control UI on iPad/Tailscale fails on image send with `RangeError: Maximum call stack size exceeded`
## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
Sending a message from the OpenClaw Control UI on an iPad via the Tailscale-served URL fails with:
`RangeError: Maximum call stack s... | open | null | false | 1 | [] | [] | 2026-04-08T23:50:32Z | 2026-05-02T21:25:30Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | christianhernandez3-afk | 274,654,821 | U_kgDOEF7mZQ | User | false |
openclaw/openclaw | 4,228,640,071 | I_kwDOQb6kR878C-1H | 63,454 | https://github.com/openclaw/openclaw/issues/63454 | https://api.github.com/repos/openclaw/openclaw/issues/63454 | Feature: Respect userTimezone in memory file timestamps and naming | ## Summary
Memory files generated during auto-flush/compaction always use UTC timestamps in both the filename and the header content (e.g. `2026-04-09 00:44:38 UTC`), even when `userTimezone` is configured (e.g. `Asia/Jakarta`).
## Current Behavior
- System prompt correctly shows local time via `userTimezone` ✅
- Me... | open | null | false | 1 | [] | [] | 2026-04-09T00:49:07Z | 2026-05-02T19:15:45Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | wheza99 | 109,982,804 | U_kgDOBo40VA | User | false |
openclaw/openclaw | 4,228,694,625 | I_kwDOQb6kR878DMJh | 63,463 | https://github.com/openclaw/openclaw/issues/63463 | https://api.github.com/repos/openclaw/openclaw/issues/63463 | MLX routing fails: model_not_found + fallback despite direct MLX API success | MLX provider model routing fails in OpenClaw (falls back to OpenAI) even though direct MLX API works.
Environment:
- OpenClaw: 2026.4.5 (issue persisted after npm update)
- macOS 26.3.1 arm64
- MLX: http://127.0.0.1:8080
Config:
- models.providers.mlx.baseUrl = http://127.0.0.1:8080
- models.providers.mlx.models[].id... | closed | completed | true | 3 | [] | [] | 2026-04-09T01:08:03Z | 2026-05-03T04:38:15Z | 2026-04-30T06:29:46Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | andhai | 20,992,727 | MDQ6VXNlcjIwOTkyNzI3 | User | false |
openclaw/openclaw | 4,228,899,443 | I_kwDOQb6kR878D-Jz | 63,488 | https://github.com/openclaw/openclaw/issues/63488 | https://api.github.com/repos/openclaw/openclaw/issues/63488 | [Bug]: ClawTalk WebSocket disconnect starves Slack Socket Mode event loop, causing stale-socket restart | ## Description
When the ClawTalk WebSocket connection (`wss://clawdtalk.com/ws`) experiences a pong timeout and disconnects, Slack Socket Mode message processing stops. Slack messages are not acknowledged within the 3-second window, the health monitor detects a stale socket, and restarts the Slack provider. Buffered m... | open | null | false | 1 | [] | [] | 2026-04-09T02:10:40Z | 2026-05-03T10:19:10Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | tiburciod-prog | 243,896,883 | U_kgDODomSMw | User | false |
openclaw/openclaw | 4,228,919,798 | I_kwDOQb6kR878EDH2 | 63,492 | https://github.com/openclaw/openclaw/issues/63492 | https://api.github.com/repos/openclaw/openclaw/issues/63492 | [Feature]: Support external unified session storage for stateless session mode | ### Summary
Introduce a pluggable session storage backend (Redis, MongoDB, etc.) to enable stateless session sharing across multiple Openclaw instances, eliminating session stickiness and improving horizontal scalability.
### Problem to solve
Currently Openclaw stores sessions only in the local filesystem. In multi‑... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-04-09T02:18:06Z | 2026-05-02T11:36:12Z | null | NONE | null | 20260502T120029Z | 2026-05-02T12:00:29Z | bianchengtuzi | 6,304,737 | MDQ6VXNlcjYzMDQ3Mzc= | User | false |
openclaw/openclaw | 4,229,599,065 | I_kwDOQb6kR878Go9Z | 63,558 | https://github.com/openclaw/openclaw/issues/63558 | https://api.github.com/repos/openclaw/openclaw/issues/63558 | Control UI: Dreaming tab has no agent selector — cannot switch between agent contexts | ## Bug Description
The Control UI has a "Dreaming" tab (introduced in v2026.4.9). This tab shows:
- A diary view
- Dreaming facts/memories
- Backfill controls and promotion hints
**However, there is no agent selector anywhere in the tab.** Unlike every other tab in the Control UI (Sessions, Channels, Cron, etc.), the... | open | null | false | 1 | [] | [] | 2026-04-09T05:14:30Z | 2026-05-02T19:15:46Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | ttomiczek | 265,786,127 | U_kgDOD9eTDw | User | false |
openclaw/openclaw | 4,229,973,496 | I_kwDOQb6kR878IEX4 | 63,591 | https://github.com/openclaw/openclaw/issues/63591 | https://api.github.com/repos/openclaw/openclaw/issues/63591 | ACP dispatch setting blocks manual sessions_spawn | Problem: failed ACP sessions can leave channel topics noisy or effectively hijacked by background tasks. When attempting to prevent this by setting `acp.dispatch.enabled=false`, it successfully stops auto-binding but also unexpectedly blocks all manual `sessions_spawn(runtime="acp")` calls with the error: "ACP dispatch... | closed | completed | true | 2 | [] | [] | 2026-04-09T06:33:14Z | 2026-04-30T04:33:54Z | 2026-04-27T13:40:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | moeedahmed | 5,780,040 | MDQ6VXNlcjU3ODAwNDA= | User | false |
openclaw/openclaw | 4,230,337,239 | I_kwDOQb6kR878JdLX | 63,610 | https://github.com/openclaw/openclaw/issues/63610 | https://api.github.com/repos/openclaw/openclaw/issues/63610 | [Bug] WhatsApp channel not displaying tool results - shows raw tool call syntax | **Description:** When model calls tools (like `web_search`), the raw tool call syntax is displayed to user instead of actual tool results.
**Observed behavior:**
- Model calls `web_search` → Tool executes successfully → Returns real data from Tavily
- But user sees: `[TOOL_CALL]{tool => "web_search", args => {...}}[/T... | closed | completed | true | 5 | [] | [] | 2026-04-09T07:30:40Z | 2026-05-04T04:34:44Z | 2026-05-02T01:38:46Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | canh0chua | 252,370,775 | U_kgDODwrfVw | User | false |
openclaw/openclaw | 4,230,927,724 | I_kwDOQb6kR878LtVs | 63,652 | https://github.com/openclaw/openclaw/issues/63652 | https://api.github.com/repos/openclaw/openclaw/issues/63652 | memory status reports embeddings unavailable even when qmd status is healthy | ## Summary
`openclaw memory status --deep` reports `Embeddings: unavailable` with `Could not determine QMD vector status from qmd status`, even though the QMD index and vectors are healthy. This makes the memory/dreaming health output misleading.
## Repro
1. Configure memory to use QMD and enable dreaming.
2. Ensure... | closed | completed | true | 2 | [] | [] | 2026-04-09T09:02:42Z | 2026-04-30T04:33:52Z | 2026-04-27T13:57:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,230,976,508 | I_kwDOQb6kR878L5P8 | 63,655 | https://github.com/openclaw/openclaw/issues/63655 | https://api.github.com/repos/openclaw/openclaw/issues/63655 | memory-lancedb: OpenAI SDK base64 encoding breaks non-OpenAI embedding providers | ## Bug Description
The `memory-lancedb` plugin fails with dimension mismatch errors when using non-OpenAI embedding providers (e.g., ZhiPu AI `embedding-3`) via the `embedding.baseUrl` configuration.
