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 | 4,152,465,670 | I_kwDOQb6kR873gZkG | 55,794 | https://github.com/openclaw/openclaw/issues/55794 | https://api.github.com/repos/openclaw/openclaw/issues/55794 | [Bug]: WhatsApp group messages stopped routing to agent sessions (regression: March 12, 2026) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
WhatsApp group messages stopped routing to agent sessions on March 12, 2026. Group messages no longer create sessions or trigger agent responses. Working on March 9 (verified screenshot), broken since March 12. Current versi... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-27T13:03:26Z | 2026-04-28T05:01:03Z | 2026-04-25T16:07:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Turk1509 | 171,039,641 | U_kgDOCjHbmQ | User | false |
openclaw/openclaw | 4,152,782,242 | I_kwDOQb6kR873hm2i | 55,811 | https://github.com/openclaw/openclaw/issues/55811 | https://api.github.com/repos/openclaw/openclaw/issues/55811 | Telegram sendChatAction retry spam during transient failures | **Problem:** When Telegram API returns transient network errors (e.g., rate limits or temporary connectivity issues), OpenClaw's Telegram plugin retries every 3 seconds with ERROR-level logs, spamming the log file.
**Observed behavior:**
- Log entries every 3 seconds: `telegram sendChatAction failed: Network request f... | closed | not_planned | true | 4 | [] | [] | 2026-03-27T13:42:57Z | 2026-04-28T05:01:02Z | 2026-04-25T16:08:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jeffvsutherland | 178,111,826 | U_kgDOCp3FUg | User | false |
openclaw/openclaw | 4,153,148,207 | I_kwDOQb6kR873jAMv | 55,829 | https://github.com/openclaw/openclaw/issues/55829 | https://api.github.com/repos/openclaw/openclaw/issues/55829 | exec tool creates orphaned grep processes that accumulate over time | ## Bug Description
** exec tool 执行管道命令时，管道中间的子进程（如 grep）会变成孤儿进程（orphaned），挂在 launchd 下不被回收。**
## Steps to Reproduce
1. 使用 exec 工具执行包含管道的命令，例如：
```
ps aux | grep chrome | grep -v grep
```
2. 多次调用后，系统会累积大量孤儿 grep 进程
3. 这些进程的父进程（PPID）显示为 1（launchd），即已被父进程抛弃
4. 只有执行 `openclaw gateway restart` 才能清理这些孤儿进程
## Exp... | closed | not_planned | true | 1 | [] | [] | 2026-03-27T14:22:04Z | 2026-04-28T05:01:01Z | 2026-04-25T16:08:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ms7-git | 264,172,505 | U_kgDOD77z2Q | User | false |
openclaw/openclaw | 4,153,975,607 | I_kwDOQb6kR873mKM3 | 55,890 | https://github.com/openclaw/openclaw/issues/55890 | https://api.github.com/repos/openclaw/openclaw/issues/55890 | [Bug] autoThreadName config not recognized despite being in v2026.3.24 release notes | ## Description
The v2026.3.24 release notes mention:
> Discord/auto threads: add optional autoThreadName: "generated" naming
However, when adding `autoThreadName: "generated"` to the Discord channel or guild config, the gateway rejects it with:
```
Invalid config at /home/local/.openclaw/openclaw.json:
- channels.disc... | closed | not_planned | true | 1 | [] | [] | 2026-03-27T16:05:01Z | 2026-04-28T05:01:00Z | 2026-04-25T16:08:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | BruceHowells | 13,401,442 | MDQ6VXNlcjEzNDAxNDQy | User | false |
openclaw/openclaw | 4,154,622,677 | I_kwDOQb6kR873ooLV | 55,934 | https://github.com/openclaw/openclaw/issues/55934 | https://api.github.com/repos/openclaw/openclaw/issues/55934 | [Feature]: Log rate limit headers from API responses into session .jsonl files | ## Summary
When OpenClaw makes LLM API calls, providers return rate limit data in response headers (e.g. `anthropic-ratelimit-requests-remaining`, `anthropic-ratelimit-unified-5h-utilization`). This data is currently discarded and not logged anywhere accessible to skills or users.
**Request:** Capture rate limit head... | closed | completed | true | 1 | [] | [] | 2026-03-27T17:24:54Z | 2026-04-28T05:00:59Z | 2026-04-25T16:08:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | vanhuelsing | 264,355,079 | U_kgDOD8G9Bw | User | false |
openclaw/openclaw | 4,154,874,672 | I_kwDOQb6kR873plsw | 55,954 | https://github.com/openclaw/openclaw/issues/55954 | https://api.github.com/repos/openclaw/openclaw/issues/55954 | Feature: agent-optimized CLI mode for programmatic/tool calls | ## Problem
When agents invoke `openclaw` CLI commands via `exec` or tool calls (e.g. `openclaw config get`, `openclaw agents list`, `openclaw cron list`), the current CLI architecture causes systemic problems:
1. **Cold-start latency**: Every `openclaw` invocation initializes ALL plugins from scratch before sending t... | closed | completed | true | 1 | [] | [] | 2026-03-27T18:03:39Z | 2026-04-28T05:00:59Z | 2026-04-25T16:08:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cygnostik | 5,272,292 | MDQ6VXNlcjUyNzIyOTI= | User | false |
openclaw/openclaw | 4,154,935,123 | I_kwDOQb6kR873p0dT | 55,964 | https://github.com/openclaw/openclaw/issues/55964 | https://api.github.com/repos/openclaw/openclaw/issues/55964 | Telegram replies intent text but does not execute tools (no exec/read/write traces) on 2026.3.24 | ### Environment
- OpenClaw: `2026.3.24 (cff6dc9)`
- OS: Linux (x64), Node 22
- Runtime: systemd gateway
- Model: `openai-codex/gpt-5.3-codex` (probe ok)
- Config: `tools.profile = coding`
### Symptom
Agent responds in Telegram with intent text (e.g. “Running it now…”) but does not execute tool actions and does not ret... | closed | completed | true | 3 | [] | [] | 2026-03-27T18:14:37Z | 2026-04-28T05:00:58Z | 2026-04-25T16:08:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kirinraw | 269,861,501 | U_kgDOEBXCfQ | User | false |
openclaw/openclaw | 4,155,758,136 | I_kwDOQb6kR873s9Y4 | 56,021 | https://github.com/openclaw/openclaw/issues/56021 | https://api.github.com/repos/openclaw/openclaw/issues/56021 | Shared tool-result ID sanitization/replay can emit orphan or invalid tool IDs across provider paths | ### Summary
OpenClaw can get a channel/session into a poisoned state where harmless follow-up turns fail immediately with tool-call/tool-result ID mismatch errors.
Originally this looked specific to Discord-backed `github-copilot/gpt-5.4` sessions on the OpenAI Responses path, but newer investigation suggests the bug... | closed | completed | true | 1 | [] | [] | 2026-03-27T20:33:46Z | 2026-04-28T05:00:57Z | 2026-04-25T16:08:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | slideshow-dingo | 258,996,445 | U_kgDOD2_43Q | User | false |
openclaw/openclaw | 4,155,933,435 | I_kwDOQb6kR873toL7 | 56,038 | https://github.com/openclaw/openclaw/issues/56038 | https://api.github.com/repos/openclaw/openclaw/issues/56038 | [Feature Request]: 微信渠道支持发送TTS语音消息为原生语音条 | ## 情况描述
使用 openclaw-weixin 插件(v2.0.0) 配合 Edge TTS 或 ElevenLabs 生成语音消息时，发送的TTS只能作为MP3音频文件在微信中播放，而不是原生语音条（语音条模式）。
## 期望行为
希望TTS生成的语音能够以微信原生语音条格式发送，类似WhatsApp的voice note，用户可以直接点击播放。
## 技术背景
- 微信消息API支持 type=3 (VOICE, SILK编码)
- 插件的 silk-transcode.ts 已包含SILK转码能力
- 但 send.ts 中尚未实现发送语音消息的函数
- WhatsApp渠道已有类似修复 (PR #48428)... | closed | not_planned | true | 1 | [] | [] | 2026-03-27T21:08:04Z | 2026-04-28T05:00:56Z | 2026-04-25T16:08:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | yj3294666927-gif | 268,633,304 | U_kgDOEAME2A | User | false |
openclaw/openclaw | 4,156,082,933 | I_kwDOQb6kR873uMr1 | 56,047 | https://github.com/openclaw/openclaw/issues/56047 | https://api.github.com/repos/openclaw/openclaw/issues/56047 | Feature: expose Anthropic unified rate-limit headers in session status / API | ## Problem
When using Claude Max (or any plan with unified rate limits), Anthropic returns `anthropic-ratelimit-unified-*` headers on every API response. These headers show the remaining budget (percentage/tokens) within the current rate-limit window — critical for understanding how much capacity is left.
