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 | 4,076,583,742 | I_kwDOQb6kR87y-7s- | 46,539 | https://github.com/openclaw/openclaw/issues/46539 | https://api.github.com/repos/openclaw/openclaw/issues/46539 | [Bug]: resetByType daily mode never resets channel conversation sessions — mergeSessionEntry() refreshes updatedAt before freshness check | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Daily session reset configured via `session.resetByType.direct` / `session.resetByType.group` with `mode: "daily"` does not appear to rotate real conversation sessions such as Telegram direct chats.
In my setup, the affected session is:
-... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-14T19:46:02Z | 2026-04-27T04:42:05Z | 2026-04-25T01:29:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | zambriguelito-commits | 264,039,815 | U_kgDOD7zthw | User | false |
openclaw/openclaw | 4,076,588,921 | I_kwDOQb6kR87y-895 | 46,540 | https://github.com/openclaw/openclaw/issues/46540 | https://api.github.com/repos/openclaw/openclaw/issues/46540 | sessions_spawn model override silently ignored for Ollama provider | # Bug: `sessions_spawn` model override silently ignored for non-Anthropic providers (e.g. Ollama)
## Summary
When spawning a sub-agent via the `sessions_spawn` tool with `model: "ollama/qwen3:14b"`, the model override is stored in `runs.json` but **not applied** to the actual session. The sub-agent always runs on the... | closed | completed | true | 2 | [] | [] | 2026-03-14T19:48:56Z | 2026-04-27T04:42:04Z | 2026-04-25T01:29:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | overwrite-bot | 268,238,486 | U_kgDOD_z-lg | User | false |
openclaw/openclaw | 4,075,829,834 | I_kwDOQb6kR87y8DpK | 46,164 | https://github.com/openclaw/openclaw/issues/46164 | https://api.github.com/repos/openclaw/openclaw/issues/46164 | [Bug]: Webchat bug: Black triangle overlay covers chat area, requires manual click to dismiss (Edge 145.0.3800.97, 8 occurrences today) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## 问题描述
在 OpenClaw Webchat 界面中，不时会出现一个黑色三角形覆盖整个聊天区域，必须手动点击才能恢复正常。该问题在多个模型（Qwen3.5、DeepSeek-R1、OpenRouter 免费模型）下均出现过，非特定模型问题。
## 重现步骤
1. 正常使用 Webchat 对话
2. 随机出现黑色三角形覆盖
3. 必须点击三角形才能继续使用
##Bug 流程（推测）：
我发送消息 → AI生成回复
回复到达前端 → 显示黑色三角形（遮罩）
遮罩没有... | closed | completed | true | 7 | [
"bug",
"bug:behavior"
] | [] | 2026-03-14T13:30:20Z | 2026-04-27T04:42:03Z | 2026-04-25T01:12:05Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lazyom | 266,662,874 | U_kgDOD-Tz2g | User | false |
openclaw/openclaw | 4,076,591,937 | I_kwDOQb6kR87y-9tB | 46,543 | https://github.com/openclaw/openclaw/issues/46543 | https://api.github.com/repos/openclaw/openclaw/issues/46543 | Feature Request: Add Chatterbox TTS Support | Feature Request: Add Chatterbox TTS Support
## Description
Add Chatterbox (local TTS) as a native TTS provider option in OpenClaw.
## Why Chatterbox?
- Local TTS (no API costs)
- High quality voice generation
- Custom voice cloning from audio samples
- Privacy-friendly (runs locally)
## Use Case
Users want to use ... | closed | not_planned | true | 3 | [] | [] | 2026-03-14T19:50:35Z | 2026-04-27T04:42:02Z | 2026-04-25T01:29:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vinaykotha433-del | 262,733,006 | U_kgDOD6j8zg | User | false |
openclaw/openclaw | 4,076,620,691 | I_kwDOQb6kR87y_EuT | 46,545 | https://github.com/openclaw/openclaw/issues/46545 | https://api.github.com/repos/openclaw/openclaw/issues/46545 | Thread-bound subagent sessions fail on second turn with Anthropic thinking blocks | ## Bug Report
### Summary
Thread-bound subagent sessions on Anthropic models (Opus 4.6, Sonnet 4.6) with thinking enabled fail on the second turn with:
```
LLM request rejected: messages.5.content.1: `thinking` or `redacted_thinking` blocks in the latest assistant message cannot be modified. These blocks must remain... | closed | completed | true | 1 | [] | [] | 2026-03-14T20:05:52Z | 2026-04-27T04:42:01Z | 2026-04-25T01:30:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | blortski | 117,885,527 | U_kgDOBwbKVw | User | false |
openclaw/openclaw | 4,076,629,372 | I_kwDOQb6kR87y_G18 | 46,547 | https://github.com/openclaw/openclaw/issues/46547 | https://api.github.com/repos/openclaw/openclaw/issues/46547 | [Feature]: Support web_search with Grok + Perplexity via OpenRouter | ### Summary
Currently, when using Grok + Perplexity (via OpenRouter), the functionality of web_search tool is dropped.
See https://github.com/openclaw/openclaw/issues/14749 for reference.
Behaviour was changed via commit https://github.com/openclaw/openclaw/commit/e11a0775e79dc5ade7e6f01edccfc7b6e024819d.
What chan... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-14T20:10:26Z | 2026-04-27T04:42:01Z | 2026-04-25T01:30:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | andres-lindo | 71,335,326 | MDQ6VXNlcjcxMzM1MzI2 | User | false |
openclaw/openclaw | 4,076,650,794 | I_kwDOQb6kR87y_MEq | 46,553 | https://github.com/openclaw/openclaw/issues/46553 | https://api.github.com/repos/openclaw/openclaw/issues/46553 | Feature Request: 跨渠道聊天记录合并/整合 | ## 背景描述
当前 OpenClaw 支持多种聊天渠道（飞书、Telegram、Discord、WhatsApp 等），但各渠道的聊天记录是分开存储的，彼此不互通。
## 问题场景
用户通过不同渠道与 AI 助手对话时，每个渠道的聊天记录是独立的。例如：
- 通过飞书发送的消息和通过网关 Control UI 发送的消息，AI 助手无法同时看到两边的对话历史
- 这导致用户体验不连贯，感觉像是在和两个不同的 AI 对话
## 功能需求
希望官方考虑以下方案之一：
1. **统一聊天记录存储**：所有渠道的消息存储到同一个数据库/文件系统，支持跨渠道查询
2. **会话关联功能**：允许将不同渠道的对话关联到同一个会话或... | closed | completed | true | 1 | [] | [] | 2026-03-14T20:22:01Z | 2026-04-27T04:42:00Z | 2026-04-25T01:30:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | huhaiting | 194,820,047 | U_kgDOC5y3zw | User | false |
openclaw/openclaw | 4,076,662,974 | I_kwDOQb6kR87y_PC- | 46,557 | https://github.com/openclaw/openclaw/issues/46557 | https://api.github.com/repos/openclaw/openclaw/issues/46557 | CLI local gateway detail probes drop device auth on localhost and localLoopback probe budget is too low | ## Summary
On local gateways using `gateway.auth.mode=token`, CLI detail calls on localhost can lose paired operator scopes because device identity is suppressed when a shared gateway token/password is present. This makes `status --deep` and `last-heartbeat` behave inconsistently even though basic health checks still w... | closed | completed | true | 2 | [] | [] | 2026-03-14T20:28:48Z | 2026-04-27T04:41:59Z | 2026-04-25T01:31:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jeffrey4341 | 77,844,634 | MDQ6VXNlcjc3ODQ0NjM0 | User | false |
openclaw/openclaw | 4,076,673,044 | I_kwDOQb6kR87y_RgU | 46,560 | https://github.com/openclaw/openclaw/issues/46560 | https://api.github.com/repos/openclaw/openclaw/issues/46560 | A2A plugin local WS dispatch lacks device identity + challenge nonce signing → operator.write downgrade on 2026.3.13 |
## Summary
After updating to OpenClaw 2026.3.13, the A2A gateway plugin's local WS dispatch fails with `missing scope: operator.write` for all agent targets. Remote A2A (HTTP) still works. The root cause is that the plugin's `GatewayRpcConnection` does not include device identity or sign the `connect.challenge` nonce... | closed | completed | true | 2 | [] | [] | 2026-03-14T20:33:25Z | 2026-04-27T04:41:58Z | 2026-04-25T01:31:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bhgtiger | 232,811,003 | U_kgDODeBp-w | User | false |
openclaw/openclaw | 4,326,523,565 | I_kwDOQb6kR88AAAABAeGCrQ | 71,325 | https://github.com/openclaw/openclaw/issues/71325 | https://api.github.com/repos/openclaw/openclaw/issues/71325 | compaction.mode=safeguard never triggers automatically when context engine windows messages or sets ownsCompaction | ## Symptom
With `compaction.mode: "safeguard"` configured, sessions can grow far past `contextTokens` (e.g. 640k tokens / 3.2x cap) without the safeguard ever firing. Hard compaction must be triggered manually with `/compact`. Cron sessions are particularly affected — observed ratios up to 756% of context cap.