**Root cause:** The bundled OpenAI SDK in `memory-lancedb` defaults `encoding_format` to `"base64"` when the user does... | closed | completed | true | 3 | [] | [] | 2026-04-09T09:10:14Z | 2026-04-30T04:32:45Z | 2026-04-28T00:13:04Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kinthaiofficial | 270,042,431 | U_kgDOEBiFPw | User | false |
openclaw/openclaw | 4,231,219,602 | I_kwDOQb6kR878M0mS | 63,673 | https://github.com/openclaw/openclaw/issues/63673 | https://api.github.com/repos/openclaw/openclaw/issues/63673 | [Bug]: Keychat Bridge receives no inbound messages after OpenClaw update to 2026.4.8 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After updating OpenClaw from 2026.3.28 to 2026.4.8, the Keychat bridge can **send** messages but **cannot receive** any inbound messages. The bridge connects to relays successfully and outgoing messages are delivered, but no... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-04-09T09:49:48Z | 2026-05-03T02:14:47Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | walter270 | 274,765,493 | U_kgDOEGCWtQ | User | false |
openclaw/openclaw | 4,231,375,019 | I_kwDOQb6kR878Nair | 63,685 | https://github.com/openclaw/openclaw/issues/63685 | https://api.github.com/repos/openclaw/openclaw/issues/63685 | [Bug]: Cannot run gemma 4 models from google ai studio | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
I want to use gemma 4 models from google ai studio but it return me error.
### Steps to reproduce
1. install your openclaw
2. config it to use gemma 4 from google ai studio with api key
3. try the telegram bot.
### Expecte... | open | null | false | 3 | [
"bug"
] | [] | 2026-04-09T10:17:06Z | 2026-05-03T12:30:59Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | behroozbc | 20,381,616 | MDQ6VXNlcjIwMzgxNjE2 | User | false |
openclaw/openclaw | 4,232,181,900 | I_kwDOQb6kR878QfiM | 63,734 | https://github.com/openclaw/openclaw/issues/63734 | https://api.github.com/repos/openclaw/openclaw/issues/63734 | Feature Request: Built-in Command Risk Classifier for exec approvals | ## Summary
Add a built-in command risk classification system to OpenClaw's exec approval mechanism, similar to Claude Code's command classification.
## Motivation
Currently, OpenClaw's exec approval system treats all commands equally - every command triggers an approval dialog, even safe read-only commands like ls, ... | open | null | false | 3 | [] | [] | 2026-04-09T12:27:55Z | 2026-05-03T08:39:08Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | Tangand | 273,061,083 | U_kgDOEEaU2w | User | false |
openclaw/openclaw | 4,232,642,953 | I_kwDOQb6kR878SQGJ | 63,757 | https://github.com/openclaw/openclaw/issues/63757 | https://api.github.com/repos/openclaw/openclaw/issues/63757 | [Bug]: Telegram reactions are received but not exposed as a stable structured hook for downstream feedback | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
On OpenClaw 2026.4.8, Telegram reaction updates are received by the gateway, but there is no stable structured internal hook/event path that downstream integrations can consume without patching core, so reactio... | open | null | false | 1 | [] | [] | 2026-04-09T13:37:52Z | 2026-05-03T06:07:15Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | alessandropcostabr | 191,734,104 | U_kgDOC22hWA | User | false |
openclaw/openclaw | 4,232,643,019 | I_kwDOQb6kR878SQHL | 63,758 | https://github.com/openclaw/openclaw/issues/63758 | https://api.github.com/repos/openclaw/openclaw/issues/63758 | [Bug]: Managed cron task runs do not populate progressSummary, leaving /tasks and session_status mostly silent during execution | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The background-task ledger supports `progressSummary`, but on OpenClaw 2026.4.8 managed cron jobs do not populate it, so `/tasks` and `session_status` remain mostly silent until the job finishes even when the t... | open | null | false | 1 | [] | [] | 2026-04-09T13:37:52Z | 2026-05-03T08:39:10Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | alessandropcostabr | 191,734,104 | U_kgDOC22hWA | User | false |
openclaw/openclaw | 4,232,646,583 | I_kwDOQb6kR878SQ-3 | 63,760 | https://github.com/openclaw/openclaw/issues/63760 | https://api.github.com/repos/openclaw/openclaw/issues/63760 | Announce runs for subagent/ACP completions waste tokens on cold cache with parent model | ## Problem
When a subagent or ACP session (e.g. Claude Code) completes, OpenClaw runs an "announce" embedded run to deliver the result to the parent agent. This announce run:
1. **Uses the parent agent's model** — e.g. Opus at $25/MTok output for what is essentially a message relay
2. **Starts with a cold cache** — l... | open | null | false | 1 | [] | [] | 2026-04-09T13:38:32Z | 2026-05-03T13:21:17Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | liu51115 | 31,350,941 | MDQ6VXNlcjMxMzUwOTQx | User | false |
openclaw/openclaw | 4,232,717,737 | I_kwDOQb6kR878SiWp | 63,769 | https://github.com/openclaw/openclaw/issues/63769 | https://api.github.com/repos/openclaw/openclaw/issues/63769 | [Bug] Gateway WebSocket chat stream may drop repeated consecutive characters in merged assistant text | ## Bug summary
When consuming OpenClaw Gateway events over WebSocket, `agent` stream payloads are correct, but subsequent `chat` delta/final payloads may drop repeated consecutive characters (digit/CJK), e.g.:
- `k0059` -> `k059`
- `k1119` -> `k19`
- `k0044` -> `k04`
- `超时时间` -> `超时间`
## Observed behavior
In logs, ... | closed | completed | true | 2 | [] | [] | 2026-04-09T13:49:35Z | 2026-04-30T04:33:08Z | 2026-04-27T20:35:58Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yon950905 | 40,497,654 | MDQ6VXNlcjQwNDk3NjU0 | User | false |
openclaw/openclaw | 4,233,619,536 | I_kwDOQb6kR878V-hQ | 63,845 | https://github.com/openclaw/openclaw/issues/63845 | https://api.github.com/repos/openclaw/openclaw/issues/63845 | Tool granularity vs server-side aggregation — best practices for high-count skills / tools (65+)? | ## Context
We're using OpenClaw 4.8 at a fintech company to build an AI investment-research assistant. Our current setup:
- **60+ structured REST endpoints** exposed through an internal data service wrapping multiple upstream providers — covering quotes, fundamentals, financials, K-line, limit-up lists, major shareho... | open | null | false | 1 | [] | [] | 2026-04-09T16:32:09Z | 2026-05-03T08:39:12Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | gucasbrg | 3,182,658 | MDQ6VXNlcjMxODI2NTg= | User | false |
openclaw/openclaw | 4,233,698,306 | I_kwDOQb6kR878WRwC | 63,855 | https://github.com/openclaw/openclaw/issues/63855 | https://api.github.com/repos/openclaw/openclaw/issues/63855 | WhatsApp: stale socket causes silent message loss (connection appears live but drops inbound messages) | ## Summary
WhatsApp connections enter a **stale socket state** where the gateway reports both accounts as "Listening" but inbound messages are silently swallowed — never surfaced to the agent. The socket appears healthy (no disconnect event) but is functionally dead.
This is **message loss**, not cosmetic reconnect n... | closed | completed | true | 2 | [] | [] | 2026-04-09T16:48:20Z | 2026-05-02T04:28:30Z | 2026-04-29T10:55:13Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | legonhilltech-jpg | 272,396,537 | U_kgDOEDxw-Q | User | false |
openclaw/openclaw | 4,234,199,107 | I_kwDOQb6kR878YMBD | 63,884 | https://github.com/openclaw/openclaw/issues/63884 | https://api.github.com/repos/openclaw/openclaw/issues/63884 | [Feature]: Per-bot context isolation in multi-bot Feishu groups - only record context when mentioned | ### Summary
Add a configuration option to isolate message context per bot in Feishu groups, ensuring each bot only records messages to its context when explicitly @mentioned, preventing context pollution from other bots' conversations.
### Problem to solve
当前 OpenClaw 的飞书插件在同一个群里部署多个机器人时存在上下文污染问题：
场景：一个飞书群里有两个 OpenC... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-09T18:32:44Z | 2026-05-03T12:31:00Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | kechangrong | 13,132,516 | MDQ6VXNlcjEzMTMyNTE2 | User | false |
openclaw/openclaw | 4,234,462,303 | I_kwDOQb6kR878ZMRf | 63,901 | https://github.com/openclaw/openclaw/issues/63901 | https://api.github.com/repos/openclaw/openclaw/issues/63901 | Feature: Pass request metadata to API providers (trace/user fields) | ## Problem
When multiple agents/tasks share the same provider account (e.g., OpenRouter), there's no way to attribute API costs to specific graphs, tasks, or agents. Each API call is anonymous beyond the model name.