Currently, ... | closed | completed | true | 1 | [] | [] | 2026-03-27T21:41:25Z | 2026-04-28T05:00:56Z | 2026-04-25T16:08:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nimzuk | 32,705,170 | MDQ6VXNlcjMyNzA1MTcw | User | false |
openclaw/openclaw | 4,156,166,781 | I_kwDOQb6kR873uhJ9 | 56,051 | https://github.com/openclaw/openclaw/issues/56051 | https://api.github.com/repos/openclaw/openclaw/issues/56051 | feat: agent-level accountId binding for multi-bot Discord setups | ### Problem
In multi-agent setups where each agent has its own Discord bot (separate token, identity, avatar), there is no way to permanently bind an agent to a specific Discord account. The current system relies on per-channel `bindings` to route inbound messages, but **outbound delivery** uses whatever `accountId` w... | closed | completed | true | 1 | [] | [] | 2026-03-27T22:03:00Z | 2026-04-28T05:00:55Z | 2026-04-25T16:08:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | DZ-Labz | 186,865,784 | U_kgDOCyNYeA | User | false |
openclaw/openclaw | 4,156,739,567 | I_kwDOQb6kR873ws_v | 56,097 | https://github.com/openclaw/openclaw/issues/56097 | https://api.github.com/repos/openclaw/openclaw/issues/56097 | feat: Add Tavily Crawl, Map, and Research API support | ## Summary
OpenClaw currently supports `tavily_search` and `tavily_extract`, but Tavily provides additional powerful APIs that would be valuable for users:
- **Crawl API** (`/crawl`) - Crawl multiple pages from a website with instructions-based discovery
- **Map API** (`/map`) - Map website structure and discover pag... | closed | not_planned | true | 1 | [] | [] | 2026-03-28T00:53:11Z | 2026-04-28T05:00:54Z | 2026-04-25T16:09:05Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mangcai1986 | 268,647,120 | U_kgDOEAM60A | User | false |
openclaw/openclaw | 4,156,751,033 | I_kwDOQb6kR873wvy5 | 56,100 | https://github.com/openclaw/openclaw/issues/56100 | https://api.github.com/repos/openclaw/openclaw/issues/56100 | Feature Request: Memory System Initialization for Ephemeral A2A Sessions | ## Problem
When using `sessions_spawn` with `mode: "run"` for native agent-to-agent (A2A) messaging, the spawned ephemeral sessions do not initialize **any** memory system. This means agent-to-agent communication is effectively memory-less, unlike human-to-agent sessions where memory persists automatically.
## Curren... | closed | not_planned | true | 1 | [] | [] | 2026-03-28T00:56:25Z | 2026-04-28T05:00:54Z | 2026-04-25T16:09:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Taimylife | 15,369,649 | MDQ6VXNlcjE1MzY5NjQ5 | User | false |
openclaw/openclaw | 4,157,035,554 | I_kwDOQb6kR873x1Qi | 56,125 | https://github.com/openclaw/openclaw/issues/56125 | https://api.github.com/repos/openclaw/openclaw/issues/56125 | Feature: Per-task model routing (route simple tasks to cheaper models) | ## Summary
Allow OpenClaw to route messages to different models based on task complexity, skill, or explicit configuration — so simple tasks (weather, reminders, smart home, basic Q&A) use a cheaper/faster model like Haiku while complex tasks (research, coding, financial analysis) use Sonnet or Opus.
## Use Case
I r... | closed | not_planned | true | 1 | [] | [] | 2026-03-28T02:34:19Z | 2026-04-28T05:00:53Z | 2026-04-25T16:09:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jaysturm | 1,745,678 | MDQ6VXNlcjE3NDU2Nzg= | User | false |
openclaw/openclaw | 3,949,557,178 | I_kwDOQb6kR87raXW6 | 18,565 | https://github.com/openclaw/openclaw/issues/18565 | https://api.github.com/repos/openclaw/openclaw/issues/18565 | [Feature]: Per-user context files for single-agent, multi-user setups | **_Note: I obviously used OpenClaw to rework/enhance the description of this issue, but I promise you that there is a human being and a real request behind it 😄_**
### Summary
Allow a single agent to maintain separate `USER.md`-style context per user via a `users/` directory in the workspace.
### Problem to solve
... | closed | not_planned | true | 6 | [
"enhancement"
] | [] | 2026-02-16T21:01:28Z | 2026-04-28T04:44:08Z | 2026-04-25T16:17:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Thanaen | 16,762,369 | MDQ6VXNlcjE2NzYyMzY5 | User | false |
openclaw/openclaw | 4,024,236,493 | I_kwDOQb6kR87v3PnN | 34,999 | https://github.com/openclaw/openclaw/issues/34999 | https://api.github.com/repos/openclaw/openclaw/issues/34999 | True Multi-Agent Group Chat: Shared Session Context with Coordinated Responses | ### Summary
Enable multiple agents to share a single conversation session, allowing simultaneous/coordinated responses with shared memory—transforming N separate agent conversations into one collaborative dialogue.
### Problem to solve
Current Pain:
Users like myself run multi-agent "Trinity" setups (specialized age... | closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-03-04T21:23:24Z | 2026-04-28T04:44:07Z | 2026-04-25T16:17:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | maverikva | 9,299,518 | MDQ6VXNlcjkyOTk1MTg= | User | false |
openclaw/openclaw | 4,158,814,292 | I_kwDOQb6kR8734nhU | 56,303 | https://github.com/openclaw/openclaw/issues/56303 | https://api.github.com/repos/openclaw/openclaw/issues/56303 | [Bug]: sessions_spawn tool hardcodes channel names instead of using plugin metadata | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The `sessions_spawn` tool in `pi-embedded-BaSvmUpW.js` hardcodes channel-specific logic (e.g., `opts?.agentChannel === "wechat"`), requiring code modifications for each new channel (feishu, telegram, discord, e... | closed | not_planned | true | 3 | [
"bug"
] | [] | 2026-03-28T09:09:25Z | 2026-04-28T05:00:07Z | 2026-04-25T16:17:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | caibutou | 28,855,802 | MDQ6VXNlcjI4ODU1ODAy | User | false |
openclaw/openclaw | 4,158,846,233 | I_kwDOQb6kR8734vUZ | 56,306 | https://github.com/openclaw/openclaw/issues/56306 | https://api.github.com/repos/openclaw/openclaw/issues/56306 | [Bug]: Telegram: read tool for images succeeds internally but fails to deliver file to user. | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When an agent is instructed to send an image file to a user on Telegram, the read(file="...") tool is executed. The agent's internal logs show that the tool successfully reads the image file from its local storage. However, ... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-28T09:15:18Z | 2026-04-28T05:00:06Z | 2026-04-25T16:18:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hoangh20 | 130,381,293 | U_kgDOB8V17Q | User | false |
openclaw/openclaw | 4,159,568,003 | I_kwDOQb6kR8737fiD | 56,371 | https://github.com/openclaw/openclaw/issues/56371 | https://api.github.com/repos/openclaw/openclaw/issues/56371 | [Bug]: WebUI message deletion is not persistent — deleted messages reappear after new message | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
In the WebUI chat interface, clicking the delete button on a message appears to remove it visually, but the message reappears after the next message is sent or the page is refreshed. The deletion is not persist... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-28T11:55:59Z | 2026-04-28T05:00:05Z | 2026-04-25T16:18:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | heichaowo | 22,045,978 | MDQ6VXNlcjIyMDQ1OTc4 | User | false |
openclaw/openclaw | 4,160,220,107 | I_kwDOQb6kR8739-vL | 56,441 | https://github.com/openclaw/openclaw/issues/56441 | https://api.github.com/repos/openclaw/openclaw/issues/56441 | Security: Provider API keys leak to all child processes spawned by exec tool | ## Summary
Resolved provider API keys (from `models.providers.<name>.apiKey`) are injected into `process.env` at startup and subsequently **inherited by every child process** spawned via the `exec` tool. This means any binary invoked by an agent — including untrusted or third-party CLIs — receives API keys it was neve... | closed | not_planned | true | 3 | [] | [] | 2026-03-28T14:39:22Z | 2026-04-28T05:00:05Z | 2026-04-25T16:18:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | toruvieI | 258,645,309 | U_kgDOD2qdPQ | User | false |
openclaw/openclaw | 4,160,306,980 | I_kwDOQb6kR873-T8k | 56,450 | https://github.com/openclaw/openclaw/issues/56450 | https://api.github.com/repos/openclaw/openclaw/issues/56450 | [Bug]: doctor --fix generates thousands of clobbered backup files | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
**Describe the bug**
doctor --fix generates thousands of clobbered backup files
**To Reproduce**
1. Corrupt config file (or have config validation fail)
2. Run `openclaw doctor --fix --yes`
3. Observe thousand... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-28T15:04:47Z | 2026-04-28T05:00:04Z | 2026-04-25T16:18:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wzwjs | 110,545,347 | U_kgDOBpbJww | User | false |
openclaw/openclaw | 4,160,354,685 | I_kwDOQb6kR873-fl9 | 56,460 | https://github.com/openclaw/openclaw/issues/56460 | https://api.github.com/repos/openclaw/openclaw/issues/56460 | Feature: Usage Dashboard — Token & Cost Tracking per Model/Day | ## Problem
OpenClaw currently stores per-session snapshots of token usage and estimated cost, but there is no way to view aggregated usage over time. Users have no visibility into:
- How many tokens they consumed per model per day/week/month
- Cost breakdown by provider (Anthropic, OpenAI, Gemini, etc.)