memory... | closed | completed | true | 1 | [] | [] | 2026-04-25T01:14:57Z | 2026-04-27T04:41:57Z | 2026-04-25T01:33:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | joeywrightphoto | 114,463,115 | U_kgDOBtKRiw | User | false |
openclaw/openclaw | 4,154,878,488 | I_kwDOQb6kR873pmoY | 55,956 | https://github.com/openclaw/openclaw/issues/55956 | https://api.github.com/repos/openclaw/openclaw/issues/55956 | [Bug]: Voice Call Plugin: Telnyx transcription field mapping is wrong | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The Telnyx provider's event mapper reads data.payload.transcription for speech transcripts, but Telnyx actually sends it at data.payload.transcription_data.transcript. This causes every transcript to resolve as... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-27T18:04:25Z | 2026-04-27T04:41:54Z | 2026-04-25T01:51:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | IbleedDialTone | 40,133,014 | MDQ6VXNlcjQwMTMzMDE0 | User | false |
openclaw/openclaw | 4,326,650,250 | I_kwDOQb6kR88AAAABAeNxig | 71,348 | https://github.com/openclaw/openclaw/issues/71348 | https://api.github.com/repos/openclaw/openclaw/issues/71348 | [BUG] Node.js v24 compile cache causes ERR_MODULE_NOT_FOUND after package update | ## Problem Description
After updating `openclaw`, the CLI crashes with `ERR_MODULE_NOT_FOUND` because Node.js v24's `module.enableCompileCache()` caches resolved filenames that include content hashes. When `openclaw` is updated, the `dist/` directory contains new hashed filenames, but the compile cache still points to... | closed | not_planned | true | 3 | [
"cannot-reproduce",
"cli",
"triage:bug",
"regression",
"bug:crash"
] | [] | 2026-04-25T01:56:49Z | 2026-04-30T04:43:02Z | 2026-04-26T04:39:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JIRBOY | 49,422,297 | MDQ6VXNlcjQ5NDIyMjk3 | User | false |
openclaw/openclaw | 4,076,699,492 | I_kwDOQb6kR87y_X9k | 46,567 | https://github.com/openclaw/openclaw/issues/46567 | https://api.github.com/repos/openclaw/openclaw/issues/46567 | [Bug]:Telegram pairing approval does not persist in Docker/container OpenClaw | ### Bug type
Regression (worked before, now fails)
### Summary
BUG: Telegram pairing approval does not persist in Docker/container OpenClaw
OpenClaw version: 2026.3.13 (Docker, Hostinger VPS, persistent /data mount)
Telegram pairing in container always reports "Approved telegram sender 5180162669." but telegram-pair... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-14T20:47:18Z | 2026-04-27T04:41:52Z | 2026-04-25T01:57:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | trinmabrker | 264,659,249 | U_kgDOD8ZhMQ | User | false |
openclaw/openclaw | 4,076,713,035 | I_kwDOQb6kR87y_bRL | 46,569 | https://github.com/openclaw/openclaw/issues/46569 | https://api.github.com/repos/openclaw/openclaw/issues/46569 | Regression: node-llama-cpp missing after upgrade to 2026.3.12 | ## Steps to reproduce
1. Install OpenClaw 2026.3.11: `npm i -g openclaw@2026.3.11`
2. Verify memory search works: `openclaw memory status` → shows `Provider: local`
3. Upgrade: `npm i -g openclaw@2026.3.12`
4. Restart gateway: `openclaw gateway restart`
5. Check memory search: `openclaw memory status` → shows `Provider... | closed | completed | true | 3 | [] | [] | 2026-03-14T20:55:25Z | 2026-04-27T04:41:51Z | 2026-04-25T01:57:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | matthewcarano | 30,149,603 | MDQ6VXNlcjMwMTQ5NjAz | User | false |
openclaw/openclaw | 4,076,724,366 | I_kwDOQb6kR87y_eCO | 46,572 | https://github.com/openclaw/openclaw/issues/46572 | https://api.github.com/repos/openclaw/openclaw/issues/46572 | User messages rendered inside thinking block context in TUI | ## Description
User messages sent via WhatsApp/Telegram are sometimes rendered in the TUI as if they are part of the agent's internal thinking/system context, rather than as distinct user turns.
## Expected behavior
User messages should render as clearly delineated user turns in the TUI output.
## Actual behavior
... | closed | completed | true | 1 | [] | [] | 2026-03-14T21:00:58Z | 2026-04-27T04:41:51Z | 2026-04-25T01:57:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | eggrollofchaos | 42,749,456 | MDQ6VXNlcjQyNzQ5NDU2 | User | false |
openclaw/openclaw | 4,076,765,322 | I_kwDOQb6kR87y_oCK | 46,582 | https://github.com/openclaw/openclaw/issues/46582 | https://api.github.com/repos/openclaw/openclaw/issues/46582 | [Bug]: banner uses cumulative Tokens in instead of Context" | ### Bug type
Regression (worked before, now fails)
### Summary
I've started to get this banner occasionally. Seems like a UI bug as the context is not actually full
### Steps to reproduce
Message in web chat and a big banner appears
### Expected behavior
No banner
### Actual behavior
<img width="960" height="2... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-14T21:21:28Z | 2026-04-27T04:41:50Z | 2026-04-25T01:57:59Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | hartmark | 12,826,053 | MDQ6VXNlcjEyODI2MDUz | User | false |
openclaw/openclaw | 4,076,765,545 | I_kwDOQb6kR87y_oFp | 46,583 | https://github.com/openclaw/openclaw/issues/46583 | https://api.github.com/repos/openclaw/openclaw/issues/46583 | Bug: Control UI re-pairs without operator.read/operator.write and local 'openclaw devices list' closes on loopback | # Bug report: Control UI re-pairs without operator.read/operator.write, and `openclaw devices list` closes the connection on local loopback
## Summary
On OpenClaw `2026.3.13`, two related issues appear during local Gateway/operator auth on a single-machine install:
1. The **Control UI / webchat client** (`clientId: ... | closed | completed | true | 3 | [] | [] | 2026-03-14T21:21:36Z | 2026-04-27T04:41:49Z | 2026-04-25T01:58:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | machinespirit-reborn | 268,110,736 | U_kgDOD_sLkA | User | false |
openclaw/openclaw | 4,326,654,147 | I_kwDOQb6kR88AAAABAeOAww | 71,349 | https://github.com/openclaw/openclaw/issues/71349 | https://api.github.com/repos/openclaw/openclaw/issues/71349 | memory-lancedb: autoCapture rarely fires; only 2 entries persisted across many sessions | ## Summary
The `memory-lancedb` plugin loads cleanly and recall works, but `autoCapture` is silently failing on most sessions. After ~weeks of use across multiple agents and channels, `openclaw ltm stats` reports only 2 entries, both from the same hour on 2026-04-23.
## Environment
- openclaw `2026.4.23` (commit `a979... | closed | completed | true | 1 | [] | [] | 2026-04-25T01:58:22Z | 2026-04-30T04:35:58Z | 2026-04-27T07:41:45Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | 19wburnett | 153,020,893 | U_kgDOCR7p3Q | User | false |
openclaw/openclaw | 4,076,809,776 | I_kwDOQb6kR87y_y4w | 46,595 | https://github.com/openclaw/openclaw/issues/46595 | https://api.github.com/repos/openclaw/openclaw/issues/46595 | sessions_spawn with agentId does not use target agent's workspace | When using `sessions_spawn` with `agentId` to spawn a subagent under a different agent, the workspace (cwd and bootstrap context) always inherits from the **requester** agent, not the **target** agent.
## Expected Behavior
When `agentId: "alex"` is passed to `sessions_spawn`, the subagent should:
1. Use the target ag... | closed | completed | true | 3 | [] | [] | 2026-03-14T21:43:20Z | 2026-04-27T04:41:48Z | 2026-04-25T01:58:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mira-solari | 268,252,643 | U_kgDOD_014w | User | false |
openclaw/openclaw | 4,076,822,914 | I_kwDOQb6kR87y_2GC | 46,599 | https://github.com/openclaw/openclaw/issues/46599 | https://api.github.com/repos/openclaw/openclaw/issues/46599 | Memory SQLite index empty after long session — memory_search returns 'database is not open' | ## Bug Report
**OpenClaw version:** 2026.3.13 (61d171a)
**OS:** macOS Darwin 24.6.0 (x64)
**Node:** v22.22.0
## Summary
After a long active session (several hours, many tool calls), `memory_search` began returning `{ disabled: true, error: 'database is not open' }`. The memory SQLite index (`~/.openclaw/memory/main.... | closed | completed | true | 1 | [] | [] | 2026-03-14T21:49:01Z | 2026-04-27T04:41:48Z | 2026-04-25T01:58:49Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | parkerati | 257,907,588 | U_kgDOD19bhA | User | false |
openclaw/openclaw | 4,326,655,223 | I_kwDOQb6kR88AAAABAeOE9w | 71,350 | https://github.com/openclaw/openclaw/issues/71350 | https://api.github.com/repos/openclaw/openclaw/issues/71350 | [Feature]: per-agent thinking level, temperature, and adaptive thinking strategies | ### Summary
Per-agent thinking level, temperature, and adaptive thinking strategies
### Problem to solve
All agents share the same `thinkingDefault` and model params (`temperature`, `top_p`, etc.) — only configurable at `agents.defaults` level. The docs mention `agents.list[].thinkingDefault` in the thinking resolut... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-25T01:58:52Z | 2026-05-03T09:49:08Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | tryxin | 20,015,204 | MDQ6VXNlcjIwMDE1MjA0 | User | false |
openclaw/openclaw | 4,076,825,555 | I_kwDOQb6kR87y_2vT | 46,600 | https://github.com/openclaw/openclaw/issues/46600 | https://api.github.com/repos/openclaw/openclaw/issues/46600 | [Bug]: Fallback model not applied to cron sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Fallback model config (agents.defaults.model.fallbacks) is not being applied to cron-triggered sessions. Regular sessions use the correct fallback, but cron jobs use a hardcoded default (gpt-5.2 in this case).
### Steps to reproduce
{
... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-14T21:50:29Z | 2026-04-27T04:41:47Z | 2026-04-25T01:59:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | VisionAIrySE | 164,838,211 | U_kgDOCdM7Qw | User | false |
openclaw/openclaw | 4,076,846,303 | I_kwDOQb6kR87y_7zf | 46,609 | https://github.com/openclaw/openclaw/issues/46609 | https://api.github.com/repos/openclaw/openclaw/issues/46609 | [Bug]: Signal plugin failed to load (Cannot find module ../../../src/infra/outbound/send-deps.js) | ### Bug type
Regression (worked before, now fails)
### Summary
I was installing OpenClaw the other day, an update popped a few hours ago and now signal plugin won't load anymore. I guess it has something to do with this commit : https://github.com/openclaw/openclaw/commit/d9c285e930d744a747ccc4ee600b4f6ffd6bd772#dif... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-14T22:02:14Z | 2026-04-27T04:41:46Z | 2026-04-25T01:59:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | YanouHD | 5,512,265 | MDQ6VXNlcjU1MTIyNjU= | User | false |
openclaw/openclaw | 4,076,888,965 | I_kwDOQb6kR87zAGOF | 46,616 | https://github.com/openclaw/openclaw/issues/46616 | https://api.github.com/repos/openclaw/openclaw/issues/46616 | [Bug]: bailian provider token usage not recorded (usage field mapping mismatch) | ### Bug type
Regression (worked before, now fails)
### Summary
## Issue: bailian provider token usage not recorded (usage field mapping mismatch)
### Description
When using the bailian (阿里百炼) provider, token usage statistics are not being recorded in session data. The `totalTokens`, `inputTokens`, and `outputToken... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-14T22:23:34Z | 2026-04-27T04:41:45Z | 2026-04-25T01:59:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sg1416-zg | 259,938,361 | U_kgDOD35YOQ | User | false |
openclaw/openclaw | 4,076,893,131 | I_kwDOQb6kR87zAHPL | 46,618 | https://github.com/openclaw/openclaw/issues/46618 | https://api.github.com/repos/openclaw/openclaw/issues/46618 | Feature Request: Context engine plugin API — topic-aware context assembly | ## Summary
A **context engine plugin** that replaces linear context management with topic-aware, relevance-driven assembly. Instead of conversations filling up and getting compacted (losing detail), the context window becomes a **viewport** onto the most relevant parts of the conversation. The agent never runs out of ... | closed | completed | true | 2 | [] | [] | 2026-03-14T22:25:09Z | 2026-04-27T04:41:45Z | 2026-04-25T01:59:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | whisky0809 | 32,197,959 | MDQ6VXNlcjMyMTk3OTU5 | User | false |
openclaw/openclaw | 4,076,914,276 | I_kwDOQb6kR87zAMZk | 46,621 | https://github.com/openclaw/openclaw/issues/46621 | https://api.github.com/repos/openclaw/openclaw/issues/46621 | [Bug]: | ### Bug type
Regression (worked before, now fails)
### Summary
After updating from 2026.3.12 to 2026.3.13, the systemd user service file retains
the old entrypoint (`dist/entry.js`) while the bundler output changed to `dist/index.js`.