## Use Case
We run a graph-based orchestration system with ~10 agent types (forge, forge-plan, atlas,... | open | null | false | 2 | [] | [] | 2026-04-09T19:27:37Z | 2026-05-03T12:31:02Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | JanPlessow | 202,702,070 | U_kgDODBT89g | User | false |
openclaw/openclaw | 4,234,492,394 | I_kwDOQb6kR878ZTnq | 63,904 | https://github.com/openclaw/openclaw/issues/63904 | https://api.github.com/repos/openclaw/openclaw/issues/63904 | Support reply_broadcast for Slack thread replies (inbound detection + outbound mirroring) | ## Feature Request: Support `reply_broadcast` for Slack thread replies
### Problem
When replying in Slack threads, OpenClaw cannot mirror the "Also send to channel" behavior. The `postSlackMessageBestEffort` function in `send-pJvSRc2i.js` builds the payload with `thread_ts` but never includes `reply_broadcast`. The `... | open | null | false | 1 | [] | [] | 2026-04-09T19:34:18Z | 2026-05-03T09:12:35Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | becksbert | 274,104,530 | U_kgDOEFaA0g | User | false |
openclaw/openclaw | 4,332,708,371 | I_kwDOQb6kR88AAAABAj_iEw | 72,514 | https://github.com/openclaw/openclaw/issues/72514 | https://api.github.com/repos/openclaw/openclaw/issues/72514 | [Feature]: OpenClaw 跨会话/跨 Agent 互通问题 | ### Summary
feedback_openclaw_cross_session_issues.md
OpenClaw 跨会话/跨 Agent 互通问题 — 用户反馈
提交者：QClaw 用户（华为安卓 + Windows 10 桌面端）
日期：2026-04-27
版本：OpenClaw + QClaw 桌面版
一、问题概述
在使用 QClaw 的过程中，发现**跨会话（webchat ↔ 微信）和跨 Agent（主 Agent ↔ 子 Agent）**的场景下，存在严重的记忆与任务上下文隔离问题，导致：
电脑和手机之间的任务无法无缝衔接
主 Agent 与子 Agent 之间几乎"失联"，复杂协作任务频繁失败
上下文... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-27T02:21:30Z | 2026-05-03T10:41:32Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | bianliang0706-ux | 266,259,596 | U_kgDOD97MjA | User | false |
openclaw/openclaw | 4,234,838,735 | I_kwDOQb6kR878aoLP | 63,920 | https://github.com/openclaw/openclaw/issues/63920 | https://api.github.com/repos/openclaw/openclaw/issues/63920 | Expose Slack message_action and global shortcut events via channels.slack | ## Context
Running OpenClaw v2026.4.9 (pin `0512059d`) with `channels.slack` in Socket Mode and a set of custom plugins. The Slack monitor currently routes three Slack interaction event families from Bolt into the OpenClaw agent runtime via `enqueueSystemEvent`:
- **`block_actions`** — wired in [`extensions/slack/src... | open | null | false | 1 | [] | [] | 2026-04-09T20:50:29Z | 2026-05-03T01:37:23Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | chrisbaker2000 | 174,035,702 | U_kgDOCl-S9g | User | false |
openclaw/openclaw | 4,235,327,127 | I_kwDOQb6kR878cfaX | 63,956 | https://github.com/openclaw/openclaw/issues/63956 | https://api.github.com/repos/openclaw/openclaw/issues/63956 | [Bug]: Streaming response creates multiple chat bubbles before collapsing into one | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When the agent streams a response in the Control UI webchat, multiple separate chat bubbles appear during streaming, then collapse/merge into a single bubble when the response completes. This creates a jarring ... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-09T22:45:06Z | 2026-05-02T20:57:14Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | contentfree | 10,958 | MDQ6VXNlcjEwOTU4 | User | false |
openclaw/openclaw | 4,236,601,890 | I_kwDOQb6kR878hWoi | 64,081 | https://github.com/openclaw/openclaw/issues/64081 | https://api.github.com/repos/openclaw/openclaw/issues/64081 | Feature: config-level default task notify policy | ## Problem
Background task completion notifications (subagent and ACP) default to `done_only` policy, which sends a delivery message to the requester channel (e.g. Telegram) for every task completion.
When running multiple sub-agents or ACP sessions, this creates notification spam with messages like:
```
Background ... | open | null | false | 2 | [] | [] | 2026-04-10T04:37:50Z | 2026-05-03T12:31:04Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | ancxlol | 209,537,080 | U_kgDODH1IOA | User | false |
openclaw/openclaw | 4,236,633,774 | I_kwDOQb6kR878heau | 64,086 | https://github.com/openclaw/openclaw/issues/64086 | https://api.github.com/repos/openclaw/openclaw/issues/64086 | feat: Plugin API for TUI footer status contributions (registerStatusProvider) | ## Feature request
Allow context engine plugins to contribute status information to the TUI footer bar.
### Current state
The TUI footer displays: agent name, session, model, token count. There's no way for plugins (especially context engine plugins) to add their own status indicators.
Currently, the only way to sh... | open | null | false | 2 | [] | [] | 2026-04-10T04:45:55Z | 2026-05-03T12:31:05Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | dodge1218 | 235,199,066 | U_kgDODgTaWg | User | false |
openclaw/openclaw | 4,236,773,210 | I_kwDOQb6kR878iAda | 64,103 | https://github.com/openclaw/openclaw/issues/64103 | https://api.github.com/repos/openclaw/openclaw/issues/64103 | Session `status` field values ("failed", "timeout", "done") mislead agents into spawning duplicate sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The session `status` field uses values like `"failed"`, `"timeout"`, and `"done"` that only reflect the last communication turn's state, but these names strongly imply the session or task has permanently ended.... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-10T05:19:20Z | 2026-05-02T13:05:33Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | lykeion-dev | 140,231,077 | U_kgDOCFvBpQ | User | false |
openclaw/openclaw | 4,236,892,137 | I_kwDOQb6kR878idfp | 64,121 | https://github.com/openclaw/openclaw/issues/64121 | https://api.github.com/repos/openclaw/openclaw/issues/64121 | [Feature]: Split chat session picker into Agent + Agent Sessions | ### Summary
Split the combined chat session picker into separate Agent and Agent Session controls, while keeping Model and Thinking bound to the active session.
### Problem to solve
Problem to solve
The current chat header uses one combined session picker for everything. Once an agent accumulates many sessions, espe... | closed | completed | false | 2 | [
"enhancement"
] | [] | 2026-04-10T05:45:16Z | 2026-05-04T12:51:10Z | 2026-05-04T12:51:10Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | pavelknezik | 274,213,128 | U_kgDOEFgpCA | User | false |
openclaw/openclaw | 4,236,931,900 | I_kwDOQb6kR878inM8 | 64,129 | https://github.com/openclaw/openclaw/issues/64129 | https://api.github.com/repos/openclaw/openclaw/issues/64129 | [Bug]: Configuring a paid Google Gemini key implicitly turned existing heartbeat traffic into paid background usage | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Configuring a paid Google Gemini API key switched the default model to Gemini and caused the existing main-workspace heartbeat to start consuming paid Gemini API in the background before any intentional user ch... | open | null | false | 1 | [
"bug"
] | [] | 2026-04-10T05:54:22Z | 2026-05-03T01:37:26Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | vegapunk9527 | 272,845,845 | U_kgDOEENMFQ | User | false |
openclaw/openclaw | 4,237,375,246 | I_kwDOQb6kR878kTcO | 64,168 | https://github.com/openclaw/openclaw/issues/64168 | https://api.github.com/repos/openclaw/openclaw/issues/64168 | [Bug]: Self-improvement note leaked to user on /new session after upgrading to 2026.4.9 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from 2026.4.x to 2026.4.9, typing /new in Telegram shows self-improvement note to user instead of starting a clean session.