- Which sessi... | closed | completed | true | 3 | [] | [] | 2026-03-28T15:19:06Z | 2026-04-28T05:00:02Z | 2026-04-25T16:18:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kmgroen | 161,303,532 | U_kgDOCZ1L7A | User | false |
openclaw/openclaw | 4,160,475,866 | I_kwDOQb6kR873-9La | 56,478 | https://github.com/openclaw/openclaw/issues/56478 | https://api.github.com/repos/openclaw/openclaw/issues/56478 | [Bug]: WeChat plugin fails to send messages from isolated cron sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
WeChat plugin fails to send messages from isolated cron sessions. The message tool returns a messageId but the message is never delivered to WeChat. This affects all cron jobs using sessionTarget: "isolated" wi... | closed | not_planned | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-28T15:42:46Z | 2026-04-28T05:00:02Z | 2026-04-25T16:18:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Etmnb | 140,947,885 | U_kgDOCGaxrQ | User | false |
openclaw/openclaw | 4,160,574,563 | I_kwDOQb6kR873_VRj | 56,489 | https://github.com/openclaw/openclaw/issues/56489 | https://api.github.com/repos/openclaw/openclaw/issues/56489 | [Bug]: System events and cron payloads appear as user messages in webchat (still broken in 2026.3.23-2) | ## Summary
System-injected events (cron job payloads, exec completion notifications) appear in the webchat UI as messages sent by the user ("You"), derailing conversations and polluting chat history. This was previously reported in #20824 and closed as stale without a fix.
## Environment
- **OpenClaw version:** 2026... | closed | completed | true | 1 | [] | [] | 2026-03-28T16:01:53Z | 2026-04-28T05:00:01Z | 2026-04-25T16:18:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sameerahmed1 | 13,081,729 | MDQ6VXNlcjEzMDgxNzI5 | User | false |
openclaw/openclaw | 4,160,878,584 | I_kwDOQb6kR874Aff4 | 56,522 | https://github.com/openclaw/openclaw/issues/56522 | https://api.github.com/repos/openclaw/openclaw/issues/56522 | Plugin register() called multiple times per gateway startup (v2026.3.24) | ## Bug
The gateway calls `register()` on plugins multiple times during a single startup and continues to re-register periodically during operation. In v2026.3.24 (stable), the `memory-lancedb` plugin`s `register()` was invoked 610 times in a single day, causing a severe memory leak.
### Observed Behavior
- `register... | closed | completed | true | 1 | [] | [] | 2026-03-28T17:06:44Z | 2026-04-28T05:00:00Z | 2026-04-25T16:18:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | josgraha | 32,680 | MDQ6VXNlcjMyNjgw | User | false |
openclaw/openclaw | 4,160,937,297 | I_kwDOQb6kR874At1R | 56,534 | https://github.com/openclaw/openclaw/issues/56534 | https://api.github.com/repos/openclaw/openclaw/issues/56534 | [Control UI] Tool parameter values incorrectly transformed (numbers get decimal points inserted) | ## Bug Description
When entering numeric account/password values in the Control UI chat input, the values are incorrectly transformed before being sent to the model/exec tool.
## Steps to Reproduce
1. Open Control UI (http://127.0.0.1:18789/)
2. Send a message asking to log into a website with credentials
3. Input c... | closed | not_planned | true | 2 | [] | [] | 2026-03-28T17:18:00Z | 2026-04-28T04:59:59Z | 2026-04-25T16:19:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | z1159998925 | 49,932,779 | MDQ6VXNlcjQ5OTMyNzc5 | User | false |
openclaw/openclaw | 4,161,134,258 | I_kwDOQb6kR874Bd6y | 56,552 | https://github.com/openclaw/openclaw/issues/56552 | https://api.github.com/repos/openclaw/openclaw/issues/56552 | [Bug]: @tencent-weixin/openclaw-weixin AbortError on QR code fetch despite successful API connectivity | Bug type: Behavior bug
Summary
@tencent-weixin/openclaw-weixin@2.1.1 fails with AbortError: This operation was aborted when attempting QR code login via openclaw channels login --channel openclaw-weixin. The plugin loads successfully, but the login flow aborts before displaying a QR code.
This is not the same as #52341... | closed | not_planned | true | 4 | [] | [] | 2026-03-28T18:04:54Z | 2026-04-28T04:59:58Z | 2026-04-25T16:19:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KaysonSear | 178,489,097 | U_kgDOCqOHCQ | User | false |
openclaw/openclaw | 4,161,503,452 | I_kwDOQb6kR874C4Dc | 56,596 | https://github.com/openclaw/openclaw/issues/56596 | https://api.github.com/repos/openclaw/openclaw/issues/56596 | [Bug]: After updating, openclaw promise to do task but never really done anthing. | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After updating, openclaw promise to do task but never really done anthing.
### Steps to reproduce
User:
open a browser and go to this website:
https://xxx.xxx
Assistant:
Okay, I’ll go to this link now.
User:
Use Peekaboo... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-28T20:03:33Z | 2026-04-28T04:59:58Z | 2026-04-25T16:19:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | celine-1984-2077 | 268,322,202 | U_kgDOD_5Fmg | User | false |
openclaw/openclaw | 4,161,851,064 | I_kwDOQb6kR874EM64 | 56,642 | https://github.com/openclaw/openclaw/issues/56642 | https://api.github.com/repos/openclaw/openclaw/issues/56642 | [Bug]: WhatsApp reconnect loop causes V8 heap growth → SIGABRT crash | ### Bug type
Regression (worked before, now fails)
### Summary
WhatsApp 503/428 disconnect-reconnect cycles (health-monitor driven) cause steady V8 heap growth in the gateway process, eventually triggering SIGABRT via `v8::internal::BaseNameDictionary::Add`. Crash occurs after ~5 hours of uptime on a 3.8GB RAM machi... | closed | completed | true | 1 | [] | [] | 2026-03-28T22:28:57Z | 2026-04-28T04:59:57Z | 2026-04-25T16:19:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | avbagrin | 262,460,389 | U_kgDOD6TT5Q | User | false |
openclaw/openclaw | 4,162,028,905 | I_kwDOQb6kR874E4Vp | 56,675 | https://github.com/openclaw/openclaw/issues/56675 | https://api.github.com/repos/openclaw/openclaw/issues/56675 | Slack streaming: mutual exclusion between streaming modes causes duplicate delivery, stale previews, and thread misrouting | ## Summary
Slack has three independent streaming mechanisms that can be active simultaneously, causing a class of bugs where content is delivered twice, previews are never cleaned up, or streams land in the wrong thread. Six open issues trace back to this root cause.
## The Problem
The three streaming layers:
| Mod... | closed | completed | true | 1 | [] | [] | 2026-03-29T00:12:16Z | 2026-04-28T04:59:56Z | 2026-04-25T16:19:27Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hsiaoa | 70,124,331 | MDQ6VXNlcjcwMTI0MzMx | User | false |
openclaw/openclaw | 4,162,089,240 | I_kwDOQb6kR874FHEY | 56,684 | https://github.com/openclaw/openclaw/issues/56684 | https://api.github.com/repos/openclaw/openclaw/issues/56684 | Claude-mem integration duplicates session init and emits incomplete observation metadata | ## Summary
When the official `claude-mem` memory plugin is enabled in OpenClaw, the integration emits duplicate session init events and sometimes sends incomplete observation metadata to the worker.