## Notes
- `openclaw doctor` detects the mismatch but `--fix` does not corr... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-14T22:34:50Z | 2026-04-27T04:41:44Z | 2026-04-25T02:00:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DariusBetelgeuse | 63,287,771 | MDQ6VXNlcjYzMjg3Nzcx | User | false |
openclaw/openclaw | 4,076,926,437 | I_kwDOQb6kR87zAPXl | 46,622 | https://github.com/openclaw/openclaw/issues/46622 | https://api.github.com/repos/openclaw/openclaw/issues/46622 | [Bug]: thinking: "minimal" causes 400 error when model falls back to NVIDIA NIM/vLLM endpoint | ### Bug type
Regression (worked before, now fails)
### Summary
## Description
Cron job configured with `thinking: "minimal"` and `model: google/gemini-flash-latest`
works on Gemini but fails with 400 when OpenClaw falls back to NVIDIA NIM (vLLM).
NVIDIA NIM accepts only `low`, `medium`, `high` — not `minimal... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-14T22:39:34Z | 2026-04-27T04:41:42Z | 2026-04-25T02:00:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DariusBetelgeuse | 63,287,771 | MDQ6VXNlcjYzMjg3Nzcx | User | false |
openclaw/openclaw | 4,076,928,714 | I_kwDOQb6kR87zAP7K | 46,623 | https://github.com/openclaw/openclaw/issues/46623 | https://api.github.com/repos/openclaw/openclaw/issues/46623 | Skills silently truncated when total prompt chars exceed 30K default limit | ## Bug Report
### Summary
When the total character count of all skill descriptions exceeds `maxSkillsPromptChars` (default 30,000), skills are **silently truncated** from the `<available_skills>` list injected into the system prompt. There is no warning or log message indicating that skills were dropped.
### Environm... | closed | completed | true | 3 | [] | [] | 2026-03-14T22:40:37Z | 2026-04-27T04:41:41Z | 2026-04-25T02:00:46Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lilac-aeon | 9,991,373 | MDQ6VXNlcjk5OTEzNzM= | User | false |
openclaw/openclaw | 4,076,928,764 | I_kwDOQb6kR87zAP78 | 46,624 | https://github.com/openclaw/openclaw/issues/46624 | https://api.github.com/repos/openclaw/openclaw/issues/46624 | [Feature]: Add GitHub Copilot as Provider | ### Summary
Please add GitHub Copilot as a model provider
### Problem to solve
--
### Proposed solution
--
### Alternatives considered
_No response_
### Impact
--
### Evidence/examples
_No response_
### Additional information
_No response_ | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-14T22:40:39Z | 2026-04-27T04:41:41Z | 2026-04-25T02:01:03Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | InvalidPandaa | 117,936,522 | U_kgDOBweRig | User | false |
openclaw/openclaw | 4,076,991,936 | I_kwDOQb6kR87zAfXA | 46,632 | https://github.com/openclaw/openclaw/issues/46632 | https://api.github.com/repos/openclaw/openclaw/issues/46632 | Chat UI shows incorrect token count (647.6k vs actual 173k) | ## Bug Description
Chat interface shows "100% context used 647.6k / 200k" but actual session tokens are 173k/200k (87%).
## Steps to Reproduce
1. Use MiniMax model with extended conversation
2. Observe token warning in chat vs actual in Sessions page
## Expected vs Actual
- Chat UI: 647.6k tokens (shown as 100% used)... | closed | completed | true | 2 | [] | [] | 2026-03-14T23:04:30Z | 2026-04-27T04:41:40Z | 2026-04-25T02:01:19Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | BorrValhal | 266,500,363 | U_kgDOD-J5Cw | User | false |
openclaw/openclaw | 4,076,997,849 | I_kwDOQb6kR87zAgzZ | 46,635 | https://github.com/openclaw/openclaw/issues/46635 | https://api.github.com/repos/openclaw/openclaw/issues/46635 | Security: agent tool deny list not enforced for cron tool | ## Issue
When an agent has `cron` in its `tools.deny` list, the deny list is **not enforced** for the `cron` tool specifically. The agent can modify any cron job in the system.
Other tools in the deny list (e.g. `subagents`) are correctly blocked — so this appears to be specific to the `cron` tool implementation.
##... | closed | not_planned | true | 7 | [] | [] | 2026-03-14T23:07:02Z | 2026-04-27T04:41:39Z | 2026-04-25T02:01:36Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bencoremans | 38,985,877 | MDQ6VXNlcjM4OTg1ODc3 | User | false |
openclaw/openclaw | 4,077,032,074 | I_kwDOQb6kR87zApKK | 46,645 | https://github.com/openclaw/openclaw/issues/46645 | https://api.github.com/repos/openclaw/openclaw/issues/46645 | [Bug]: ACP sessions_send replies are not mirrored into the target worker’s bound Discord thread | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
sessions_send reaches the target ACP worker session and gets an internal reply, but the reply is not posted visibly into the target worker’s bound Discord thread/channel.`
### Steps to reproduce
1. Spawn a persistent ACP worker thread in... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-14T23:22:28Z | 2026-04-27T04:41:38Z | 2026-04-25T02:01:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | BenjaminCarro11 | 163,931,244 | U_kgDOCcVkbA | User | false |
openclaw/openclaw | 4,077,073,956 | I_kwDOQb6kR87zAzYk | 46,655 | https://github.com/openclaw/openclaw/issues/46655 | https://api.github.com/repos/openclaw/openclaw/issues/46655 | Inbound media ghost attachment: previous image re-attached to unrelated subsequent messages | ## Bug Report
**Version:** OpenClaw 2026.3.13
**Channel:** Telegram
## Description
When a user sends a photo via Telegram, a previously received/cached inbound image gets ghost-attached to subsequent unrelated messages. The extra attachment appears in the agent context as if the user sent it again, even though the n... | closed | not_planned | true | 4 | [] | [] | 2026-03-14T23:39:14Z | 2026-04-27T04:41:38Z | 2026-04-25T02:02:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | duclamvan | 5,202,937 | MDQ6VXNlcjUyMDI5Mzc= | User | false |
openclaw/openclaw | 4,077,139,398 | I_kwDOQb6kR87zBDXG | 46,669 | https://github.com/openclaw/openclaw/issues/46669 | https://api.github.com/repos/openclaw/openclaw/issues/46669 | Browser proxy blocks event loop, kills Telegram polling | ## Summary
Browser proxy actions (screenshot, click, navigate) that take >20s block the Node.js event loop, preventing Telegram getUpdates polling. After ~115s of stall, Telegram disconnects and never reconnects.
## Environment
- OpenClaw 2026.3.13
- EC2 (gateway) + Mac node host (browser proxy)
- Chrome DevTools MCP ... | closed | completed | true | 2 | [] | [] | 2026-03-15T00:10:45Z | 2026-04-27T04:41:37Z | 2026-04-25T02:02:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | durang | 14,999,651 | MDQ6VXNlcjE0OTk5NjUx | User | false |
openclaw/openclaw | 4,077,143,367 | I_kwDOQb6kR87zBEVH | 46,671 | https://github.com/openclaw/openclaw/issues/46671 | https://api.github.com/repos/openclaw/openclaw/issues/46671 | [Bug] memory_search returns empty results in 2026.3.13 (regression from #29112 fix) | ## Summary
memory_search returns no results for any query on OpenClaw 2026.3.13, despite successful indexing and data existing in SQLite database. This appears to be a regression or incomplete fix of issue #29112 which was reported as closed on 2026-02-28.