### Steps to reproduce
1. Start a Telegram DM with the OpenClaw agent.
2. Type /ne... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-10T07:26:03Z | 2026-05-01T04:44:21Z | 2026-04-28T08:57:18Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jisi-assist | 262,780,846 | U_kgDOD6m3rg | User | false |
openclaw/openclaw | 4,237,813,883 | I_kwDOQb6kR878l-h7 | 64,199 | https://github.com/openclaw/openclaw/issues/64199 | https://api.github.com/repos/openclaw/openclaw/issues/64199 | [Bug]: ACP configured binding uses parent channel ID for session key — all threads under same channel share one persistent Claude Code process | ## Bug
When an agent uses `runtime.type: \"acp\"` with `mode: \"persistent\"` and a Discord channel binding, **all Discord threads spawned from the same parent channel share a single persistent Claude Code process**. This causes cross-thread context contamination — thread A's agent is aware of thread B's conversation ... | open | null | false | 2 | [] | [] | 2026-04-10T08:38:38Z | 2026-05-03T12:31:09Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | hoh-dev-bot | 219,382,756 | U_kgDODROD5A | User | false |
openclaw/openclaw | 4,237,830,821 | I_kwDOQb6kR878mCql | 64,201 | https://github.com/openclaw/openclaw/issues/64201 | https://api.github.com/repos/openclaw/openclaw/issues/64201 | [Bug]: OpenClaw: Crash loop on plugin config reload (ECONNREFUSED on loopback port 18789) | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
OpenClaw enters a crash loop for ~6 minutes (≈25 restarts at 15s intervals)
whenever the `plugins.allow` or `plugins.entries.*` configuration is changed
at runtime. The container self-recovers thanks to `restart: unless-sto... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-10T08:41:18Z | 2026-04-29T04:32:07Z | 2026-04-27T03:50:36Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jpdeuster | 32,199,147 | MDQ6VXNlcjMyMTk5MTQ3 | User | false |
openclaw/openclaw | 4,237,906,542 | I_kwDOQb6kR878mVJu | 64,205 | https://github.com/openclaw/openclaw/issues/64205 | https://api.github.com/repos/openclaw/openclaw/issues/64205 | Agent should be notified when MCP server fails to start | ## Problem
When a configured MCP server fails to start (e.g. npm package removed from registry), the error is logged at gateway level but **the agent is never informed**. The agent simply receives a reduced tool set without knowing tools are missing.
This leads to a terrible user experience: the agent "pretends" to w... | open | null | false | 1 | [] | [] | 2026-04-10T08:54:13Z | 2026-05-03T08:39:15Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | SvenPB | 269,757,202 | U_kgDOEBQrEg | User | false |
openclaw/openclaw | 4,237,970,554 | I_kwDOQb6kR878mkx6 | 64,212 | https://github.com/openclaw/openclaw/issues/64212 | https://api.github.com/repos/openclaw/openclaw/issues/64212 | [Bug]: Image tool fails with "Request was aborted" for NVIDIA Kimi K2.5 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using custom provider for NVIDIA Kimi K2.5, the image tool fails with "Request was aborted". Direct API calls work fine.
### Steps to reproduce
1. Configure custom provider for NVIDIA Kimi K2.5 with inp... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-10T09:07:28Z | 2026-05-03T12:31:10Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | Iderty | 264,356,664 | U_kgDOD8HDOA | User | false |
openclaw/openclaw | 4,332,715,035 | I_kwDOQb6kR88AAAABAj_8Gw | 72,516 | https://github.com/openclaw/openclaw/issues/72516 | https://api.github.com/repos/openclaw/openclaw/issues/72516 | [Bug]: Control UI Chat shows default agent identity for all agent sessions (race condition) | ### Bug type
Regression / Race condition
### Summary
When navigating to a non-default agent session in Control UI Chat (e.g. `?session=agent%3Alottery%3Amain`), the message bubble sidebar always shows the **default agent** name and emoji instead of the correct agent identity.
**Example:**
- URL: `http://127.0.0.1:1... | closed | completed | true | 2 | [] | [] | 2026-04-27T02:24:08Z | 2026-04-30T04:34:26Z | 2026-04-27T11:45:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | antnft | 275,393,373 | U_kgDOEGorXQ | User | false |
openclaw/openclaw | 4,207,322,344 | I_kwDOQb6kR876xqTo | 61,320 | https://github.com/openclaw/openclaw/issues/61320 | https://api.github.com/repos/openclaw/openclaw/issues/61320 | [Bug]: Moonshot/Kimi API returns 400: invalid scalar type [object boolean] | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
OpenClaw fails to call Moonshot (Kimi) API with HTTP 400 error: "invalid scalar type [object boolean]". The same API key works fine with direct curl requests.
### Steps to reproduce
1. Configure moonshot/kimi-k2.5 as de... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T11:44:16Z | 2026-04-29T04:32:37Z | 2026-04-27T02:24:46Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | unrealfantacy | 138,662,677 | U_kgDOCEPTFQ | User | false |
openclaw/openclaw | 4,225,263,910 | I_kwDOQb6kR8772Gkm | 63,187 | https://github.com/openclaw/openclaw/issues/63187 | https://api.github.com/repos/openclaw/openclaw/issues/63187 | [Feature]: 需要能够获取模型推理的TTFT | ### Summary
同标题所言，目前通过openclaw可以获取模型推理的token in、token out、cache token等数据，但无法获取TTFT，希望添加相关支持。
### Problem to solve
解决log不全问题。
### Proposed solution
在pi-agent流式输出事件中增加首chunk接收时间戳，作差获得TTFT
### Alternatives considered
_No response_
### Impact
Affected systems
Severity annoying
Frequency always
Consequence extra m... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-08T14:23:57Z | 2026-04-29T04:32:35Z | 2026-04-27T02:25:04Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | zzHezzo | 100,818,680 | U_kgDOBgJe-A | User | false |
openclaw/openclaw | 4,225,775,275 | I_kwDOQb6kR8774Dar | 63,227 | https://github.com/openclaw/openclaw/issues/63227 | https://api.github.com/repos/openclaw/openclaw/issues/63227 | Windows: CLI commands hang for 30-90s after output completes (event loop not exiting) | ## Bug Description
On Windows 11, all openclaw CLI commands that go through the full initialization path hang for 30-90 seconds **after** all output has been printed, before the process finally exits. The actual command execution completes in under 1 second.
## Environment
- **OS**: Windows 11 Pro 10.0.26300 (x64)
-... | closed | completed | true | 1 | [] | [] | 2026-04-08T15:42:50Z | 2026-04-29T04:32:34Z | 2026-04-27T02:25:09Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | 863108316 | 127,532,543 | U_kgDOB5n9_w | User | false |
openclaw/openclaw | 4,226,219,315 | I_kwDOQb6kR8775v0z | 63,260 | https://github.com/openclaw/openclaw/issues/63260 | https://api.github.com/repos/openclaw/openclaw/issues/63260 | Add `showPrompt` to heartbeat visibility config | # Feature Requests & Issues
## OpenClaw Gateway: Add `showPrompt` to heartbeat visibility config
**Status:** Documented for upstream submission
**Date:** 2026-04-08
**Reporter:** Bill (via Karl)
### Problem
The heartbeat system prompt ("Read HEARTBEAT.md if it exists...") is always displayed in Control UI chat, ... | closed | completed | true | 1 | [] | [] | 2026-04-08T17:01:23Z | 2026-04-29T04:32:34Z | 2026-04-27T02:25:14Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | winslowb | 71,040,996 | MDQ6VXNlcjcxMDQwOTk2 | User | false |
openclaw/openclaw | 4,232,758,993 | I_kwDOQb6kR878SsbR | 63,773 | https://github.com/openclaw/openclaw/issues/63773 | https://api.github.com/repos/openclaw/openclaw/issues/63773 | Bug: Subagent session JSONL does not record token usage (usage fields always 0) | # OpenClaw Issue: Subagent session JSONL does not record token usage
## Summary
Subagent session JSONL files record `usage` fields as all zeros (`inputTokens: 0, outputTokens: 0, cacheReadInputTokens: 0`) even though the LLM API returned valid token counts and the assistant message `content` contains actual output te... | closed | completed | true | 1 | [] | [] | 2026-04-09T13:56:05Z | 2026-04-29T04:32:32Z | 2026-04-27T02:25:50Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | mfkanb | 4,747,373 | MDQ6VXNlcjQ3NDczNzM= | User | false |
openclaw/openclaw | 4,332,599,125 | I_kwDOQb6kR88AAAABAj43VQ | 72,488 | https://github.com/openclaw/openclaw/issues/72488 | https://api.github.com/repos/openclaw/openclaw/issues/72488 | [Bug]: plugin service failed | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
更新至v2026.4.24后，出现
<img width="1709" height="422" alt="Image" src="https://github.com/user-attachments/assets/e2735691-76f8-4182-a280-c94ddab44ea5" />
### Steps to reproduce
重启网关/gateway后仍会出现
### Expected behavior
.
###... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-04-27T01:42:16Z | 2026-04-30T04:36:05Z | 2026-04-27T07:24:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | xiao4d-lyf | 247,534,011 | U_kgDODsERuw | User | false |
openclaw/openclaw | 4,332,736,520 | I_kwDOQb6kR88AAAABAkBQCA | 72,521 | https://github.com/openclaw/openclaw/issues/72521 | https://api.github.com/repos/openclaw/openclaw/issues/72521 | [Bug] doctor false-misses bundled runtime deps and bundled channel loader can hit ENOTEMPTY on plugin-sdk alias | ## Summary
On OpenClaw `2026.4.24`, I hit two related bundled-runtime-deps problems on macOS:
1. `openclaw doctor` kept reporting `Bundled plugin runtime deps are missing` and `openclaw gateway restart` stayed slow because bundled runtime deps were repeatedly restaged.
2. During `doctor` / bundled channel loading, ch... | closed | completed | true | 1 | [] | [] | 2026-04-27T02:32:21Z | 2026-04-30T04:36:35Z | 2026-04-27T06:00:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lamkan0210 | 65,824,253 | MDQ6VXNlcjY1ODI0MjUz | User | false |
openclaw/openclaw | 4,131,775,554 | I_kwDOQb6kR872ReRC | 54,142 | https://github.com/openclaw/openclaw/issues/54142 | https://api.github.com/repos/openclaw/openclaw/issues/54142 | Several provider API fetch calls missing timeout — can hang indefinitely | ## Problem
Multiple provider integration modules make `fetch` calls without `AbortSignal.timeout()`, meaning they can hang indefinitely if the remote API is unresponsive.
### Affected Locations
| File | Line | Endpoint | Has Timeout |
|------|------|----------|-------------|
| `src/plugins/provider-ollama-setup.ts` ... | closed | completed | true | 2 | [] | [] | 2026-03-25T01:36:58Z | 2026-04-29T04:32:19Z | 2026-04-27T02:34:31Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,332,742,867 | I_kwDOQb6kR88AAAABAkBo0w | 72,523 | https://github.com/openclaw/openclaw/issues/72523 | https://api.github.com/repos/openclaw/openclaw/issues/72523 | Google Meet realtime should clear playback on Gemini Live interruptions | ## Summary
Google Meet realtime sessions can resume stale assistant audio after Gemini Live reports an interruption. The provider already emits `onClearAudio` when `serverContent.interrupted` is true, and the shared realtime runtime forwards that to `audioSink.clearAudio`, but the Google Meet Chrome command-pair sinks ... | closed | completed | true | 0 | [] | [] | 2026-04-27T02:34:47Z | 2026-04-30T04:36:16Z | 2026-04-27T06:49:11Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | BsnizND | 199,837,910 | U_kgDOC-lI1g | User | false |
openclaw/openclaw | 4,332,743,811 | I_kwDOQb6kR88AAAABAkBsgw | 72,525 | https://github.com/openclaw/openclaw/issues/72525 | https://api.github.com/repos/openclaw/openclaw/issues/72525 | Google Meet Chrome realtime bridge should support higher-fidelity PCM audio | ## Summary
The Google Meet Chrome realtime command-pair path is documented and implemented as 8 kHz G.711 mu-law in both directions. That makes the assistant voice sound telephone-band even when Gemini Live is producing higher-rate native audio and Chrome/Meet/CoreAudio can handle 48 kHz device paths.
## Evidence
- Op... | closed | completed | true | 1 | [] | [] | 2026-04-27T02:35:10Z | 2026-04-30T04:34:18Z | 2026-04-27T11:55:17Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | BsnizND | 199,837,910 | U_kgDOC-lI1g | User | false |
openclaw/openclaw | 4,113,056,622 | I_kwDOQb6kR871KENu | 51,767 | https://github.com/openclaw/openclaw/issues/51767 | https://api.github.com/repos/openclaw/openclaw/issues/51767 | sessionRetention does not apply to subagent sessions — only cron-run sessions are swept | ## Summary
`cron.sessionRetention` (e.g. `"3h"`) only prunes sessions matching `isCronRunSessionKey()` (pattern `cron:*:run:*`). Subagent sessions (`agent:*:subagent:*`) are **never cleaned up** by the retention sweep, even after they complete.
## Impact
After any sprint with 10-30+ subagent spawns, `sessions.json` ... | open | null | false | 5 | [] | [] | 2026-03-21T17:01:49Z | 2026-05-04T20:23:15Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | 100yenadmin | 239,388,517 | U_kgDODkTHZQ | User | false |
openclaw/openclaw | 4,116,342,398 | I_kwDOQb6kR871WmZ- | 52,327 | https://github.com/openclaw/openclaw/issues/52327 | https://api.github.com/repos/openclaw/openclaw/issues/52327 | Feature: Persistent session retention for subagent/ACP sessions (analogous to cron.sessionRetention) | We hit a production incident caused by subagent/ACP session accumulation. The existing `archiveAfterMinutes` mechanism has gaps that allow completed sessions to persist indefinitely, eventually inflating `sessions.json` past the `rotateBytes` threshold and causing destructive rotation that wipes the active main session... | closed | duplicate | true | 2 | [] | [] | 2026-03-22T15:33:20Z | 2026-04-29T04:32:18Z | 2026-04-27T02:39:54Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | Madruru | 265,138,534 | U_kgDOD82xZg | User | false |
openclaw/openclaw | 4,332,758,821 | I_kwDOQb6kR88AAAABAkCnJQ | 72,527 | https://github.com/openclaw/openclaw/issues/72527 | https://api.github.com/repos/openclaw/openclaw/issues/72527 | [Bug]: Downgrading from 2026.4.x to 2026.3.2 leaves openclaw.json in invalid/broken state with no warning | ## Summary
When downgrading from OpenClaw 2026.4.x back to 2026.3.2, the `openclaw.json` written by the newer version contains keys and value formats that are invalid in 2026.3.2. The gateway starts with silent config errors that break Discord channel connectivity and LLM responses, with no actionable guidance.
## En... | open | null | false | 1 | [] | [] | 2026-04-27T02:40:44Z | 2026-05-03T08:39:40Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | Schlaflied | 187,913,759 | U_kgDOCzNWHw | User | false |
openclaw/openclaw | 4,332,760,876 | I_kwDOQb6kR88AAAABAkCvLA | 72,528 | https://github.com/openclaw/openclaw/issues/72528 | https://api.github.com/repos/openclaw/openclaw/issues/72528 | Plugin register fails with ENOTEMPTY in openclaw-unknown-* shadow runtime-deps dir on every gateway boot | ## Summary
On every `ai.openclaw.gateway` launch, the plugin loader fails to register the bundled Telegram plugin (and several other channels) with `ENOTEMPTY`. Two related directories live under `~/.openclaw/plugin-runtime-deps/`:
- `openclaw-2026.4.24-02f98c7ef353/` — the expected versioned dir
- `openclaw-unknown-... | closed | completed | true | 2 | [] | [] | 2026-04-27T02:41:21Z | 2026-04-30T04:36:35Z | 2026-04-27T06:00:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JanPlessow | 202,702,070 | U_kgDODBT89g | User | false |
openclaw/openclaw | 4,098,571,232 | I_kwDOQb6kR870Szvg | 50,103 | https://github.com/openclaw/openclaw/issues/50103 | https://api.github.com/repos/openclaw/openclaw/issues/50103 | [Feature]: Add agents.defaults.systemPrompt for config-injectable rules | ### Summary
Add support for injecting agent rules/system prompts from `openclaw.json` instead of (or in addition to) bootstrap files like AGENTS.md.