This causes:
- duplicate prompt records for a single user message
- inconsistent project/session scoping across the sa... | closed | not_planned | true | 1 | [] | [] | 2026-03-29T00:49:44Z | 2026-04-28T04:59:55Z | 2026-04-25T16:19:33Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | GigiTiti-Kai | 146,460,333 | U_kgDOCLrOrQ | User | false |
openclaw/openclaw | 4,162,348,191 | I_kwDOQb6kR874GGSf | 56,760 | https://github.com/openclaw/openclaw/issues/56760 | https://api.github.com/repos/openclaw/openclaw/issues/56760 | Plugin hooks (before_agent_start, agent_end) don't fire for per-channel-peer sessions | ## Bug
When `session.dmScope` is set to `per-channel-peer`, plugin hooks (`before_agent_start`, `agent_end`) never fire for the per-channel-peer sessions (e.g. `agent:main:telegram:direct:<user_id>`). They only fire for:
- `agent:main:main` (base session)
- `agent:main:cron:*` (cron-spawned sessions)
This means plugi... | closed | completed | true | 2 | [] | [] | 2026-03-29T03:25:20Z | 2026-04-28T04:59:55Z | 2026-04-25T16:19:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Veda-openclaw | 267,342,076 | U_kgDOD-9Q_A | User | false |
openclaw/openclaw | 4,162,686,725 | I_kwDOQb6kR874HY8F | 56,817 | https://github.com/openclaw/openclaw/issues/56817 | https://api.github.com/repos/openclaw/openclaw/issues/56817 | [Feature]: Conversation Branch Management | ### Summary
Add Git-like branch system to conversations for context isolation and topic switching.
### Problem to solve
When conversations jump between unrelated topics (e.g., working on Project A → asking about Git → back to Project A), context gets lost and there's no way to view or resume previous discussion thre... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-29T05:19:53Z | 2026-04-28T04:59:54Z | 2026-04-25T16:19:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Iamwangli | 48,752,449 | MDQ6VXNlcjQ4NzUyNDQ5 | User | false |
openclaw/openclaw | 4,162,841,417 | I_kwDOQb6kR874H-tJ | 56,829 | https://github.com/openclaw/openclaw/issues/56829 | https://api.github.com/repos/openclaw/openclaw/issues/56829 | Feature request: per-agent skill pinning (agents.list[].skills.always) | ## Feature Request
**Per-agent skill pinning** — allow `agents.list[].skills.always` (array of skill names) so spawned sub-agents automatically load specified skills without relying on description-keyword matching.
### Use Case
Specialized agents like a processor architecture assistant that always needs its domain s... | closed | completed | true | 2 | [] | [] | 2026-03-29T05:58:10Z | 2026-04-28T04:59:53Z | 2026-04-25T16:19:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | WolvenRA | 35,547,932 | MDQ6VXNlcjM1NTQ3OTMy | User | false |
openclaw/openclaw | 4,162,937,858 | I_kwDOQb6kR874IWQC | 56,832 | https://github.com/openclaw/openclaw/issues/56832 | https://api.github.com/repos/openclaw/openclaw/issues/56832 | [Bug]: Intermittent deactivated_workspace after /new (session reset race with previous embedded run) !!! | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After triggering /new, OpenClaw intermittently returns a raw internal error payload to chat ({"detail":{"code":"deactivated_workspace"}}) instead of cleanly resetting the session.
This appears to happen when a previous embed... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-29T06:17:46Z | 2026-04-28T04:59:52Z | 2026-04-25T16:19:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jokedul | 54,840,321 | MDQ6VXNlcjU0ODQwMzIx | User | false |
openclaw/openclaw | 4,163,794,833 | I_kwDOQb6kR874LneR | 56,949 | https://github.com/openclaw/openclaw/issues/56949 | https://api.github.com/repos/openclaw/openclaw/issues/56949 | [Bug]: 飞书插件 footer 配置项（elapsed/status）在新版本中丢失 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## 问题描述
之前版本可以配置飞书消息的 footer 相关参数： ... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-29T10:55:44Z | 2026-04-28T04:59:52Z | 2026-04-25T16:20:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zhanglichangning-jpg | 261,853,115 | U_kgDOD5uPuw | User | false |
openclaw/openclaw | 4,164,382,232 | I_kwDOQb6kR874N24Y | 57,068 | https://github.com/openclaw/openclaw/issues/57068 | https://api.github.com/repos/openclaw/openclaw/issues/57068 | Agent overview: primary model selection does not save (reverts to 'Not set' after Refresh) | ## Bug Report
**Summary:** Changing the primary model on the Agent overview page and pressing Save does not persist the change. After pressing Refresh, the primary model reverts to 'Not set'.
## Steps to Reproduce
1. Open the Gateway Dashboard → Agents
2. Select an agent
3. In the Overview tab, change the Primary Mod... | closed | completed | true | 1 | [] | [] | 2026-03-29T13:47:09Z | 2026-04-28T04:59:51Z | 2026-04-25T16:20:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fwends | 4,113,483 | MDQ6VXNlcjQxMTM0ODM= | User | false |
openclaw/openclaw | 4,164,927,545 | I_kwDOQb6kR874P8A5 | 57,144 | https://github.com/openclaw/openclaw/issues/57144 | https://api.github.com/repos/openclaw/openclaw/issues/57144 | LaunchAgent plist loses custom env vars on npm global update | ## Summary
When updating OpenClaw via `npm install -g openclaw@<version>`, the LaunchAgent plist (`~/Library/LaunchAgents/ai.openclaw.gateway.plist`) is regenerated without preserving user-defined environment variables (e.g. `OPENCLAW_TELEGRAM_BOT_TOKEN`, `TELEGRAM_BOT_TOKEN_FAMILY`).
Config values using `${...}` env... | closed | completed | true | 1 | [] | [] | 2026-03-29T16:13:28Z | 2026-04-28T04:59:50Z | 2026-04-25T16:20:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | artdaal | 35,874,009 | MDQ6VXNlcjM1ODc0MDA5 | User | false |
openclaw/openclaw | 4,165,279,552 | I_kwDOQb6kR874RR9A | 57,212 | https://github.com/openclaw/openclaw/issues/57212 | https://api.github.com/repos/openclaw/openclaw/issues/57212 | Discord voice: audio receive pipeline silent — no STT, no decrypt errors, no speaking events | ## Environment
- **OpenClaw:** 2026.3.28 (f9b1079)
- **Node.js:** v25.8.2
- **OS:** macOS Darwin 24.6.0 x86_64 (also tested in Docker Linux amd64)
- **@discordjs/voice:** 0.19.2 (vendored in discord extension)
- **opusscript:** 0.1.1 (installed at top level — runtime requires it explicitly)
- **@discordjs/opus:** built... | closed | completed | true | 1 | [] | [] | 2026-03-29T17:46:54Z | 2026-04-28T04:59:49Z | 2026-04-25T16:20:46Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | szponeczek | 266,308,501 | U_kgDOD9-LlQ | User | false |
openclaw/openclaw | 4,328,650,362 | I_kwDOQb6kR88AAAABAgH2eg | 71,656 | https://github.com/openclaw/openclaw/issues/71656 | https://api.github.com/repos/openclaw/openclaw/issues/71656 | Memory dreaming: corpus pre-filtering, weighted scoring, and session scan stall | ## Summary
The memory-core dreaming pipeline runs nightly but has three scaling/quality issues that prevent meaningful promotion from light sleep to REM as the corpus grows.
## Environment
- OpenClaw 2026.4.22 (00bd2cf)
- memory-core plugin, dreaming enabled
- 19 successful nightly runs since deployment
## Issues
#... | open | null | false | 1 | [] | [] | 2026-04-25T16:30:12Z | 2026-05-02T15:23:21Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | Lead2Legacy | 268,115,986 | U_kgDOD_sgEg | User | false |
openclaw/openclaw | 4,328,672,388 | I_kwDOQb6kR88AAAABAgJMhA | 71,657 | https://github.com/openclaw/openclaw/issues/71657 | https://api.github.com/repos/openclaw/openclaw/issues/71657 | [Bug] Telegram DM sends silent reply messages like 'No further word from me' after update to 2026.4.23 | ## Bug Description
Telegram direct-message turns that end with no visible final reply emit synthetic user-facing bubbles like 'No further word from me' or 'Nothing new to add from my side' instead of staying silent.
This started after updating to OpenClaw 2026.4.23 (from 2026.4.14).
## Environment
- OpenClaw version:... | closed | not_planned | true | 1 | [] | [] | 2026-04-25T16:38:30Z | 2026-04-29T04:41:20Z | 2026-04-26T08:24:27Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | aecom | 5,202,392 | MDQ6VXNlcjUyMDIzOTI= | User | false |
openclaw/openclaw | 4,328,679,241 | I_kwDOQb6kR88AAAABAgJnSQ | 71,659 | https://github.com/openclaw/openclaw/issues/71659 | https://api.github.com/repos/openclaw/openclaw/issues/71659 | Plugin-SDK `DEFAULT_PROVIDER` hardcoded to "openai" breaks plugin model resolution for users on `openai-codex`-only setups | ## Problem
`DEFAULT_PROVIDER` is exported from the plugin-SDK as a hardcoded string `"openai"`. Plugins import this constant and use it as the fallback provider when parsing user-provided model refs (e.g., `parseModelRef("gpt-5.5", DEFAULT_PROVIDER)`).