## Environment
- OpenClaw version: 2026.3.13 (61d171a)
- OS: ... | closed | completed | true | 3 | [] | [] | 2026-03-15T00:12:33Z | 2026-04-27T04:41:36Z | 2026-04-25T02:02:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,077,158,427 | I_kwDOQb6kR87zBIAb | 46,673 | https://github.com/openclaw/openclaw/issues/46673 | https://api.github.com/repos/openclaw/openclaw/issues/46673 | [Feature]: Built-in skill regression testing — `/edd save` + `/edd` workflow | ### Summary
Skills have no quality feedback loop. You edit a SKILL.md, and the only way to know if you broke something is to manually re-test every scenario. As ClawHub scales to 13,000+ skills, this gap affects authors, consumers, and teams sharing workspaces. Proposing a lightweight edd.yaml convention for golden te... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-15T00:20:14Z | 2026-04-27T04:41:35Z | 2026-04-25T02:02:59Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Belyenochi | 26,409,132 | MDQ6VXNlcjI2NDA5MTMy | User | false |
openclaw/openclaw | 4,077,182,213 | I_kwDOQb6kR87zBN0F | 46,681 | https://github.com/openclaw/openclaw/issues/46681 | https://api.github.com/repos/openclaw/openclaw/issues/46681 | OpenClaw + gstack Integration - Phase 1 & 2 Complete (5/8 workflows) | # 🎉 OpenClaw + gstack Integration - Major Progress Update
## Status: Phase 1 & 2 Successfully Completed
### Phase 1: Core Workflow Engine ✅ 100% Complete
**Components Created (4 files, ~1,000 lines):**
- `workflow-manager.ts` (250 lines) - Multi-step workflow orchestration
- `workflow-types.ts` (200 lines) - TypeSc... | closed | not_planned | true | 2 | [] | [] | 2026-03-15T00:32:49Z | 2026-04-27T04:41:34Z | 2026-04-25T02:03:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | emilystardev | 265,783,721 | U_kgDOD9eJqQ | User | false |
openclaw/openclaw | 4,077,241,548 | I_kwDOQb6kR87zBcTM | 46,687 | https://github.com/openclaw/openclaw/issues/46687 | https://api.github.com/repos/openclaw/openclaw/issues/46687 | Memory file naming inconsistency between AGENTS.md template and session-memory hook | ## Problem
There's a naming mismatch between what AGENTS.md tells agents to read and what session-memory hook generates:
| Source | Format | Design |
|--------|--------|--------|
| **AGENTS.md template** | `YYYY-MM-DD.md` | One file per day, agent writes content |
| **session-memory hook** | `YYYY-MM-DD-slug.md` | On... | closed | completed | true | 1 | [] | [] | 2026-03-15T01:02:26Z | 2026-04-27T04:41:34Z | 2026-04-25T02:03:33Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | yaoduan | 17,714,063 | MDQ6VXNlcjE3NzE0MDYz | User | false |
openclaw/openclaw | 4,077,243,350 | I_kwDOQb6kR87zBcvW | 46,688 | https://github.com/openclaw/openclaw/issues/46688 | https://api.github.com/repos/openclaw/openclaw/issues/46688 | Extensions discovered from workspace/.openclaw/extensions/ without guard or warning | ## Bug Summary
Extensions placed inside `workspace/.openclaw/extensions/` are silently discovered and loaded by the gateway, even though the canonical path is `~/.openclaw/extensions/`. This creates ghost plugins that:
1. Load without proper install provenance (logged as "untracked local code" but still loaded)
2. Ar... | closed | completed | true | 1 | [] | [] | 2026-03-15T01:03:15Z | 2026-04-27T04:41:33Z | 2026-04-25T02:03:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,077,244,875 | I_kwDOQb6kR87zBdHL | 46,689 | https://github.com/openclaw/openclaw/issues/46689 | https://api.github.com/repos/openclaw/openclaw/issues/46689 | [Bug]:ontrol UI auto-pairs on loopback but is granted only admin/approvals/pairing scopes, then fails all read RPCs with `missing scope: operator.read` | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
n a macOS local-loopback install, the OpenClaw Control UI/dashboard successfully connects and is auto-paired as an operator device, but the paired device is granted only these scopes:
- `operator.admin`
- `operator.approvals`
- `operator.p... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T01:03:57Z | 2026-04-27T04:41:32Z | 2026-04-25T02:04:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | babafetta888 | 268,272,665 | U_kgDOD_2EGQ | User | false |
openclaw/openclaw | 4,077,280,311 | I_kwDOQb6kR87zBlw3 | 46,700 | https://github.com/openclaw/openclaw/issues/46700 | https://api.github.com/repos/openclaw/openclaw/issues/46700 | [Bug]: Invalid model ID after switching model | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to the 2026-03-12 version, I encountered an issue when switching to a new model using the session command.
### Steps to reproduce
1. Use the command /model openrouter/deepseek/deepseek_v3
2. Wait for a while, then ask the agent to c... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-15T01:22:14Z | 2026-04-27T04:41:32Z | 2026-04-25T02:04:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | liuandrew1997 | 180,085,287 | U_kgDOCrviJw | User | false |
openclaw/openclaw | 4,326,671,346 | I_kwDOQb6kR88AAAABAePD8g | 71,355 | https://github.com/openclaw/openclaw/issues/71355 | https://api.github.com/repos/openclaw/openclaw/issues/71355 | Plugin tools intermittently dropped from tools array between turns of the same session | # Plugin tools intermittently dropped from `tools` array between turns of the same session
## Summary
In `2026.4.23-beta.6`, the `tools` array sent to the model on a single Telegram session non-deterministically alternates between **14 tools** (core only) and **23 tools** (core + plugin) across turns of the same sess... | closed | completed | true | 1 | [] | [] | 2026-04-25T02:04:36Z | 2026-04-27T04:37:01Z | 2026-04-25T03:17:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | c-g14 | 59,234,834 | MDQ6VXNlcjU5MjM0ODM0 | User | false |
openclaw/openclaw | 4,077,294,920 | I_kwDOQb6kR87zBpVI | 46,709 | https://github.com/openclaw/openclaw/issues/46709 | https://api.github.com/repos/openclaw/openclaw/issues/46709 | [Bug]: Cloudflare AI Gateway - OpenClaw does not pass Anthropic cache control settings to gateway | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Hello,
I am using Cloudflare AI Gateway with Anthropic models. OpenClaw does not seem to pass the cache control settings in the model definitions in the requests to the AI gateway. When I view the request content in the gateway logs it sho... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T01:29:36Z | 2026-04-27T04:41:31Z | 2026-04-25T02:04:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | AaronFaby | 41,449,666 | MDQ6VXNlcjQxNDQ5NjY2 | User | false |
openclaw/openclaw | 4,077,312,370 | I_kwDOQb6kR87zBtly | 46,717 | https://github.com/openclaw/openclaw/issues/46717 | https://api.github.com/repos/openclaw/openclaw/issues/46717 | Gemini provider: function_response.name empty on session replay → 400 INVALID_ARGUMENT | ## Bug
When using a Gemini-backed agent (google-generative-ai provider, gemini-3.1-pro-preview), tool results are serialized into the session history with an empty `name` field in the `function_response` part. On session replay, the Google API rejects the request:
```
GenerateContentRequest.contents[2].parts[0].funct... | closed | completed | true | 3 | [] | [] | 2026-03-15T01:39:26Z | 2026-04-27T04:41:30Z | 2026-04-25T02:04:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Javacabra | 10,149,584 | MDQ6VXNlcjEwMTQ5NTg0 | User | false |
openclaw/openclaw | 4,077,313,829 | I_kwDOQb6kR87zBt8l | 46,719 | https://github.com/openclaw/openclaw/issues/46719 | https://api.github.com/repos/openclaw/openclaw/issues/46719 | sessions_yield with pending child subagents can be announced upstream as 'completed successfully' | ## Summary
A subagent that calls `sessions_yield()` after spawning child subagents can still be announced to its parent/grandparent as **"completed successfully"**, even though it is not actually finished and still has pending descendants.
In the observed case, the parent/orchestrator subagent yielded with a message ... | closed | completed | true | 4 | [] | [] | 2026-03-15T01:40:19Z | 2026-04-27T04:41:29Z | 2026-04-25T02:05:13Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | joeykrug | 5,925,937 | MDQ6VXNlcjU5MjU5Mzc= | User | false |
openclaw/openclaw | 4,077,322,007 | I_kwDOQb6kR87zBv8X | 46,723 | https://github.com/openclaw/openclaw/issues/46723 | https://api.github.com/repos/openclaw/openclaw/issues/46723 | Feishu: text content lost when sending image+text mixed message in group chat | ## Bug Description
When a user sends a mixed image+text message in a Feishu group chat, the text content is lost during dispatch to the agent. The agent only receives the `image_key` JSON, not the accompanying text.
## Steps to Reproduce
1. In a Feishu group chat with an OpenClaw bot
2. Send a message that contains ... | closed | not_planned | true | 1 | [] | [] | 2026-03-15T01:45:03Z | 2026-04-27T04:41:29Z | 2026-04-25T02:05:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 120106835 | 57,889,251 | MDQ6VXNlcjU3ODg5MjUx | User | false |
openclaw/openclaw | 4,077,354,269 | I_kwDOQb6kR87zB30d | 46,734 | https://github.com/openclaw/openclaw/issues/46734 | https://api.github.com/repos/openclaw/openclaw/issues/46734 | Model ID case transformation breaks gpt-5.3-codex-spark on ChatGPT OAuth accounts | ## Summary
When configuring `gpt-5.3-codex-spark` as the primary model in `openclaw.json`, the gateway transforms the model ID to `GPT-5.3-Codex-Spark` (title-cased) before sending it to the OpenAI API. The API rejects this with:
```
"The 'GPT-5.3-Codex-Spark' model is not supported when using Codex with a ChatGPT ac... | closed | not_planned | true | 2 | [] | [] | 2026-03-15T02:03:21Z | 2026-04-27T04:41:28Z | 2026-04-25T02:05:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | atd93 | 80,082,302 | MDQ6VXNlcjgwMDgyMzAy | User | false |
openclaw/openclaw | 4,077,361,449 | I_kwDOQb6kR87zB5kp | 46,736 | https://github.com/openclaw/openclaw/issues/46736 | https://api.github.com/repos/openclaw/openclaw/issues/46736 | [Bug]: WebChat Message Injection | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
# OpenClaw Bug Report: WebChat Message Injection
## Bug 描述
从 WebChat (`openclaw-control-ui`) 发送的消息会被错误地附加 "Skills store policy" 配置文本。
## 影响范围
- **受影响渠道**: WebChat (openclaw-control-ui)
- **正常渠道**: Telegram, ClawAI Relay (openclaw-macos)... | closed | not_planned | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T02:07:26Z | 2026-04-27T04:41:27Z | 2026-04-25T02:06:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | zoutaihua1975-dot | 260,827,014 | U_kgDOD4vnhg | User | false |
openclaw/openclaw | 4,077,385,448 | I_kwDOQb6kR87zB_bo | 46,743 | https://github.com/openclaw/openclaw/issues/46743 | https://api.github.com/repos/openclaw/openclaw/issues/46743 | Add option to disable local model discovery (Ollama) | For users who only use cloud-based models (like MiniMax), this warning is unnecessary and confusing.
## Proposed Solution
Add a configuration option to completely disable local model discovery. For example:
```json
{
"models": {
"localDiscovery": false
}
} | closed | completed | true | 1 | [] | [] | 2026-03-15T02:21:43Z | 2026-04-27T04:41:26Z | 2026-04-25T02:06:20Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lanmeiyishengforgit | 262,987,079 | U_kgDOD6zdRw | User | false |
openclaw/openclaw | 4,198,795,959 | I_kwDOQb6kR876RIq3 | 60,118 | https://github.com/openclaw/openclaw/issues/60118 | https://api.github.com/repos/openclaw/openclaw/issues/60118 | [Bug]: voice-call responseModel config ignored — LiveSessionModelSwitchError overrides to global default | ## Bug Description
The voice-call plugin's `responseModel` config option is ignored at runtime. When `responseModel` is set to a fast model (e.g., `openai/gpt-4o-mini` or `openai/gpt-4.1-nano`) but the global agent default model is different (e.g., `anthropic/claude-opus-4-6`), the embedded PI runner throws `LiveSessi... | closed | completed | true | 2 | [
"bug",
"channel: voice-call"
] | [] | 2026-04-03T06:51:54Z | 2026-04-27T04:41:25Z | 2026-04-25T02:06:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | xinbenlv | 640,325 | MDQ6VXNlcjY0MDMyNQ== | User | false |
openclaw/openclaw | 4,077,408,085 | I_kwDOQb6kR87zCE9V | 46,751 | https://github.com/openclaw/openclaw/issues/46751 | https://api.github.com/repos/openclaw/openclaw/issues/46751 | [Feature]:Prevent clicks on occluded elements (Z-Index/Overlays) to resolve Action-State Hallucinations | ### Summary
**One-line statement of the requested capability.**
Introduce a lightweight document.elementFromPoint pre-click validation to prevent the Agent from interacting with obscured elements and hallucinating success.