### Problem to solve
Currently, safety rules and behavioral guidelines live in workspace files (AGENTS.md, SOUL.md, etc.). This works but:
- Rules are tied to the works... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-19T01:02:35Z | 2026-05-04T20:14:36Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | aaronyyleon-beep | 252,489,375 | U_kgDODwyunw | User | false |
openclaw/openclaw | 4,332,675,642 | I_kwDOQb6kR88AAAABAj9iOg | 72,506 | https://github.com/openclaw/openclaw/issues/72506 | https://api.github.com/repos/openclaw/openclaw/issues/72506 | BlueBubbles native iOS voice-memo delivery broken end-to-end with ElevenLabs (and other non-Azure TTS providers) | ## Summary
Sending TTS audio to a BlueBubbles iMessage chat using the bundled `tts` agent tool (or `tts.convert` RPC) currently always renders as a **plain audio attachment** in iMessage, never as a **native iOS voice memo** (the bubble with the waveform / scrubber UI). Two distinct upstream gaps in the same pipeline ... | closed | completed | true | 3 | [
"maintainer"
] | [
"omarshahine"
] | 2026-04-27T02:09:22Z | 2026-04-29T04:32:12Z | 2026-04-27T03:34:09Z | CONTRIBUTOR | null | 20260429T060025Z | 2026-04-29T06:00:25Z | omarshahine | 10,343,873 | MDQ6VXNlcjEwMzQzODcz | User | false |
openclaw/openclaw | 4,112,324,344 | I_kwDOQb6kR871HRb4 | 51,598 | https://github.com/openclaw/openclaw/issues/51598 | https://api.github.com/repos/openclaw/openclaw/issues/51598 | Feature: agent-level systemPrompt config (agents.defaults.systemPrompt / agents.<id>.systemPrompt) | ## Problem
Currently, custom system prompt text can only be injected at the **channel level** (e.g. `channels.discord.guilds.<id>.channels.<id>.systemPrompt`). There is no way to inject custom text into the **actual system prompt** at the agent level.
Workspace `.md` files (BOOTSTRAP.md, AGENTS.md, etc.) are injected... | closed | duplicate | true | 2 | [] | [] | 2026-03-21T12:26:20Z | 2026-04-29T04:32:17Z | 2026-04-27T02:46:15Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | david6ca | 45,101,457 | MDQ6VXNlcjQ1MTAxNDU3 | User | false |
openclaw/openclaw | 4,332,774,108 | I_kwDOQb6kR88AAAABAkDi3A | 72,529 | https://github.com/openclaw/openclaw/issues/72529 | https://api.github.com/repos/openclaw/openclaw/issues/72529 | Feature Request: Allow cron jobs to route to subagent persistent sessions | ## Summary
Currently, cron jobs only support `sessionTarget: main` or `sessionTarget: isolated`. There is no way to route a cron job to an existing subagent persistent session (created via `sessions_spawn` with `thread: true`).
## Use Case
In a multi-agent setup, we have specialized sub-agents (e.g., a fandom news c... | closed | completed | true | 2 | [] | [] | 2026-04-27T02:46:25Z | 2026-05-03T04:34:00Z | 2026-05-01T04:15:07Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | sylviazhang2006-design | 279,586,936 | U_kgDOEKooeA | User | false |
openclaw/openclaw | 4,077,475,187 | I_kwDOQb6kR87zCVVz | 46,812 | https://github.com/openclaw/openclaw/issues/46812 | https://api.github.com/repos/openclaw/openclaw/issues/46812 | Control UI: Pinned context panel for user-defined persistent facts | ## Feature Request
### Summary
Add a pinned context panel to the Control UI that lets users surface and edit key facts that are always injected into the agent's context.
### Motivation
Users currently manage persistent context by editing MEMORY.md and USER.md directly. There is no lightweight way to pin a quick fact ... | open | null | false | 3 | [] | [] | 2026-03-15T03:26:08Z | 2026-05-04T00:44:44Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | allisonshawai-ctrl | 262,603,479 | U_kgDOD6cC1w | User | false |
openclaw/openclaw | 4,115,107,979 | I_kwDOQb6kR871R5CL | 52,115 | https://github.com/openclaw/openclaw/issues/52115 | https://api.github.com/repos/openclaw/openclaw/issues/52115 | [Feature]: memory_search returns stale/invalid results when gateway is long-running | ### Summary
## Summary
The `memory_search` tool (invoked via the agent/tool chain, not CLI) has two related bugs when the gateway process has been running for a while:
1. **Stale index** — After workspace files are created, renamed, or deleted, `memory_search` continues returning results based on the old index state... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-22T06:25:19Z | 2026-05-04T20:16:04Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | FicheallADa | 270,066,732 | U_kgDOEBjkLA | User | false |
openclaw/openclaw | 4,120,748,139 | I_kwDOQb6kR871naBr | 52,861 | https://github.com/openclaw/openclaw/issues/52861 | https://api.github.com/repos/openclaw/openclaw/issues/52861 | perf: reduce per-turn token overhead via conditional tool/skill injection | ## Problem
OpenClaw's system prompt includes **all** registered tool schemas and skill descriptions on every turn, regardless of whether they're relevant to the current conversation. This creates significant per-turn token overhead.
### Measured impact (real-world example)
In a typical Feishu-connected session with ... | open | null | false | 1 | [] | [] | 2026-03-23T13:08:03Z | 2026-05-04T21:00:59Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | upsightx | 127,651,198 | U_kgDOB5vNfg | User | false |
openclaw/openclaw | 4,121,031,466 | I_kwDOQb6kR871ofMq | 52,894 | https://github.com/openclaw/openclaw/issues/52894 | https://api.github.com/repos/openclaw/openclaw/issues/52894 | [Feature]: Automatically install brew if not installed | ### Summary
<img width="1084" height="530" alt="Image" src="https://github.com/user-attachments/assets/efe4449a-4749-4759-9751-f8bd12f956e9" />
<img width="1078" height="571" alt="Image" src="https://github.com/user-attachments/assets/2ffc370d-f57c-459a-8ddb-c456439ddc49" />
The installer asks if you'd like to see t... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-23T13:46:51Z | 2026-05-01T21:59:26Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | RaphaelFakhri | 153,192,858 | U_kgDOCSGJmg | User | false |
openclaw/openclaw | 4,121,363,449 | I_kwDOQb6kR871pwP5 | 52,928 | https://github.com/openclaw/openclaw/issues/52928 | https://api.github.com/repos/openclaw/openclaw/issues/52928 | feat(gateway): add agents.setDefault RPC method | ## Problem
Currently there is no dedicated Gateway RPC method to change the default agent. The only way is through `config.patch`, which requires:
1. Reading the full config to locate the target agent's index in `agents.list`
2. Constructing a merge-patch to set `default: true` on the target entry
3. Handling `baseHa... | open | null | false | 1 | [] | [] | 2026-03-23T14:33:46Z | 2026-05-04T21:01:03Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | samzong | 13,782,141 | MDQ6VXNlcjEzNzgyMTQx | User | false |
openclaw/openclaw | 4,121,698,483 | I_kwDOQb6kR871rCCz | 52,963 | https://github.com/openclaw/openclaw/issues/52963 | https://api.github.com/repos/openclaw/openclaw/issues/52963 | Support Slack user-level events (events on behalf of users) for DM routing | ## Feature Request
**Use case:** I have a dedicated Slack user account for my AI agent (Mandy Monday at monday.com). I want people to DM the regular Slack user and have the agent respond as that user - no APP badge, fully natural conversation.