For users on a Codex Pro OAuth setup (where the only configured p... | closed | completed | true | 1 | [] | [] | 2026-04-25T16:40:49Z | 2026-04-28T04:50:24Z | 2026-04-25T18:41:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | solomonneas | 41,877,493 | MDQ6VXNlcjQxODc3NDkz | User | false |
openclaw/openclaw | 4,328,679,218 | I_kwDOQb6kR88AAAABAgJnMg | 71,658 | https://github.com/openclaw/openclaw/issues/71658 | https://api.github.com/repos/openclaw/openclaw/issues/71658 | `executionContract: "strict-agentic"` planning-only retry skips on common imperatives and present-continuous narration | ## Problem
`executionContract: "strict-agentic"` advertises that GPT-5-family runs will keep acting through plan-only turns. In practice the retry silently skips when:
- **Gap A:** The user prompt uses a common imperative verb not in the actionable allowlist
- **Gap B:** The assistant's planning narration uses a pres... | closed | not_planned | true | 2 | [] | [] | 2026-04-25T16:40:49Z | 2026-04-29T04:34:27Z | 2026-04-26T18:59:58Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | solomonneas | 41,877,493 | MDQ6VXNlcjQxODc3NDkz | User | false |
openclaw/openclaw | 4,328,689,532 | I_kwDOQb6kR88AAAABAgKPfA | 71,662 | https://github.com/openclaw/openclaw/issues/71662 | https://api.github.com/repos/openclaw/openclaw/issues/71662 | claude-cli provider: session abort triggers gateway-wide SIGTERM (v2026.4.23) | ## Bug Report
**Version:** 2026.4.23 (a979721)
**OS:** macOS 25.4.0 (arm64)
**Provider:** `claude-cli/claude-opus-4-7`
## Summary
When the main agent uses `claude-cli/claude-opus-4-7` as its primary model, a session abort causes the entire gateway process to receive SIGTERM and restart. This was introduced in v2026.... | closed | completed | true | 0 | [] | [] | 2026-04-25T16:45:42Z | 2026-04-28T04:46:02Z | 2026-04-25T21:08:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zjy282 | 8,154,648 | MDQ6VXNlcjgxNTQ2NDg= | User | false |
openclaw/openclaw | 4,328,690,615 | I_kwDOQb6kR88AAAABAgKTtw | 71,663 | https://github.com/openclaw/openclaw/issues/71663 | https://api.github.com/repos/openclaw/openclaw/issues/71663 | Telegram: partial streaming leaks tool call names into chat | ## Description
When `channels.telegram.streaming.mode` is set to `"partial"`, tool call names and content leak into Telegram as live message edits before filtering can strip them. Users see raw tool call identifiers flash in the chat during agent execution.
## Reproduction
1. Set `channels.telegram.streaming.mode: "... | closed | not_planned | true | 1 | [] | [] | 2026-04-25T16:46:17Z | 2026-04-29T04:41:19Z | 2026-04-26T08:24:31Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | BCofSD | 187,073,902 | U_kgDOCyaFbg | User | false |
openclaw/openclaw | 4,328,687,556 | I_kwDOQb6kR88AAAABAgKHxA | 71,661 | https://github.com/openclaw/openclaw/issues/71661 | https://api.github.com/repos/openclaw/openclaw/issues/71661 | Add headless option to browser profile config | ## Problem
OpenClaw's browser tool fails to start on headless Mac minis (no display attached) because Chrome is launched in non-headless mode by default. The error is:
```
Error: Chrome CDP websocket for profile "openclaw" is not reachable after start.
```
Chrome logs show `CVDisplayLinkCreateWithCGDisplay failed. C... | closed | not_planned | true | 1 | [] | [] | 2026-04-25T16:44:46Z | 2026-04-28T04:59:05Z | 2026-04-25T16:46:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sids | 16,904 | MDQ6VXNlcjE2OTA0 | User | false |
openclaw/openclaw | 4,328,698,729 | I_kwDOQb6kR88AAAABAgKzaQ | 71,664 | https://github.com/openclaw/openclaw/issues/71664 | https://api.github.com/repos/openclaw/openclaw/issues/71664 | [Bug]: Block streaming: same-channel text blocks not delivered before tool execution on WhatsApp | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When an agent emits text and then makes tool calls in the same turn, the text is not delivered to WhatsApp until all tool calls complete and the entire turn finishes. This means a simple "Checking now..." ackno... | closed | not_planned | true | 1 | [
"bug"
] | [] | 2026-04-25T16:50:12Z | 2026-04-29T04:41:18Z | 2026-04-26T08:24:34Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | sbmilburn | 11,755,869 | MDQ6VXNlcjExNzU1ODY5 | User | false |
openclaw/openclaw | 4,328,710,915 | I_kwDOQb6kR88AAAABAgLjAw | 71,665 | https://github.com/openclaw/openclaw/issues/71665 | https://api.github.com/repos/openclaw/openclaw/issues/71665 | Feature: Handle native Slack Slash Commands via Socket Mode | ## Feature Request: Handle Native Slack Slash Commands via Socket Mode
### Problem
When Slash Commands are registered in a Slack App's manifest, Slack intercepts messages starting with `/` and routes them as `slash_commands` events through Socket Mode, instead of delivering them as regular text messages.
Currently, ... | closed | completed | true | 1 | [] | [] | 2026-04-25T16:55:05Z | 2026-04-28T04:46:32Z | 2026-04-25T19:35:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | BigBoom20 | 123,953,520 | U_kgDOB2NhcA | User | false |
openclaw/openclaw | 4,328,717,666 | I_kwDOQb6kR88AAAABAgL9Yg | 71,667 | https://github.com/openclaw/openclaw/issues/71667 | https://api.github.com/repos/openclaw/openclaw/issues/71667 | openclaw update: completion cache refresh fails because onboard gets duplicate --openai-api-key registration | ## Bug
`openclaw update` fails at the final completion-cache refresh step with:
```text
Completion cache update failed ([openclaw] Failed to start CLI: Error: Cannot add option '--openai-api-key <key>' to command 'onboard' due to conflicting flag '--openai-api-key'
```
## What I checked
- `openclaw --help` still wo... | closed | completed | true | 2 | [] | [] | 2026-04-25T16:57:45Z | 2026-04-28T04:48:54Z | 2026-04-25T19:12:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Tethys0 | 187,063,395 | U_kgDOCyZcYw | User | false |
openclaw/openclaw | 4,328,727,491 | I_kwDOQb6kR88AAAABAgMjww | 71,669 | https://github.com/openclaw/openclaw/issues/71669 | https://api.github.com/repos/openclaw/openclaw/issues/71669 | [Feature]: Emit runtime warning when Control UI config will silently reject non-secure connections | ### Feature type
Enhancement (developer experience / operator UX)
### Summary
When the gateway starts with `controlUi.enabled: true` and `dangerouslyDisableDeviceAuth` is not set, non-localhost HTTP connections are silently rejected with `DEVICE_IDENTITY_REQUIRED`. The user sees a cryptic "Disconnected from Gateway"... | open | null | false | 1 | [] | [] | 2026-04-25T17:01:35Z | 2026-05-02T14:22:32Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | profbernardoj | 260,372,225 | U_kgDOD4T3AQ | User | false |
openclaw/openclaw | 4,328,732,574 | I_kwDOQb6kR88AAAABAgM3ng | 71,670 | https://github.com/openclaw/openclaw/issues/71670 | https://api.github.com/repos/openclaw/openclaw/issues/71670 | [Bug]: `parseFeishuCardActionEventPayload` field name mismatch for v2 card callbacks | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Clicking any button on a Feishu Schema 2.0 interactive card is silently discarded. The card.action.trigger callback fires but the gateway drops it with ignoring malformed card action payload. Card buttons are c... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-25T17:03:35Z | 2026-04-28T04:45:33Z | 2026-04-25T23:58:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | eddy1068 | 279,298,293 | U_kgDOEKXA9Q | User | false |
openclaw/openclaw | 4,165,414,384 | I_kwDOQb6kR874Ry3w | 57,234 | https://github.com/openclaw/openclaw/issues/57234 | https://api.github.com/repos/openclaw/openclaw/issues/57234 | Browser downloads via CDP go to Playwright temp dir instead of user Downloads folder | ## Problem
When the browser automation triggers downloads via CDP (Chrome DevTools Protocol), Playwright routes them to its temp `artifactsDir` (e.g., `/var/folders/.../T/playwright-artifacts-...`) instead of the user's Downloads folder. These temp directories are auto-cleaned, so downloaded files disappear.