### Problem to solve
**What user pain this solves and why current behavior is insufficient.**... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-15T02:36:48Z | 2026-04-27T04:41:25Z | 2026-04-25T02:06:37Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | FangcunOS | 260,646,088 | U_kgDOD4kkyA | User | false |
openclaw/openclaw | 4,077,430,381 | I_kwDOQb6kR87zCKZt | 46,767 | https://github.com/openclaw/openclaw/issues/46767 | https://api.github.com/repos/openclaw/openclaw/issues/46767 | Managed OAuth connectors for common business integrations (QuickBooks, Salesforce, etc.) | ## Summary
OpenClaw currently requires users to manually handle OAuth flows for third-party business integrations. Competing platforms like Perplexity Computer offer one-click "login and approve" authorization that takes ~30 seconds. OpenClaw should offer the same.
## The Problem
Connecting an OpenClaw agent to QuickB... | closed | not_planned | true | 1 | [] | [] | 2026-03-15T02:53:56Z | 2026-04-27T04:41:24Z | 2026-04-25T02:06:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bo-blue | 257,455,029 | U_kgDOD1hztQ | User | false |
openclaw/openclaw | 4,077,431,109 | I_kwDOQb6kR87zCKlF | 46,768 | https://github.com/openclaw/openclaw/issues/46768 | https://api.github.com/repos/openclaw/openclaw/issues/46768 | Orchestrator sessions go idle after spawning sub-agents — no progress updates without manual ping | ## Problem
When an orchestrator agent spawns multiple sub-agents (e.g., a swarm of 8-15 coding agents), the orchestrator session goes idle immediately after dispatching. No progress updates are posted to the delivery channel until a human manually pings the session.
This is a fundamental architectural issue, not a be... | closed | completed | true | 2 | [] | [] | 2026-03-15T02:54:32Z | 2026-04-27T04:41:23Z | 2026-04-25T02:07:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bharatindia123-a11y | 252,865,299 | U_kgDODxJrEw | User | false |
openclaw/openclaw | 4,077,431,568 | I_kwDOQb6kR87zCKsQ | 46,769 | https://github.com/openclaw/openclaw/issues/46769 | https://api.github.com/repos/openclaw/openclaw/issues/46769 | Gateway RPC timeout causes intermittent `openclaw cron list` failures (1000 normal closure) | # Gateway RPC timeout causes intermittent `openclaw cron list` failures (1000 normal closure)
## Environment
- **OpenClaw Version**: 2026.3.13 (61d171a)
- **OS**: macOS 15.3.0 (Darwin Kernel 25.3.0)
- **Node.js**: v22.22.1
- **npm**: 10.9.4
- **Installation**: npm global install via nvm
- **Gateway**: LaunchAgent ser... | closed | completed | true | 1 | [] | [] | 2026-03-15T02:54:54Z | 2026-04-27T04:41:22Z | 2026-04-25T02:07:27Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | onion214 | 10,388,223 | MDQ6VXNlcjEwMzg4MjIz | User | false |
openclaw/openclaw | 4,077,433,614 | I_kwDOQb6kR87zCLMO | 46,772 | https://github.com/openclaw/openclaw/issues/46772 | https://api.github.com/repos/openclaw/openclaw/issues/46772 | Doctor fabricates phantom accounts.default in multi-account Telegram config even with no top-level fields | ## Summary
`openclaw doctor` reports "Moved channels.telegram single-account top-level values into channels.telegram.accounts.default" even when **no top-level values exist** (`dmPolicy`, `groupPolicy`, `streaming` are absent). It then warns about the phantom `accounts.default` it fabricated:
> channels.telegram.acco... | closed | completed | true | 1 | [] | [] | 2026-03-15T02:56:36Z | 2026-04-27T04:41:22Z | 2026-04-25T02:07:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | freivokh | 77,433,046 | MDQ6VXNlcjc3NDMzMDQ2 | User | false |
openclaw/openclaw | 4,077,455,042 | I_kwDOQb6kR87zCQbC | 46,783 | https://github.com/openclaw/openclaw/issues/46783 | https://api.github.com/repos/openclaw/openclaw/issues/46783 | Add context compression middleware to reduce token usage in long-running agents | ### Summary
Long-running agents accumulate large conversation history which increases token usage and cost.
As tasks run longer, context grows quickly and models start hitting token limits or burning excessive tokens.
A mechanism to compress or summarize context while preserving important information could significa... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-15T03:11:52Z | 2026-04-27T04:41:21Z | 2026-04-25T02:08:00Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Rothschildiuk | 15,833,131 | MDQ6VXNlcjE1ODMzMTMx | User | false |
openclaw/openclaw | 4,077,459,670 | I_kwDOQb6kR87zCRjW | 46,785 | https://github.com/openclaw/openclaw/issues/46785 | https://api.github.com/repos/openclaw/openclaw/issues/46785 | feat: Add maxAnnounceChars parameter to sessions_spawn | ## Problem
Sub-agent completion messages that exceed Telegram's 4096 character limit are silently dropped (related to upstream issue #28851 — sendPayload bypasses chunker).
The current mitigation is a protocol-level instruction in sub-agent prompts limiting output to ~3500 chars, but LLMs cannot reliably count charac... | closed | not_planned | true | 1 | [] | [] | 2026-03-15T03:15:06Z | 2026-04-27T04:41:20Z | 2026-04-25T02:08:17Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | salvador-clawi | 261,618,414 | U_kgDOD5f67g | User | false |
openclaw/openclaw | 4,326,351,573 | I_kwDOQb6kR88AAAABAd7i1Q | 71,300 | https://github.com/openclaw/openclaw/issues/71300 | https://api.github.com/repos/openclaw/openclaw/issues/71300 | WebChat screenshot paste fails — OSS URL blocked by SSRF policy | When pasting images directly into the WebChat input field, images are uploaded to Aliyun OSS and return a public URL (e.g. `https://xxx.oss-cn-wulanchabu.aliyuncs.com/...`). However, OpenClaw's image tool (and image generation tools) cannot fetch these images, returning:
`Blocked: resolves to private/internal/special-u... | closed | completed | true | 0 | [] | [] | 2026-04-25T00:18:47Z | 2026-04-27T04:41:19Z | 2026-04-25T02:08:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tonyxiaogui-cpu | 266,668,221 | U_kgDOD-UIvQ | User | false |
openclaw/openclaw | 4,077,466,303 | I_kwDOQb6kR87zCTK_ | 46,795 | https://github.com/openclaw/openclaw/issues/46795 | https://api.github.com/repos/openclaw/openclaw/issues/46795 | ACP sessions_spawn streamTo=parent stalls: acpx emits console.log JSON frames, parser drops them → no text_delta / no emitAgentEvent | ## Summary
`sessions_spawn(runtime: "acp", streamTo: "parent")` accepts the spawn and the child agent responds within seconds, but the parent session never receives the result. The parent relay only emits `start` + `stall` (60s idle) events.
## Root Cause
The acpx adapter parser (`extensions/acpx/src/runtime-interna... | closed | completed | true | 4 | [] | [] | 2026-03-15T03:19:50Z | 2026-04-27T04:41:18Z | 2026-04-25T02:08:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sumurtk2 | 53,351,545 | MDQ6VXNlcjUzMzUxNTQ1 | User | false |
openclaw/openclaw | 4,077,466,741 | I_kwDOQb6kR87zCTR1 | 46,798 | https://github.com/openclaw/openclaw/issues/46798 | https://api.github.com/repos/openclaw/openclaw/issues/46798 | [Bug]: Possible regression: heartbeat sessions can re-trigger on local exec completed events and spam duplicate heartbeat log entries | ### Bug type
Regression (worked before, now fails)
### Summary
I’m seeing what looks like a remaining or regressed variant of the earlier local exec / heartbeat-routing bugs.
The visible heartbeat cron job is correctly scheduled hourly, but heartbeat-related sessions appear to be re-triggered by local exec complete... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-15T03:20:10Z | 2026-04-27T04:41:17Z | 2026-04-25T02:08:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mogglemoss | 10,966,195 | MDQ6VXNlcjEwOTY2MTk1 | User | false |
openclaw/openclaw | 4,326,685,738 | I_kwDOQb6kR88AAAABAeP8Kg | 71,357 | https://github.com/openclaw/openclaw/issues/71357 | https://api.github.com/repos/openclaw/openclaw/issues/71357 | Compaction summary quality issues: snowballing, no cooldown, keepRecentTokens not working | ## Compaction Summary Quality Issues (v2026.4.23)
### Environment
- **Version**: OpenClaw 2026.4.23 (a979721)
- **Model**: opencode-go/glm-5.1 (primary), same for compaction
- **Context Window**: 198k tokens
- **Compaction mode**: safeguard
- **keepRecentTokens**: 50000 (config)
### Observed Issues
I've been running... | closed | completed | true | 1 | [] | [] | 2026-04-25T02:08:54Z | 2026-04-27T04:38:00Z | 2026-04-25T03:03:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | WhiteGiverMa | 152,406,589 | U_kgDOCRWKPQ | User | false |
openclaw/openclaw | 4,077,469,756 | I_kwDOQb6kR87zCUA8 | 46,804 | https://github.com/openclaw/openclaw/issues/46804 | https://api.github.com/repos/openclaw/openclaw/issues/46804 | ReferenceError: Cannot access 'ANTHROPIC_MODEL_ALIASES' before initialization on startup | **Version:** 2026.3.12
**Platform:** Linux (x64)
**Description:**
On every `openclaw` CLI invocation, the following error is printed to stderr before the command runs. Functionality is not affected — it appears to be non-fatal.
**Error:**
```
Failed to read config at /home/.../.openclaw/openclaw.json ReferenceError:... | closed | completed | true | 1 | [
"close:duplicate",
"close:already-fixed",
"dedupe:child"
] | [
"vincentkoc"
] | 2026-03-15T03:22:21Z | 2026-04-27T04:31:21Z | 2026-04-25T02:09:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | fin70bot | 268,273,160 | U_kgDOD_2GCA | User | false |
openclaw/openclaw | 4,077,473,884 | I_kwDOQb6kR87zCVBc | 46,806 | https://github.com/openclaw/openclaw/issues/46806 | https://api.github.com/repos/openclaw/openclaw/issues/46806 | Feature: External webhook → session event injection | ## Feature Request: External Webhook → Session Event
### Problem
External services (n8n, monitoring tools, etc.) cannot trigger events in an OpenClaw session. For example:
- n8n workflow error → want to alert the agent
- External webhook → want to wake the session
Currently, OpenClaw hooks only support specific prese... | closed | completed | true | 1 | [] | [] | 2026-03-15T03:25:16Z | 2026-04-27T04:41:16Z | 2026-04-25T02:09:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sabo961 | 191,984 | MDQ6VXNlcjE5MTk4NA== | User | false |
openclaw/openclaw | 4,077,474,994 | I_kwDOQb6kR87zCVSy | 46,811 | https://github.com/openclaw/openclaw/issues/46811 | https://api.github.com/repos/openclaw/openclaw/issues/46811 | Control UI: Cron job visual manager | ## Feature Request
### Summary
Add a cron job management panel to the Control UI so users can view, enable/disable, run, and delete scheduled jobs without using chat commands.