**Current behavior:** OpenClaw's Slack channel only processes bot-level ev... | open | null | false | 1 | [] | [] | 2026-03-23T15:22:40Z | 2026-05-01T21:59:27Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | mondaymandy | 266,874,969 | U_kgDOD-gwWQ | User | false |
openclaw/openclaw | 4,121,759,432 | I_kwDOQb6kR871rQ7I | 52,972 | https://github.com/openclaw/openclaw/issues/52972 | https://api.github.com/repos/openclaw/openclaw/issues/52972 | Bug: Incorrect 'I did not schedule a reminder' note appended to messages after scheduling cron reminders | ## Description
When I successfully schedule a one-time reminder using `openclaw cron add --at`, the system incorrectly appends this note to my response:
```
Note: I did not schedule a reminder in this turn, so this will not trigger automatically.
```
## Steps to Reproduce
1. Ask the assistant to schedule a reminder... | open | null | false | 5 | [] | [] | 2026-03-23T15:30:36Z | 2026-05-01T21:59:29Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | vitobotta | 283,811 | MDQ6VXNlcjI4MzgxMQ== | User | false |
openclaw/openclaw | 4,122,182,832 | I_kwDOQb6kR871s4Sw | 53,008 | https://github.com/openclaw/openclaw/issues/53008 | https://api.github.com/repos/openclaw/openclaw/issues/53008 | Memory compaction blocks main processing lane, causing unresponsive bot for 10+ minutes | ## Bug Report
### What happened
A session memory compaction (memoryFlush) run started and hung for the full 10-minute timeout (600,000ms). During this time, the main processing lane was blocked (`totalActive=1`), which meant **all inbound Telegram messages were queued and never processed**. The bot appeared completely... | open | null | false | 2 | [] | [] | 2026-03-23T16:37:05Z | 2026-05-04T21:01:05Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | petiblaster | 73,549,597 | MDQ6VXNlcjczNTQ5NTk3 | User | false |
openclaw/openclaw | 4,122,328,164 | I_kwDOQb6kR871tbxk | 53,023 | https://github.com/openclaw/openclaw/issues/53023 | https://api.github.com/repos/openclaw/openclaw/issues/53023 | Feature Request: Configurable Session Lane Concurrency + Auto-Yield | # Feature Request: Configurable Session Lane Concurrency + Auto-Yield
## Problem
When an agent is executing sequential tool calls, new user messages queue behind them in the session lane. Session lanes have hardcoded `maxConcurrent: 1` (in `getLaneState()`), causing 100-268+ second delays that are visible to users as... | open | null | false | 2 | [] | [] | 2026-03-23T17:02:17Z | 2026-05-04T21:01:07Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | marcelolebre | 5,076,535 | MDQ6VXNlcjUwNzY1MzU= | User | false |
openclaw/openclaw | 4,122,571,052 | I_kwDOQb6kR871uXEs | 53,056 | https://github.com/openclaw/openclaw/issues/53056 | https://api.github.com/repos/openclaw/openclaw/issues/53056 | [Slack] Double HTML encoding causes & to appear instead of & | ## Bug Description
When messages are sent to Slack, the ampersand character (`&`) gets double-encoded, resulting in `&` appearing in Slack instead of `&`.
## Root Cause
In the OpenClaw send module (`dist/send-*.js`), there's an `escapeHtml` function that pre-escapes ampersands:
```javascript
function escapeHtml(t... | open | null | false | 4 | [] | [] | 2026-03-23T17:47:27Z | 2026-05-04T21:01:09Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | immanwell | 70,859,080 | MDQ6VXNlcjcwODU5MDgw | User | false |
openclaw/openclaw | 4,123,004,746 | I_kwDOQb6kR871wA9K | 53,088 | https://github.com/openclaw/openclaw/issues/53088 | https://api.github.com/repos/openclaw/openclaw/issues/53088 | [Feature]: Control console session list should show full date | ### Summary
Control console session list should directly show full date/time (e.g., "2026-03-20 14:30") instead of only showing time (hour:minute).
### Problem to solve
The control console session list currently only shows time (e.g., "14:30") without any date information. Users cannot identify when a session occurr... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-23T18:54:40Z | 2026-05-04T21:01:11Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | matingxiu | 24,728,784 | MDQ6VXNlcjI0NzI4Nzg0 | User | false |
openclaw/openclaw | 4,123,070,790 | I_kwDOQb6kR871wRFG | 53,092 | https://github.com/openclaw/openclaw/issues/53092 | https://api.github.com/repos/openclaw/openclaw/issues/53092 | Search provider config schema should validate against extension-registered providers | ## Summary
The `tools.web.search.provider` config field was changed from a strict enum (`z.union([z.literal("brave"), z.literal("perplexity")])`) to `z.string().optional()` to accommodate extension-provided search providers (tavily, duckduckgo, exa). While this unblocks extensions, it removes all validation — any typo... | closed | completed | true | 2 | [] | [] | 2026-03-23T19:04:06Z | 2026-05-05T04:28:24Z | 2026-05-02T07:05:18Z | CONTRIBUTOR | null | 20260505T060034Z | 2026-05-05T06:00:34Z | TinyTb | 5,957,298 | MDQ6VXNlcjU5NTcyOTg= | User | false |
openclaw/openclaw | 4,123,260,390 | I_kwDOQb6kR871w_Xm | 53,107 | https://github.com/openclaw/openclaw/issues/53107 | https://api.github.com/repos/openclaw/openclaw/issues/53107 | feat: Cron job tags/categories for organization and filtering | ## Feature Request: Cron job tags / categories
### Problem
As the number of cron jobs grows, it becomes hard to manage and understand what's running.
Currently, jobs only have a `name` and `id`. There's no way to group, filter, or prioritize jobs by purpose.
### Use case
In a multi-agent setup (e.g. a primary inst... | open | null | false | 1 | [] | [] | 2026-03-23T19:38:55Z | 2026-05-04T21:01:13Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | stevenbaert | 43,472,567 | MDQ6VXNlcjQzNDcyNTY3 | User | false |
openclaw/openclaw | 4,123,622,416 | I_kwDOQb6kR871yXwQ | 53,145 | https://github.com/openclaw/openclaw/issues/53145 | https://api.github.com/repos/openclaw/openclaw/issues/53145 | Inter-session announce plumbing can leak raw placeholder messages and echo NO_REPLY / follow-ups | Title: Inter-session announce plumbing can leak raw placeholder messages and echo NO_REPLY/user follow-ups
Repo: openclaw/openclaw
Summary
When using agent-to-agent messaging via `sessions_send`, the inter-session announce path can leak internal placeholder text into the sender-visible chat, and can also forward `NO_... | closed | completed | true | 3 | [] | [] | 2026-03-23T20:45:33Z | 2026-05-05T04:28:31Z | 2026-05-02T06:22:30Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | TarahAssistant | 258,112,504 | U_kgDOD2J7-A | User | false |
openclaw/openclaw | 4,123,950,387 | I_kwDOQb6kR871zn0z | 53,185 | https://github.com/openclaw/openclaw/issues/53185 | https://api.github.com/repos/openclaw/openclaw/issues/53185 | Feature: Integrate content-sanitizer.py into web_fetch/web_search pipeline | ## Summary
Add content-sanitizer.py (rule-based HTML stripping, injection detection) into the web_fetch and web_search tool pipelines.