**What w... | closed | completed | true | 1 | [] | [] | 2026-03-29T18:23:41Z | 2026-04-28T04:59:01Z | 2026-04-25T17:04:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | autobot177-sa | 263,484,899 | U_kgDOD7R14w | User | false |
openclaw/openclaw | 4,165,414,922 | I_kwDOQb6kR874RzAK | 57,235 | https://github.com/openclaw/openclaw/issues/57235 | https://api.github.com/repos/openclaw/openclaw/issues/57235 | [Feature]: opencalw chatgpt like interface + new sessions / chat button | ### Summary
<img width="1469" height="987" alt="Image" src="https://github.com/user-attachments/assets/ab41d482-c77e-4bc1-88f9-ed2eb853052c" />
I would like to have abiltity to talk through openclaw over webapp like I talk with other chat-agents (chatgpt.com, gemini.google.com, grok.com). So evry task or research - n... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-29T18:23:53Z | 2026-04-28T04:59:00Z | 2026-04-25T17:04:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dzianisv | 2,119,348 | MDQ6VXNlcjIxMTkzNDg= | User | false |
openclaw/openclaw | 4,165,675,229 | I_kwDOQb6kR874Syjd | 57,274 | https://github.com/openclaw/openclaw/issues/57274 | https://api.github.com/repos/openclaw/openclaw/issues/57274 | [Bug:] Dashboard and CLI fail to render responses with Ollama integration (Qwen2.5) | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
Yes
### Summary
Troubleshooting Report: OpenClaw + Ollama (Qwen2.5)
Context
Environment: Windows 10/11, PowerShell
Gateway Command: openclaw gateway --port 18789
Target Model: ollama/qwen2.5:latest
Goal: Use OpenClaw’s dashboard to interac... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-03-29T20:04:27Z | 2026-04-28T04:58:59Z | 2026-04-25T17:04:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ClovisChProgrammer | 225,098,580 | U_kgDODWq7VA | User | false |
openclaw/openclaw | 4,165,767,059 | I_kwDOQb6kR874TI-T | 57,292 | https://github.com/openclaw/openclaw/issues/57292 | https://api.github.com/repos/openclaw/openclaw/issues/57292 | context1m beta header stripped for OAuth token auth — breaks Max plan users | ## Bug Description
OpenClaw strips the `anthropic-beta: max-model-output-200k-2025-02-19` (context1m) header when the Anthropic auth profile uses OAuth token mode (`mode: token`). The log emits:
```
ignoring context1m for OAuth token auth on anthropic/claude-opus-4-6; Anthropic rejects context-1m beta with OAuth auth... | closed | not_planned | true | 1 | [] | [] | 2026-03-29T20:48:44Z | 2026-04-28T04:58:58Z | 2026-04-25T17:04:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dagny-holland | 261,707,352 | U_kgDOD5lWWA | User | false |
openclaw/openclaw | 4,165,997,374 | I_kwDOQb6kR874UBM- | 57,323 | https://github.com/openclaw/openclaw/issues/57323 | https://api.github.com/repos/openclaw/openclaw/issues/57323 | Telegram channel silently fails to connect on fresh Fly.io deploy (v2026.3.23) | ## Description
On a fresh Fly.io deploy, the Telegram channel is marked as "configured, enabled automatically" in gateway startup logs, but never actually connects or polls for updates. No error is logged. The gateway starts normally, port 3000 binds, other plugins (diagnostics-otel) load successfully, but Telegram is... | closed | completed | true | 2 | [] | [
"vincentkoc"
] | 2026-03-29T22:53:13Z | 2026-04-28T04:58:58Z | 2026-04-25T17:04:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Inkvi | 374,203 | MDQ6VXNlcjM3NDIwMw== | User | false |
openclaw/openclaw | 4,166,408,554 | I_kwDOQb6kR874Vllq | 57,390 | https://github.com/openclaw/openclaw/issues/57390 | https://api.github.com/repos/openclaw/openclaw/issues/57390 | Global npm update drops node-llama-cpp (optional peerDep not re-installed) | ## Summary
`npm i -g openclaw@latest` drops `node-llama-cpp` on every update, breaking local memory-search embeddings until manually re-installed.
## Environment
- macOS 15.3.1 (arm64, Apple M2)
- Node.js 22.22.0
- npm 10.x (global install)
- OpenClaw versions affected: at least 2026.3.24 → 2026.3.28 (and 2026.3.x p... | closed | completed | true | 2 | [] | [] | 2026-03-30T01:36:58Z | 2026-04-28T04:58:57Z | 2026-04-25T17:04:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1sexywoo8 | 262,731,466 | U_kgDOD6j2yg | User | false |
openclaw/openclaw | 4,166,721,882 | I_kwDOQb6kR874WyFa | 57,430 | https://github.com/openclaw/openclaw/issues/57430 | https://api.github.com/repos/openclaw/openclaw/issues/57430 | [Bug]: Reasoning is required for this model endpoint. Use /think minimal (or any non-off level) and try again. | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
after upgarding to
v2026.3.28 for every call i'm seeing the error **Reasoning is required for this model endpoint. Use /think minimal (or any non-off level) and try again.**
It was working as is without issues before upgra... | closed | completed | true | 8 | [
"bug",
"regression"
] | [] | 2026-03-30T02:58:44Z | 2026-04-28T04:58:56Z | 2026-04-25T17:05:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dnkmoc | 269,955,571 | U_kgDOEBcx8w | User | false |
openclaw/openclaw | 4,167,785,303 | I_kwDOQb6kR874a1tX | 57,515 | https://github.com/openclaw/openclaw/issues/57515 | https://api.github.com/repos/openclaw/openclaw/issues/57515 | [Bug]: openclaw backup create --verify fails with duplicate entry path when workspace contains symlinks | ## Bug type
Behavior bug
## Summary
`openclaw backup create --verify` fails with `Error: Archive contains duplicate entry path` when the workspace directory contains symbolic links pointing to paths already included in the backup.
## Steps to reproduce
1. Create a symlink inside the workspace: `ln -s /some/path ~/.op... | closed | completed | true | 1 | [] | [] | 2026-03-30T06:17:00Z | 2026-04-28T04:58:55Z | 2026-04-25T17:05:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | MilongWong | 13,269,445 | MDQ6VXNlcjEzMjY5NDQ1 | User | false |
openclaw/openclaw | 4,167,868,944 | I_kwDOQb6kR874bKIQ | 57,523 | https://github.com/openclaw/openclaw/issues/57523 | https://api.github.com/repos/openclaw/openclaw/issues/57523 | [Bug]: Kimi k2p5 (anthropic-messages): "Unexpected event order: message_start before message_stop" — chat unusable | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Summary
When using **Kimi** with model **`k2p5`** via the **Anthropic Messages–compatible** streaming path (`anthropic-messages`), OpenClaw fails on assistant runs with:
`Unexpected event order, got message_start before re... | closed | not_planned | true | 55 | [
"bug",
"regression"
] | [] | 2026-03-30T06:31:08Z | 2026-04-28T04:58:54Z | 2026-04-25T17:05:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kanmanhu-less | 194,345,236 | U_kgDOC5V5FA | User | false |
openclaw/openclaw | 4,168,725,502 | I_kwDOQb6kR874ebP- | 57,589 | https://github.com/openclaw/openclaw/issues/57589 | https://api.github.com/repos/openclaw/openclaw/issues/57589 | [Bug]: Openclaw command shows incorrect current version | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
After upgrading from 2026.3.24 to 2026.3.28, openclaw shows incorrect current version except 'openclaw --version'.
### Steps to reproduce
<img width="873" height="111" alt="Image" src="https://github.com/user... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T08:34:42Z | 2026-04-28T04:58:50Z | 2026-04-25T17:05:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dong-jh | 46,438,624 | MDQ6VXNlcjQ2NDM4NjI0 | User | false |
openclaw/openclaw | 4,169,427,188 | I_kwDOQb6kR874hGj0 | 57,636 | https://github.com/openclaw/openclaw/issues/57636 | https://api.github.com/repos/openclaw/openclaw/issues/57636 | [Feature]: Plugin hooks cannot distinguish main agent runs from subagent runs | ## Summary
Add subagent identification to `PluginHookAgentContext` so that plugin hooks can distinguish between main agent runs and subagent runs.