### Motivation
The cron system is powerful but invisible. Users who set reminders or recurring tasks have no way to see what is scheduled with... | closed | completed | true | 1 | [] | [] | 2026-03-15T03:25:59Z | 2026-04-27T04:41:15Z | 2026-04-25T02:09:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | allisonshawai-ctrl | 262,603,479 | U_kgDOD6cC1w | User | false |
openclaw/openclaw | 4,077,484,941 | I_kwDOQb6kR87zCXuN | 46,818 | https://github.com/openclaw/openclaw/issues/46818 | https://api.github.com/repos/openclaw/openclaw/issues/46818 | [Web UI] 白色三角形感叹号图标随机覆盖整个屏幕 | ## 问题描述
Web UI 界面上随机出现一个**白色三角形+感叹号 ❕** 的图标，完全覆盖了输入框和显示区域，导致无法使用。
## 重现步骤
1. 访问 Web UI (http://127.0.0.1:18789/chat?session=main)
2. 正常使用一段时间后，该三角形图标会随机出现
3. 点击该图标后消失
4. 问题会频繁复发
## 预期行为
- Web UI 应该正常显示聊天界面
- 不应该出现覆盖整个屏幕的三角形图标
## 实际行为
- 白色三角形+感叹号图标覆盖整个屏幕
- 图标内部为白色，有一个 ❕ 感叹号
- 背景为深色
- 点击后消失
- 刷新页面或重启 Gateway 后暂时恢复... | closed | completed | true | 8 | [] | [] | 2026-03-15T03:33:51Z | 2026-04-27T04:41:14Z | 2026-04-25T02:09:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Bingo117 | 155,721,502 | U_kgDOCUgfHg | User | false |
openclaw/openclaw | 4,077,485,841 | I_kwDOQb6kR87zCX8R | 46,820 | https://github.com/openclaw/openclaw/issues/46820 | https://api.github.com/repos/openclaw/openclaw/issues/46820 | [Bug]: openclaw models list takes 55s with heavy futex contention (73% CPU) — only 3 OpenRouter models configured | ### Bug type
Regression (worked before, now fails)
### Summary
## Environment
- OpenClaw: 2026.3.13 (61d171a)
- OS: Linux 6.1.0-43-amd64 (Debian, Virtualmin managed, bhyve)
- Hardware: 4 vCPUs (Xeon E5-2698 v4 @ 2.20GHz), 8GB RAM
- Node: v22.22.1
- Install: npm global (non-Docker)
## Problem
`openclaw models list` ... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-15T03:34:36Z | 2026-04-27T04:41:14Z | 2026-04-25T02:10:14Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 4044ever | 3,244,771 | MDQ6VXNlcjMyNDQ3NzE= | User | false |
openclaw/openclaw | 4,077,499,127 | I_kwDOQb6kR87zCbL3 | 46,826 | https://github.com/openclaw/openclaw/issues/46826 | https://api.github.com/repos/openclaw/openclaw/issues/46826 | Show & Tell: ComfyUI Bridge — generate images from iMessage/Discord via ComfyUI on your own hardware | ## What I built
A FastAPI bridge server + OpenClaw skill that connects a self-hosted [ComfyUI](https://github.com/comfyanonymous/ComfyUI) instance to any chat channel. Generate images, faceswap, animate expressions, and more — all from a chat message, on your own GPU.
```
Chat message → OpenClaw bot → Bridge server →... | closed | not_planned | true | 1 | [] | [] | 2026-03-15T03:45:44Z | 2026-04-27T04:41:13Z | 2026-04-25T02:10:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Bortlesboat | 169,967,362 | U_kgDOCiF_Ag | User | false |
openclaw/openclaw | 4,077,507,533 | I_kwDOQb6kR87zCdPN | 46,832 | https://github.com/openclaw/openclaw/issues/46832 | https://api.github.com/repos/openclaw/openclaw/issues/46832 | voice-call plugin EADDRINUSE error - webhook server tries to bind twice | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The voice-call plugin fails to start with EADDRINUSE error because the webhook server attempts to bind to the same port twice. This appears to be a regression or unresolved issue related to #8926, which was closed without fixing the voice-c... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T03:52:42Z | 2026-04-27T04:41:12Z | 2026-04-25T02:10:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | aspyyyn | 266,942,185 | U_kgDOD-k26Q | User | false |
openclaw/openclaw | 4,175,530,265 | I_kwDOQb6kR8744YkZ | 58,115 | https://github.com/openclaw/openclaw/issues/58115 | https://api.github.com/repos/openclaw/openclaw/issues/58115 | Bug: voice_call tool fails with EADDRINUSE when called from any agent session | # Bug: `voice_call` tool fails with EADDRINUSE when called from any session outside the gateway process
## Summary
All `voice_call` tool actions (`end_call`, `speak_to_user`, `get_status`, `continue_call`) fail with:
```
listen EADDRINUSE: address already in use 127.0.0.1:3334
```
when invoked from any agent sessio... | closed | completed | true | 11 | [
"bug",
"channel: voice-call"
] | [] | 2026-03-31T03:51:22Z | 2026-04-27T04:41:10Z | 2026-04-25T02:12:08Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | legonhilltech-jpg | 272,396,537 | U_kgDOEDxw-Q | User | false |
openclaw/openclaw | 4,165,090,126 | I_kwDOQb6kR874QjtO | 57,172 | https://github.com/openclaw/openclaw/issues/57172 | https://api.github.com/repos/openclaw/openclaw/issues/57172 | [voice-call] CLI and agent tool EADDRINUSE when gateway plugin is running — tool should RPC to gateway, not create its own runtime | ## Bug Report
### Environment
- OpenClaw version: 2026.3.28 (f9b1079)
- OS: macOS 15.5 (Apple Silicon M1)
- Node: v24.14.0
- Voice-call provider: Twilio
- Webhook exposure: Tailscale Funnel
### Description
When the voice-call plugin is running inside the Gateway (webhook server bound to port 3334), both the CLI (`o... | closed | completed | true | 2 | [] | [] | 2026-03-29T16:53:05Z | 2026-04-27T04:41:09Z | 2026-04-25T02:12:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | binarytested | 45,023,617 | MDQ6VXNlcjQ1MDIzNjE3 | User | false |
openclaw/openclaw | 4,165,163,100 | I_kwDOQb6kR874Q1hc | 57,186 | https://github.com/openclaw/openclaw/issues/57186 | https://api.github.com/repos/openclaw/openclaw/issues/57186 | voice-call: EADDRINUSE on port 3334 — tool/CLI tries to start webhook server already owned by gateway | ## Bug Description
When using the `voice_call` tool (or `openclaw voicecall call` CLI), it fails with:
```
Error: listen EADDRINUSE: address already in use 127.0.0.1:3334
```
## Root Cause
The gateway starts the voice-call webhook server on port 3334 during boot:
```
[voice-call] Webhook server listening on http:/... | closed | completed | true | 1 | [] | [] | 2026-03-29T17:10:39Z | 2026-04-27T04:41:06Z | 2026-04-25T02:15:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mrcyke | 38,428,133 | MDQ6VXNlcjM4NDI4MTMz | User | false |
openclaw/openclaw | 4,123,484,486 | I_kwDOQb6kR871x2FG | 53,126 | https://github.com/openclaw/openclaw/issues/53126 | https://api.github.com/repos/openclaw/openclaw/issues/53126 | voice-call plugin: CLI conflicts with gateway on port 3334 — voicecall.initiate RPC does not trigger call (source install) | ## Environment
- OpenClaw 2026.3.23 (df18063), **source install** (`git clone`)
- Linux (Ubuntu), Node 22.22.1
- Provider: Twilio (upgraded paid account)
- Tailscale Funnel for public webhook URL
## Expected behavior
`openclaw voicecall call --to "+1xxx" --mode notify --message "test"` initiates an outbound Twilio cal... | closed | completed | true | 1 | [] | [] | 2026-03-23T20:20:38Z | 2026-04-27T04:41:05Z | 2026-04-25T02:15:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | markasulzen | 196,271,753 | U_kgDOC7LeiQ | User | false |
openclaw/openclaw | 4,079,498,836 | I_kwDOQb6kR87zKDZU | 47,697 | https://github.com/openclaw/openclaw/issues/47697 | https://api.github.com/repos/openclaw/openclaw/issues/47697 | voicecall CLI: EADDRINUSE when gateway voice-call plugin is already running | ## Summary
`openclaw voicecall call` fails with `EADDRINUSE` when the voice-call plugin is already running inside the gateway, because both try to bind to the same webhook port.
## Steps to Reproduce
1. Configure voice-call plugin in `plugins.entries.voice-call`
2. Start gateway (plugin binds to `serve.port`, e.g. 8... | closed | completed | true | 2 | [] | [] | 2026-03-15T23:59:12Z | 2026-04-27T04:41:05Z | 2026-04-25T02:15:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mira-solari | 268,252,643 | U_kgDOD_014w | User | false |
openclaw/openclaw | 4,080,638,130 | I_kwDOQb6kR87zOZiy | 47,974 | https://github.com/openclaw/openclaw/issues/47974 | https://api.github.com/repos/openclaw/openclaw/issues/47974 | voice-call tool/CLI causes EADDRINUSE - tries to bind webhook port instead of using gateway | ## Description
The `voice_call` tool and `openclaw voicecall` CLI both fail with `EADDRINUSE` because they try to start their own webhook server instead of communicating with the running gateway process.