## Problem
Current behavior:
- web_fetch/web_search add a SECURITY NOTICE wrapper
- Each result is wrapped in EXTERNAL_UNTRUSTED_CONTENT tags
- Content is marked as untrusted in metada... | open | null | false | 3 | [] | [] | 2026-03-23T21:58:03Z | 2026-05-04T21:01:15Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | BruceHowells | 13,401,442 | MDQ6VXNlcjEzNDAxNDQy | User | false |
openclaw/openclaw | 4,123,955,714 | I_kwDOQb6kR871zpIC | 53,186 | https://github.com/openclaw/openclaw/issues/53186 | https://api.github.com/repos/openclaw/openclaw/issues/53186 | replyToMode: "off" and replyToMode: "first" have no effect on Mattermost DM replies — replies are always threaded | **OpenClaw version:** 2026.3.22
**Channel:** Mattermost
### Description
In Mattermost direct message conversations, DM replies are sometimes posted as threaded replies despite the intended `replyToMode: "off"` semantics for DMs. The DM thread-context resolver correctly returns no reply target for direct chats, but th... | closed | completed | true | 5 | [] | [] | 2026-03-23T21:59:07Z | 2026-04-30T04:35:02Z | 2026-04-27T10:22:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | joho0 | 78,939,530 | MDQ6VXNlcjc4OTM5NTMw | User | false |
openclaw/openclaw | 4,124,047,960 | I_kwDOQb6kR871z_pY | 53,198 | https://github.com/openclaw/openclaw/issues/53198 | https://api.github.com/repos/openclaw/openclaw/issues/53198 | [Bug]: Discord elevated allowFrom fallback from channels.discord.allowFrom does not work on 2026.3.22 | ### Bug type
Regression (worked before, now fails)
### Summary
On OpenClaw `2026.3.22`, Discord elevated exec no longer works when relying on the documented fallback from `channels.discord.allowFrom` to `tools.elevated.allowFrom.discord`.
Docs say this fallback should work:
- `tools.elevated.allowFrom.discord` omit... | open | null | false | 2 | [] | [] | 2026-03-23T22:23:42Z | 2026-05-04T21:01:17Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | fdovarela | 3,207,861 | MDQ6VXNlcjMyMDc4NjE= | User | false |
openclaw/openclaw | 4,124,152,595 | I_kwDOQb6kR8710ZMT | 53,208 | https://github.com/openclaw/openclaw/issues/53208 | https://api.github.com/repos/openclaw/openclaw/issues/53208 | WhatsApp trusted relay contact in separate DM is incorrectly told relay to main user is unavailable | # Bug: Trusted WhatsApp relay contact in separate DM is incorrectly told relay to main user is unavailable
## Summary
A trusted relay contact (Jill) is explicitly approved to send messages, links, media, and voice notes to Jack through OpenClaw on WhatsApp. Outbound sending to both contacts works. However, when Jill a... | open | null | false | 1 | [] | [] | 2026-03-23T22:49:47Z | 2026-05-04T21:01:19Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | stumcmurtrie | 31,492,084 | MDQ6VXNlcjMxNDkyMDg0 | User | false |
openclaw/openclaw | 4,124,274,288 | I_kwDOQb6kR871025w | 53,228 | https://github.com/openclaw/openclaw/issues/53228 | https://api.github.com/repos/openclaw/openclaw/issues/53228 | [Feature]: openclaw notify — admin-scoped push to connected UI sessions | ### Summary
openclaw notify CLI command to push visible notifications to connected UI sessions from shell scripts
### Problem to solve
A new openclaw notify <message> CLI command that broadcasts a visible notification bubble to all connected UI sessions using admin-scoped broadcast internally. No agent turn required... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-23T23:24:52Z | 2026-05-04T21:01:21Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | wow-leeroy-jenkins05 | 266,177,925 | U_kgDOD92NhQ | User | false |
openclaw/openclaw | 4,131,669,377 | I_kwDOQb6kR872REWB | 54,128 | https://github.com/openclaw/openclaw/issues/54128 | https://api.github.com/repos/openclaw/openclaw/issues/54128 | Add maxThreads config for local embedding (node-llama-cpp) | ### Problem
When using `memorySearch.provider = "local"` with `node-llama-cpp`, the embedding process doesn't fully utilize available CPU cores. On a 16-core system, only ~6 cores are used (~42% CPU), leaving significant compute resources idle during memory indexing.
### Current Behavior
OpenClaw calls `getLlama()` ... | open | null | false | 2 | [] | [] | 2026-03-25T01:05:15Z | 2026-05-04T21:19:17Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | hsuaaron | 24,956,301 | MDQ6VXNlcjI0OTU2MzAx | User | false |
openclaw/openclaw | 4,131,762,668 | I_kwDOQb6kR872RbHs | 54,138 | https://github.com/openclaw/openclaw/issues/54138 | https://api.github.com/repos/openclaw/openclaw/issues/54138 | [Bug]: Heartbeat mode leaks [TOOL_CALL] bracket-format blocks into user messages | ## Bug Description
When a heartbeat handler runs tool calls (e.g. `exec`, `cron list`), the embedded agent outputs `[TOOL_CALL]{tool: "exec", args: {...}}[/TOOL_CALL]` blocks in its response. These leak into user-facing Telegram/Discord messages instead of being silently discarded.
## Steps to Reproduce
1. Have HEAR... | closed | completed | true | 6 | [] | [] | 2026-03-25T01:33:34Z | 2026-05-04T04:33:53Z | 2026-05-02T04:20:43Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | Deniable9570 | 266,249,687 | U_kgDOD96l1w | User | false |
openclaw/openclaw | 4,131,790,319 | I_kwDOQb6kR872Rh3v | 54,149 | https://github.com/openclaw/openclaw/issues/54149 | https://api.github.com/repos/openclaw/openclaw/issues/54149 | Volcano Engine (Doubao/BytePlus) model catalog: Kimi K2.5 reasoning=false and maxTokens=4096, inconsistent with Together provider | ## Problem
The Volcano Engine shared model catalog (`src/agents/volc-models.shared.ts`) defines Kimi K2.5 with incorrect/outdated metadata:
```typescript
export const VOLC_MODEL_KIMI_K2_5 = {
id: "kimi-k2-5-260127",
name: "Kimi K2.5",
reasoning: false, // ← Should be true
input: ["text", "image"] as const,... | open | null | false | 1 | [] | [] | 2026-03-25T01:42:24Z | 2026-05-04T21:19:20Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,131,804,194 | I_kwDOQb6kR872RlQi | 54,153 | https://github.com/openclaw/openclaw/issues/54153 | https://api.github.com/repos/openclaw/openclaw/issues/54153 | signal-cli-install: downloadToFile uses raw http.request with no timeout, no SSRF guard, no size limit | ## Problem
`src/plugins/signal-cli-install.ts:97-119` implements a `downloadToFile` function using Node.js `https.request()` directly, without timeout, SSRF protection, or download size limits:
```typescript
import { request } from "node:https";
async function downloadToFile(url: string, dest: string, maxRedirects =... | closed | completed | true | 4 | [] | [] | 2026-03-25T01:47:38Z | 2026-05-03T04:36:56Z | 2026-04-30T14:35:23Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,131,808,001 | I_kwDOQb6kR872RmMB | 54,155 | https://github.com/openclaw/openclaw/issues/54155 | https://api.github.com/repos/openclaw/openclaw/issues/54155 | Gateway memory leak: 389MB → 14.7GB over 4 days with session accumulation | ## Description
The `openclaw-gateway` process leaks memory steadily over multi-day uptime, growing from ~389MB at startup to 14.7GB (58% of 24GB RAM on Mac Mini) after approximately 4 days of continuous operation.
## Environment
- **OpenClaw**: 2026.3.13 → 2026.3.23-2 (observed on both)
- **OS**: macOS 15.7.4 (arm64... | open | null | false | 3 | [] | [] | 2026-03-25T01:49:09Z | 2026-05-04T21:19:22Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | the-lobsternaut | 263,267,851 | U_kgDOD7EmCw | User | false |
openclaw/openclaw | 4,131,813,246 | I_kwDOQb6kR872Rnd- | 54,157 | https://github.com/openclaw/openclaw/issues/54157 | https://api.github.com/repos/openclaw/openclaw/issues/54157 | Doubao/BytePlus model catalog: cacheRead/cacheWrite pricing set to 0, missing context caching cost | ## Problem
Both `src/agents/doubao-models.ts` and `src/agents/byteplus-models.ts` define `cacheRead: 0` and `cacheWrite: 0` in their cost config:
```typescript
// Doubao:
export const DOUBAO_DEFAULT_COST = {
input: 0.0001,
output: 0.0002,
cacheRead: 0, // ← Missing pricing
cacheWrite: 0, // ← Missing pri... | open | null | false | 3 | [] | [] | 2026-03-25T01:51:13Z | 2026-05-04T21:19:24Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.