## Problem to solve
When registering a `before_prompt_build` (or other agent hooks) hook, there is no way to determine if the current run is the **main agent** or a **sub... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T10:10:39Z | 2026-04-28T04:58:50Z | 2026-04-25T17:05:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Rheasilvia | 29,389,840 | MDQ6VXNlcjI5Mzg5ODQw | User | false |
openclaw/openclaw | 4,165,102,234 | I_kwDOQb6kR874Qmqa | 57,175 | https://github.com/openclaw/openclaw/issues/57175 | https://api.github.com/repos/openclaw/openclaw/issues/57175 | Feature: Session isolation audit — detect cross-session context leakage in multi-agent setups | ## Summary
In multi-agent deployments, agents sometimes reference context from other sessions or make assumptions about conversations they weren't part of. There's no built-in tool to detect or prevent this "session leakage." We built a detection script and propose native support.
## The Problem
### Symptom: Ghost C... | closed | not_planned | true | 4 | [] | [] | 2026-03-29T16:55:42Z | 2026-04-28T04:58:49Z | 2026-04-25T16:20:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1052326311 | 65,798,732 | MDQ6VXNlcjY1Nzk4NzMy | User | false |
openclaw/openclaw | 4,170,876,459 | I_kwDOQb6kR874moYr | 57,725 | https://github.com/openclaw/openclaw/issues/57725 | https://api.github.com/repos/openclaw/openclaw/issues/57725 | [bug] Embedded ws-stream websocket upgrade fails with HTTP 500 (falls back to HTTP) | ## Summary
Control UI repeatedly fails to establish embedded ws-stream connections with `Unexpected server response: 500`, then falls back to HTTP transport.
## Version / Environment
- OpenClaw: `2026.3.28`
- OS: Windows 10.0.22621 x64
- Node: 22.22.0
- Gateway: local loopback `ws://127.0.0.1:18789`
- Client: openclaw... | closed | completed | true | 2 | [] | [] | 2026-03-30T13:46:16Z | 2026-04-28T04:58:47Z | 2026-04-25T17:06:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | RacecarGuy | 13,582,683 | MDQ6VXNlcjEzNTgyNjgz | User | false |
openclaw/openclaw | 4,170,879,837 | I_kwDOQb6kR874mpNd | 57,726 | https://github.com/openclaw/openclaw/issues/57726 | https://api.github.com/repos/openclaw/openclaw/issues/57726 | Add built-in auth health monitoring with proactive expiry/failure notifications | ## Problem
OpenClaw already exposes auth health via:
```bash
openclaw models status --check
```
with exit codes for:
- `0`: healthy
- `1`: broken or missing auth
- `2`: expiring soon
That is enough for users to build their own monitoring, but not enough as a product default. In practice, auth issues are often disc... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T13:46:48Z | 2026-04-28T04:58:46Z | 2026-04-25T17:06:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Phaven910 | 192,037,295 | U_kgDOC3JBrw | User | false |
openclaw/openclaw | 4,165,029,436 | I_kwDOQb6kR874QU48 | 57,162 | https://github.com/openclaw/openclaw/issues/57162 | https://api.github.com/repos/openclaw/openclaw/issues/57162 | Feature Proposal: External Health Monitor & Auto-Recovery Framework for Multi-Agent Deployments | ## Summary
OpenClaw lacks a native health monitoring and auto-recovery system for multi-agent deployments. We built a production Watchdog system (cron-based, with Feishu alerts, session health checks, token overflow warnings, and auto-recovery) that has been running for 2+ weeks across 17 agents. Proposing to upstream... | closed | not_planned | true | 5 | [] | [] | 2026-03-29T16:40:22Z | 2026-04-28T04:58:45Z | 2026-04-25T16:20:33Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1052326311 | 65,798,732 | MDQ6VXNlcjY1Nzk4NzMy | User | false |
openclaw/openclaw | 4,171,881,099 | I_kwDOQb6kR874qdqL | 57,801 | https://github.com/openclaw/openclaw/issues/57801 | https://api.github.com/repos/openclaw/openclaw/issues/57801 | Feishu WebSocket reconnect every 5 minutes causes webchat console disconnect (code=1006) | ## Environment
- OpenClaw version: 2026.3.28 (f9b1079)
- OS: macOS (Mac mini)
- Deployment: local mode
- Feishu channel: enabled, connectionMode: websocket
## Bug Description
Feishu WebSocket triggers an abort signal and reconnects every 5 minutes. This simultaneously causes the webchat console WebSocket to disconnec... | closed | completed | true | 1 | [] | [] | 2026-03-30T15:51:56Z | 2026-04-28T04:58:44Z | 2026-04-25T17:06:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fx18646118 | 36,415,654 | MDQ6VXNlcjM2NDE1NjU0 | User | false |
openclaw/openclaw | 4,172,303,548 | I_kwDOQb6kR874sEy8 | 57,834 | https://github.com/openclaw/openclaw/issues/57834 | https://api.github.com/repos/openclaw/openclaw/issues/57834 | [Bug]: Session force-resets immediately after user questions in kimi-claw channel | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Sessions force-reset immediately after users ask questions in the kimi-claw channel, losing all recent context and requiring restart of conversation.
### Steps to reproduce
1. User starts a session in kimi-cl... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T16:43:28Z | 2026-04-28T04:58:43Z | 2026-04-25T17:06:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hydrogenbondss | 266,451,670 | U_kgDOD-G61g | User | false |
openclaw/openclaw | 4,172,407,661 | I_kwDOQb6kR874seNt | 57,851 | https://github.com/openclaw/openclaw/issues/57851 | https://api.github.com/repos/openclaw/openclaw/issues/57851 | Gateway does not resume orphaned sessions after crash/restart | ## Bug Report
### Summary
When the gateway process crashes (or its WebSocket connection to the LLM provider drops), sessions that were mid-turn are orphaned. On restart, the gateway does not detect or resume these sessions — they remain in `running` state indefinitely, blocking the agent's `maxConcurrent` slot so no n... | closed | completed | true | 1 | [] | [] | 2026-03-30T16:56:56Z | 2026-04-28T04:58:42Z | 2026-04-25T17:06:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ctcreel | 1,657,759 | MDQ6VXNlcjE2NTc3NTk= | User | false |
openclaw/openclaw | 4,173,366,878 | I_kwDOQb6kR874wIZe | 57,890 | https://github.com/openclaw/openclaw/issues/57890 | https://api.github.com/repos/openclaw/openclaw/issues/57890 | feat(cron): add structured job-completion record to detect partial/incomplete cron cycles | ## Problem
When a cron job modifies state files (e.g. `heartbeat-state.json`) but is killed before writing its completion log, there is no way for subsequent cycles to detect that the previous run was incomplete.
**Observed behavior**: A self-evolution cron cycle updated `heartbeat-state.json` (setting `lastEvolveRun... | closed | completed | true | 1 | [] | [] | 2026-03-30T19:09:54Z | 2026-04-28T04:58:41Z | 2026-04-25T17:07:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,173,424,268 | I_kwDOQb6kR874wWaM | 57,894 | https://github.com/openclaw/openclaw/issues/57894 | https://api.github.com/repos/openclaw/openclaw/issues/57894 | [Feature] Time-based session context window (send only recent N hours of conversation) | ## Problem
In long-running (endless) sessions, every request sends the full conversation history. This wastes tokens on stale context and can hit context limits, forcing users to manually run `/new` to start a fresh session just to shed old context.
## Proposed Solution
Implement a configurable time-based session co... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T19:18:07Z | 2026-04-28T04:58:41Z | 2026-04-25T17:07:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | apoapostolov | 4,083,812 | MDQ6VXNlcjQwODM4MTI= | User | false |
openclaw/openclaw | 4,173,955,057 | I_kwDOQb6kR874yX_x | 57,935 | https://github.com/openclaw/openclaw/issues/57935 | https://api.github.com/repos/openclaw/openclaw/issues/57935 | Telegram incoming messages not received after gateway restart | **Bug Description:**
- Bot can send messages (outgoing works)
- Bot cannot receive messages (incoming broken)
- Telegram API getUpdates call returns empty array
- Issue started after gateway restart at 3:58 PM ET on March 30, 2026
- Multiple gateway restarts do not fix the issue
**Environment:**
- OpenClaw version: ... | closed | completed | true | 2 | [] | [] | 2026-03-30T20:42:46Z | 2026-04-28T04:58:39Z | 2026-04-25T17:07:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | SephyAI-sys | 269,233,011 | U_kgDOEAwrcw | User | false |
openclaw/openclaw | 4,174,434,428 | I_kwDOQb6kR8740NB8 | 57,964 | https://github.com/openclaw/openclaw/issues/57964 | https://api.github.com/repos/openclaw/openclaw/issues/57964 | [Bug]: repeated characters collapse when displayed in dashboard or tui | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
often when same symbol repeats in web dashboard or tui, its collapsed into a single symbol.
initially i thought it was a perplexity issue, but i've tried with 16bit for both model and kv-cache.
also seen it happen with mult... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-30T22:24:49Z | 2026-04-28T04:58:38Z | 2026-04-25T17:07:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | SGC41 | 104,068,405 | U_kgDOBjP1NQ | User | false |
openclaw/openclaw | 4,175,103,075 | I_kwDOQb6kR8742wRj | 58,049 | https://github.com/openclaw/openclaw/issues/58049 | https://api.github.com/repos/openclaw/openclaw/issues/58049 | Failover loop: timeout-heavy candidates starve later fallbacks in run time budget | ## Summary
When a session's `modelOverride` points to an unavailable model (e.g. a Vertex AI preview model that returns 404), the failover mechanism enters an infinite retry loop instead of converging on a working fallback candidate.