## Environment
- OpenClaw version: 2026.3.13
- Plugin version: @openclaw/voice-call 2026.3.13
- OS: Ubuntu Linux (... | closed | completed | true | 2 | [] | [] | 2026-03-16T07:02:42Z | 2026-04-27T04:41:04Z | 2026-04-25T02:15:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jack-jackhui | 65,952,706 | MDQ6VXNlcjY1OTUyNzA2 | User | false |
openclaw/openclaw | 4,077,513,332 | I_kwDOQb6kR87zCep0 | 46,838 | https://github.com/openclaw/openclaw/issues/46838 | https://api.github.com/repos/openclaw/openclaw/issues/46838 | Show & Tell: bitcoin-mcp + Satoshi API — 47 Bitcoin tools in any OpenClaw agent, zero config | ## What I built
Two skills that bring Bitcoin intelligence into any OpenClaw agent:
### 1. [satoshi-api](https://clawhub.ai/skills/satoshi-api) — Ask Bitcoin questions in chat
Fee recommendations, mempool status, BTC price, block info, address lookups, halving countdown — all from a chat message. No Bitcoin node, no... | closed | not_planned | true | 1 | [] | [] | 2026-03-15T03:56:49Z | 2026-04-27T04:41:03Z | 2026-04-25T02:16:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Bortlesboat | 169,967,362 | U_kgDOCiF_Ag | User | false |
openclaw/openclaw | 4,077,514,947 | I_kwDOQb6kR87zCfDD | 46,840 | https://github.com/openclaw/openclaw/issues/46840 | https://api.github.com/repos/openclaw/openclaw/issues/46840 | gateway stop 后再 start 需要重新安装 gateway | ## 问题描述
执行 `openclaw gateway stop` 停止服务后，再次执行 `openclaw gateway start` 启动服务时，提示需要重新安装 gateway，而不是直接启动已有的服务。
## 预期行为
```bash
openclaw gateway stop # 停止服务
openclaw gateway start # 应该直接启动已安装的服务
```
## 实际行为
```bash
openclaw gateway stop # 停止服务
openclaw gateway start # 提示需要重新安装 gateway
```
## 环境信息
- **系统**: mac... | closed | completed | true | 5 | [] | [] | 2026-03-15T03:57:51Z | 2026-04-27T04:41:02Z | 2026-04-25T02:16:59Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | teng2-doudou | 197,468,326 | U_kgDOC8Ugpg | User | false |
openclaw/openclaw | 4,077,516,985 | I_kwDOQb6kR87zCfi5 | 46,841 | https://github.com/openclaw/openclaw/issues/46841 | https://api.github.com/repos/openclaw/openclaw/issues/46841 | [Feature]: Telegram inline button callback routing to webhook (zero-token bypass) | ### Summary
Add an optional telegram.callbackRoutes config that routes specific callback_data patterns directly to a webhook URL, bypassing the agent session entirely.
### Problem to solve
Every Telegram inline button tap currently spawns an agent turn — callback_data is wrapped in a synthetic message and sent to pr... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-03-15T03:59:07Z | 2026-04-27T04:41:02Z | 2026-04-25T02:17:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TylerWanner | 3,697,600 | MDQ6VXNlcjM2OTc2MDA= | User | false |
openclaw/openclaw | 4,077,542,079 | I_kwDOQb6kR87zClq_ | 46,849 | https://github.com/openclaw/openclaw/issues/46849 | https://api.github.com/repos/openclaw/openclaw/issues/46849 | Bug: openclaw gateway status hangs / spawns nested gateway when invoked from inside running gateway service | # Bug draft: `openclaw gateway status` hangs / spawns nested gateway when invoked from inside the running gateway service context
## Summary
When `openclaw gateway status` is executed from an agent/tool process that is itself running under the live `openclaw-gateway` systemd user service, the command can hang, get ki... | closed | completed | true | 1 | [] | [] | 2026-03-15T04:12:51Z | 2026-04-27T04:41:01Z | 2026-04-25T02:17:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | leonliuyang | 1,224,875 | MDQ6VXNlcjEyMjQ4NzU= | User | false |
openclaw/openclaw | 4,077,569,537 | I_kwDOQb6kR87zCsYB | 46,862 | https://github.com/openclaw/openclaw/issues/46862 | https://api.github.com/repos/openclaw/openclaw/issues/46862 | Bug: Telegram Pairing Fails in a Loop | # Telegram Pairing Failure
**Summary:**
The setup process for the Telegram channel failed repeatedly despite extensive troubleshooting. The bot comes online and responds, but the final pairing/approval step consistently fails with a variety of contradictory and nonsensical error messages, suggesting a bug or corrupted... | closed | completed | true | 2 | [] | [] | 2026-03-15T04:26:05Z | 2026-04-27T04:41:00Z | 2026-04-25T02:17:49Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wesmagyar | 95,952,786 | U_kgDOBbgfkg | User | false |
openclaw/openclaw | 4,077,581,254 | I_kwDOQb6kR87zCvPG | 46,872 | https://github.com/openclaw/openclaw/issues/46872 | https://api.github.com/repos/openclaw/openclaw/issues/46872 | Feishu streaming cards: tool calls split reply into multiple cards | ## Bug
When using Feishu with streaming card mode (`renderMode: "card"`, `streaming: true`, `blockStreaming: true`), tool calls during a response cause the current card to close and a new card to be created. This results in a single logical reply being displayed as multiple separate card bubbles in Feishu.
### Enviro... | closed | completed | true | 1 | [] | [] | 2026-03-15T04:33:04Z | 2026-04-27T04:40:59Z | 2026-04-25T02:18:05Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ZenoRewn | 16,939,151 | MDQ6VXNlcjE2OTM5MTUx | User | false |
openclaw/openclaw | 4,077,589,063 | I_kwDOQb6kR87zCxJH | 46,878 | https://github.com/openclaw/openclaw/issues/46878 | https://api.github.com/repos/openclaw/openclaw/issues/46878 | doctor: improve memory search health check when memory plugin slot is active | ## Problem
When a memory plugin (e.g. `memory-neo4j`) owns the `plugins.slots.memory` slot, `openclaw doctor --fix` shows:
```
Memory search is explicitly disabled (enabled: false).
```
This is misleading. The built-in memory search is correctly disabled — it's been replaced by the plugin — but the doctor gives no i... | closed | completed | true | 1 | [] | [] | 2026-03-15T04:39:24Z | 2026-04-27T04:40:59Z | 2026-04-25T02:18:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tsukhani | 156,283 | MDQ6VXNlcjE1NjI4Mw== | User | false |
openclaw/openclaw | 4,077,597,591 | I_kwDOQb6kR87zCzOX | 46,882 | https://github.com/openclaw/openclaw/issues/46882 | https://api.github.com/repos/openclaw/openclaw/issues/46882 | [Bug] Write/Edit tools fail to create new files when tools.fs.workspaceOnly: true | ## Description
When an agent has `tools.fs.workspaceOnly: true` configured, the **Write tool fails for all new file creation** with the error:
```
path is not a regular file under root
```
This affects both relative and absolute paths within the workspace. Existing file edits are partially affected as well.
## Step... | closed | completed | true | 1 | [] | [] | 2026-03-15T04:46:12Z | 2026-04-27T04:40:57Z | 2026-04-25T02:18:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vincent-srp | 180,138,700 | U_kgDOCryyzA | User | false |
openclaw/openclaw | 4,077,601,174 | I_kwDOQb6kR87zC0GW | 46,885 | https://github.com/openclaw/openclaw/issues/46885 | https://api.github.com/repos/openclaw/openclaw/issues/46885 | [Bug] openclaw logs: WS handshake timeout due to event loop starvation during module loading | ## Summary
`openclaw logs` fails with `gateway connect failed: Error: gateway closed (1000):` on every invocation. The gateway is healthy — `openclaw doctor`, control UI, and all channels work fine.
## Root Cause
Event loop starvation race condition between lazy chunk loading and the gateway's 3-second WebSocket han... | closed | completed | true | 2 | [] | [] | 2026-03-15T04:49:01Z | 2026-04-27T04:40:56Z | 2026-04-25T02:18:55Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Oceanswave | 760,674 | MDQ6VXNlcjc2MDY3NA== | User | false |
openclaw/openclaw | 4,326,508,873 | I_kwDOQb6kR88AAAABAeFJSQ | 71,320 | https://github.com/openclaw/openclaw/issues/71320 | https://api.github.com/repos/openclaw/openclaw/issues/71320 | Bug: Tool call progress messages leaking to Telegram as visible messages | ## Description
When the agent makes tool calls (exec, read, etc.), the internal progress/working status messages are being exposed to Telegram chats as visible messages. Users see things like:
```
Working... • exec show...
Working... • exec ls ~/Desktop
Working... • read /path/to/file
```
These should be hidden inte... | closed | completed | true | 1 | [] | [] | 2026-04-25T01:09:55Z | 2026-04-27T04:40:56Z | 2026-04-25T02:19:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Vubration | 275,097,429 | U_kgDOEGWnVQ | User | false |
openclaw/openclaw | 4,077,605,425 | I_kwDOQb6kR87zC1Ix | 46,888 | https://github.com/openclaw/openclaw/issues/46888 | https://api.github.com/repos/openclaw/openclaw/issues/46888 | [Bug]: Telegram polling loop never initializes in proxy environments (China/HTTP_PROXY required) | # Bug: Telegram polling loop never initializes in proxy environments (China/HTTP_PROXY required)
## Summary
Telegram provider fails to initialize polling loop when HTTP_PROXY is required for Telegram API access (e.g., China region). The `bot.start()` method in grammY appears to not inherit the `HTTP_PROXY` environment... | closed | completed | true | 2 | [] | [] | 2026-03-15T04:52:42Z | 2026-04-27T04:40:55Z | 2026-04-25T02:19:12Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Quanle-Studio | 234,681,802 | U_kgDODfz1yg | User | false |
openclaw/openclaw | 4,077,619,918 | I_kwDOQb6kR87zC4rO | 46,897 | https://github.com/openclaw/openclaw/issues/46897 | https://api.github.com/repos/openclaw/openclaw/issues/46897 | `openclaw status` shows "missing scope: operator.read" on macOS local loopback after 2026.3.13 update | **Environment:**
- OS: macOS 26.3.1 (arm64)
- OpenClaw Version: 2026.3.13
- Node Version: v25.8.0
**Steps to Reproduce:**
1. Update to 2026.3.13 on a macOS machine.
2. Run `openclaw status` in the terminal.
**Expected Behavior:**
The CLI should successfully authenticate via local loopback and display the full gateway... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:04:16Z | 2026-04-27T04:40:54Z | 2026-04-25T02:19:28Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | acewong7 | 16,329,701 | MDQ6VXNlcjE2MzI5NzAx | User | false |
openclaw/openclaw | 4,077,641,048 | I_kwDOQb6kR87zC91Y | 46,913 | https://github.com/openclaw/openclaw/issues/46913 | https://api.github.com/repos/openclaw/openclaw/issues/46913 | Feature request: universal tool-result sanitizer hook | ## Problem
There's no single point where all tool results pass through before reaching the LLM context. Users who want to sanitize external content (prompt injection defense) must patch individual tool handlers separately — currently requiring 4+ patches across `wrapExternalContent`, `normalizeMessage`, `buildTextTool... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:21:24Z | 2026-04-27T04:40:53Z | 2026-04-25T02:19:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | esulcj | 11,564,095 | MDQ6VXNlcjExNTY0MDk1 | User | false |
openclaw/openclaw | 4,077,646,059 | I_kwDOQb6kR87zC_Dr | 46,917 | https://github.com/openclaw/openclaw/issues/46917 | https://api.github.com/repos/openclaw/openclaw/issues/46917 | [Feature Request] Add config option to control tool status output in DM | ## Problem
Since OpenClaw 3.13, tool execution status messages (like `Read with from xxx Completed`) are sent to users in DM channels. This behavior is hardcoded and cannot be disabled via configuration.
**Current behavior in source code** (`src/auto-reply/reply/dispatch-from-config.ts`):
```typescript
const shouldSe... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:25:26Z | 2026-04-27T04:40:53Z | 2026-04-25T02:20:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sunwedge | 13,697,363 | MDQ6VXNlcjEzNjk3MzYz | User | false |
openclaw/openclaw | 4,077,646,217 | I_kwDOQb6kR87zC_GJ | 46,918 | https://github.com/openclaw/openclaw/issues/46918 | https://api.github.com/repos/openclaw/openclaw/issues/46918 | 中文社区技术交流群 |  | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-15T05:25:35Z | 2026-04-27T04:40:52Z | 2026-04-25T02:20:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | aiqubits | 35,585,232 | MDQ6VXNlcjM1NTg1MjMy | User | false |
openclaw/openclaw | 4,077,657,380 | I_kwDOQb6kR87zDB0k | 46,921 | https://github.com/openclaw/openclaw/issues/46921 | https://api.github.com/repos/openclaw/openclaw/issues/46921 | [Feature Request] Add configurable Permissions-Policy header for Control UI microphone access | ## Problem
The Control UI (WebChat) shows a microphone button, but clicking it immediately shows a slash/disabled icon and disappears. This is because the Gateway sets a default HTTP response header:
```
Permissions-Policy: microphone=()
```
This security header blocks browser microphone access, making voice input i... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:34:15Z | 2026-04-27T04:40:51Z | 2026-04-25T02:20:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sunwedge | 13,697,363 | MDQ6VXNlcjEzNjk3MzYz | User | false |
openclaw/openclaw | 4,077,664,342 | I_kwDOQb6kR87zDDhW | 46,923 | https://github.com/openclaw/openclaw/issues/46923 | https://api.github.com/repos/openclaw/openclaw/issues/46923 | Bug: CLI 连接导臻 gateway 退出 | **描述**: 执行 openclaw CLI 命令（如 `openclaw cron list`）时，gateway 会崩溫退出。
**复现步骨**:
1. 启动 gateway（openclaw gateway start）
2. 执行 CLI 命令（openclaw cron list）
3. Gateway 退出
**错误信息**:
```
gateway connect failed: Error: gateway closed (1000):
Error: gateway closed (1000 normal closure): no close reason
```
**环境**:
- Windows
- O... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:38:27Z | 2026-04-27T04:40:51Z | 2026-04-25T02:20:51Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Furingg | 93,340,739 | U_kgDOBZBEQw | User | false |
openclaw/openclaw | 4,077,681,437 | I_kwDOQb6kR87zDHsd | 46,937 | https://github.com/openclaw/openclaw/issues/46937 | https://api.github.com/repos/openclaw/openclaw/issues/46937 | [Bug]: Agent model auto-switches to zenmux without user input, causing 402 quota errors | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Bug Description
When starting a new session for `discord-agent-coder`, the model automatically switches from the configured `openai-codex/gpt-5.3-codex` to `zenmux/anthropic/claude-opus-4.6` immediately after session startup, without an... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T05:52:52Z | 2026-04-27T04:40:49Z | 2026-04-25T02:21:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vincez-builder | 7,855,442 | MDQ6VXNlcjc4NTU0NDI= | User | false |
openclaw/openclaw | 4,077,683,646 | I_kwDOQb6kR87zDIO- | 46,941 | https://github.com/openclaw/openclaw/issues/46941 | https://api.github.com/repos/openclaw/openclaw/issues/46941 | Bug: /new re-fires command:new hooks because maybeEmitMissingResetHooks ignores resetHookTriggered mutation | # Bug Report: `/new` command triggers `command:new` hook multiple times
## Summary
When executing `/new` or `/reset` commands, the `command:new` hook is triggered multiple times for the same session, causing plugins that listen to this hook (e.g., `memory-lancedb-pro` reflection) to execute redundantly.
## Environmen... | closed | completed | true | 1 | [] | [] | 2026-03-15T05:54:48Z | 2026-04-27T04:40:48Z | 2026-04-25T02:21:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Louise-Qiuqiu | 253,923,225 | U_kgDODyKPmQ | User | false |
openclaw/openclaw | 4,077,692,314 | I_kwDOQb6kR87zDKWa | 46,951 | https://github.com/openclaw/openclaw/issues/46951 | https://api.github.com/repos/openclaw/openclaw/issues/46951 | openclaw logs fails with handshake timeout when TypeScript plugins are installed | ## Summary
`openclaw logs` (and potentially other non-model CLI commands) fails with `gateway closed (1000)` when a TypeScript plugin is installed. The root cause is an eager context-window warmup that triggers full plugin discovery at module import time, blocking the event loop during the gateway WebSocket handshake.... | closed | completed | true | 2 | [] | [] | 2026-03-15T06:02:39Z | 2026-04-27T04:40:48Z | 2026-04-25T02:21:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | KrE80r | 13,027,962 | MDQ6VXNlcjEzMDI3OTYy | User | false |
openclaw/openclaw | 4,077,728,217 | I_kwDOQb6kR87zDTHZ | 46,964 | https://github.com/openclaw/openclaw/issues/46964 | https://api.github.com/repos/openclaw/openclaw/issues/46964 | [Feature]: DO WE NEED A VSCODE PLUGIN FOR THIS? | ### Summary
Literally meaning.
I just made a VSCode Plugin called `openclaw luna`, and this plugin is hungerly needed to be experienced and issued asap 'cause there's still 0 issues for that plugin.
This plugin will play the role of a console that support the majority of functions for openclaw. And the most essenti... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-15T06:26:04Z | 2026-04-27T04:40:46Z | 2026-04-25T02:21:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | LunaticLegacy | 125,584,680 | U_kgDOB3xFKA | User | false |
openclaw/openclaw | 4,077,732,591 | I_kwDOQb6kR87zDULv | 46,966 | https://github.com/openclaw/openclaw/issues/46966 | https://api.github.com/repos/openclaw/openclaw/issues/46966 | [Bug]: Gateway dev-build startup has ~80s silent period before first log line | ## Summary
After a full process restart (`supervisor restart` mode), the gateway process spawns but emits **zero log output for ~80 seconds**, then all startup log lines appear simultaneously. Makes it impossible to diagnose startup issues and causes `openclaw gateway restart` to appear hung.
## Environment
- **Vers... | closed | completed | true | 1 | [] | [] | 2026-03-15T06:29:59Z | 2026-04-27T04:40:45Z | 2026-04-25T02:22:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jellyAI-dev | 263,263,160 | U_kgDOD7ETuA | User | false |
openclaw/openclaw | 4,077,748,719 | I_kwDOQb6kR87zDYHv | 46,975 | https://github.com/openclaw/openclaw/issues/46975 | https://api.github.com/repos/openclaw/openclaw/issues/46975 | [Bug]: Discord /model picker drops openrouter / prefix for OpenRouter nested ids (deepseek/deepseek-v3.2) → "model not allowed" | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
On Discord, using the '/model' interactive picker and selecting the DeepSeek option shown as something like:
'deepseek/deepseek-v3.2 • openrouter' and other openrouter model
### Steps to reproduce
configure more than one or probably any o... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T06:39:59Z | 2026-04-27T04:40:45Z | 2026-04-25T02:22:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | omar-0000 | 24,511,132 | MDQ6VXNlcjI0NTExMTMy | User | false |
openclaw/openclaw | 4,077,785,858 | I_kwDOQb6kR87zDhMC | 47,001 | https://github.com/openclaw/openclaw/issues/47001 | https://api.github.com/repos/openclaw/openclaw/issues/47001 | 建议增加find-skills镜像安装源解决GitHub限流问题 | ## 问题描述
在国内安装find-skills技能时，由于GitHub API限流和网络访问问题，经常会遇到：
1. `npx clawhub install find-skills` 报 Rate limit exceeded
2. `git clone` 访问超时或连接重置
## 建议方案
建议在文档中增加国内镜像加速源的安装方式，比如：
```bash
# 使用GitHub镜像加速克隆
cd ~/.openclaw/workspace/skills
git clone https://github.moeyy.xyz/https://github.com/openclaw/openclaw-find-skills.g... | closed | not_planned | true | 2 | [] | [] | 2026-03-15T07:03:17Z | 2026-04-27T04:40:44Z | 2026-04-25T02:22:46Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerryMa007 | 93,064,573 | U_kgDOBYwNfQ | User | false |
openclaw/openclaw | 4,077,789,153 | I_kwDOQb6kR87zDh_h | 47,004 | https://github.com/openclaw/openclaw/issues/47004 | https://api.github.com/repos/openclaw/openclaw/issues/47004 | [Bug]: | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
代码块
√
1 【Bug 报告】Synology Chat 插件无法发送回复
3 环境：
4﹣群晖 DS925+
5 - OpenClaw Docker 版本：2026.3.13
6 -Synology Chat Server
8 问题：
9- 消息接收正常
10 - 回复发送失败
11- 日志：fetchChatUsers: API returned success=false
12 - 手动调用 API 成功（curl 测试通过）
13 ﹣插件调用失败
14
15 尝试... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T07:05:15Z | 2026-04-27T04:40:43Z | 2026-04-25T02:23:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | gdlove0825 | 49,310,061 | MDQ6VXNlcjQ5MzEwMDYx | User | false |
openclaw/openclaw | 4,077,812,009 | I_kwDOQb6kR87zDnkp | 47,019 | https://github.com/openclaw/openclaw/issues/47019 | https://api.github.com/repos/openclaw/openclaw/issues/47019 | [Bug]: Agent-level `skills` filters do not constrain injected skills prompt | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When an agent is configured with `skills: []`, OpenClaw still appears to inject the full skills prompt block for that agent, so agent-level skill filtering does not seem to affect final prompt assembly as expected.
This may be either:
1. pr... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T07:21:11Z | 2026-04-27T04:40:42Z | 2026-04-25T02:23:19Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | danielwanwx | 144,515,713 | U_kgDOCJ0igQ | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.