## Root Cause
The fallback chain is evaluated **per embedded run**, and each run ha... | closed | completed | true | 1 | [] | [] | 2026-03-31T01:47:46Z | 2026-04-28T04:58:38Z | 2026-04-25T17:08:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | justin7974 | 46,263,464 | MDQ6VXNlcjQ2MjYzNDY0 | User | false |
openclaw/openclaw | 4,176,627,591 | I_kwDOQb6kR8748keH | 58,195 | https://github.com/openclaw/openclaw/issues/58195 | https://api.github.com/repos/openclaw/openclaw/issues/58195 | [Bug]: Reasoning issue | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Model reasoning content is not displayed in the OpenClaw Control UI (webchat) on Linux x64, even though streaming works and the model generates thinking correctly. Same version 2026.3.28 works on RPi ARM.
Environment
OpenC... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-31T07:19:39Z | 2026-04-28T04:58:37Z | 2026-04-25T17:08:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | NewIdeaSrls | 3,061,685 | MDQ6VXNlcjMwNjE2ODU= | User | false |
openclaw/openclaw | 4,177,334,556 | I_kwDOQb6kR874_REc | 58,255 | https://github.com/openclaw/openclaw/issues/58255 | https://api.github.com/repos/openclaw/openclaw/issues/58255 | Gemini memory indexing fails with fetch failed in WSL2/OpenClaw 2026.3.28, while direct Node fetch to Gemini embeddings succeeds | ## Summary
OpenClaw memory indexing with `memorySearch.provider = "gemini"` fails with:
- `Memory index failed: fetch failed`
- `Embeddings error: fetch failed`
- `Index error: fetch failed`
However, direct `curl` and direct Node `fetch()` requests to the Gemini embeddings API succeed on the same machine.
This sugg... | closed | completed | true | 1 | [] | [] | 2026-03-31T08:55:54Z | 2026-04-28T04:58:36Z | 2026-04-25T17:08:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | knightHuo | 26,109,947 | MDQ6VXNlcjI2MTA5OTQ3 | User | false |
openclaw/openclaw | 4,177,359,254 | I_kwDOQb6kR874_XGW | 58,258 | https://github.com/openclaw/openclaw/issues/58258 | https://api.github.com/repos/openclaw/openclaw/issues/58258 | Live model switch creates duplicate messages during cross-provider failover loop | ## Bug: Live model switch creates duplicate messages during failover loop
### Environment
- OpenClaw 2026.3.28 (f9b1079)
- macOS Darwin arm64
- Telegram channel
### Config
```json
{
"model": {
"primary": "anthropic/claude-opus-4-6",
"fallbacks": ["anthropic/claude-sonnet-4-6", "google/gemini-3-flash-preview... | closed | completed | true | 1 | [] | [] | 2026-03-31T08:59:10Z | 2026-04-28T04:58:35Z | 2026-04-25T17:08:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | JarvisShpigel | 258,616,481 | U_kgDOD2osoQ | User | false |
openclaw/openclaw | 4,328,744,083 | I_kwDOQb6kR88AAAABAgNkkw | 71,673 | https://github.com/openclaw/openclaw/issues/71673 | https://api.github.com/repos/openclaw/openclaw/issues/71673 | [Bug]: Synthetic `card-action-c-*` message IDs cause delivery layer 400 errors | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When the agent replies to a Feishu card action callback, OpenClaw uses the synthetic message ID (card-action-{token}) as the open_message_id for Feishu API calls. Feishu rejects these with HTTP 400, causing 10+... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-25T17:08:37Z | 2026-04-28T04:45:32Z | 2026-04-25T23:59:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | eddy1068 | 279,298,293 | U_kgDOEKXA9Q | User | false |
openclaw/openclaw | 4,215,761,025 | I_kwDOQb6kR877R2iB | 62,325 | https://github.com/openclaw/openclaw/issues/62325 | https://api.github.com/repos/openclaw/openclaw/issues/62325 | video_generate: expose xAI reference_images mode separately from image-to-video | ## Summary
xAI's video API supports a `reference_images` generation mode that preserves character/style cues without forcing the input image to become the first frame.
OpenClaw currently appears to expose only these shared video modes:
- `generate`
- `imageToVideo`
- `videoToVideo`
As a result, when using `video_gen... | closed | completed | true | 1 | [] | [] | 2026-04-07T05:58:24Z | 2026-04-28T04:57:54Z | 2026-04-25T17:15:15Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zqchris | 4,436,110 | MDQ6VXNlcjQ0MzYxMTA= | User | false |
openclaw/openclaw | 4,328,768,437 | I_kwDOQb6kR88AAAABAgPDtQ | 71,682 | https://github.com/openclaw/openclaw/issues/71682 | https://api.github.com/repos/openclaw/openclaw/issues/71682 | long model timeouts . | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
LLM generated summary : When performing complex tasks, the system's model selection is unreliable (e.g., asking for a specific model fails) and performance degrades severely as context grows, even if context cl... | closed | not_planned | true | 3 | [
"bug"
] | [] | 2026-04-25T17:21:14Z | 2026-04-30T04:33:44Z | 2026-04-27T09:52:59Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | peter247 | 3,247,383 | MDQ6VXNlcjMyNDczODM= | User | false |
openclaw/openclaw | 4,328,753,575 | I_kwDOQb6kR88AAAABAgOJpw | 71,675 | https://github.com/openclaw/openclaw/issues/71675 | https://api.github.com/repos/openclaw/openclaw/issues/71675 | Feature: Configurable timeout for image generation models | ## Problem
The image generation provider hardcodes `DEFAULT_TIMEOUT_MS = 90000` (90s) in `image-generation-provider-*.js`. Models like OpenAI GPT-Image-2 via OpenRouter have average E2E latency of ~127s, causing consistent timeouts and silent fallbacks to lower-quality models (e.g. minimax/image-01).
## Current Worka... | closed | completed | true | 1 | [] | [] | 2026-04-25T17:13:31Z | 2026-04-28T04:56:52Z | 2026-04-25T17:26:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | RoseKongPS | 270,016,663 | U_kgDOEBgglw | User | false |
openclaw/openclaw | 4,081,613,424 | I_kwDOQb6kR87zSHpw | 48,136 | https://github.com/openclaw/openclaw/issues/48136 | https://api.github.com/repos/openclaw/openclaw/issues/48136 | fix(acpx): ACP bridge passes OpenClaw agent ID instead of runtime.acp.agent to ACPX harness | ## Bug
When `sessions_spawn` with `runtime: "acp"` is called, the ACP bridge passes the **OpenClaw agent ID** (e.g. `opus`) directly to the ACPX harness instead of resolving `agents.list[].runtime.acp.agent` (e.g. `claude`).
This means any OpenClaw agent whose ID does not exactly match a built-in ACPX harness name (`... | closed | not_planned | true | 4 | [] | [] | 2026-03-16T10:29:44Z | 2026-04-28T04:56:51Z | 2026-04-25T17:26:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | CryptoKrad | 88,065,315 | MDQ6VXNlcjg4MDY1MzE1 | User | false |
openclaw/openclaw | 4,328,766,909 | I_kwDOQb6kR88AAAABAgO9vQ | 71,680 | https://github.com/openclaw/openclaw/issues/71680 | https://api.github.com/repos/openclaw/openclaw/issues/71680 | [Feature]: memory_get JSON parse errors silently kill tool requests with no error visibility | ### Summary
When MEMORY.md contains special characters (em-dashes, curly quotes, etc.), memory_get fails with JSON parse error on the backend silently. OpenClaw never see an error.
### Problem to solve
Tool: memory_get,
Problem: When MEMORY.md contains special characters (em-dashes, curly quotes, etc.), memory_get f... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-04-25T17:20:27Z | 2026-04-28T04:46:31Z | 2026-04-25T19:35:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rickjan130 | 18,168,141 | MDQ6VXNlcjE4MTY4MTQx | User | false |
openclaw/openclaw | 4,328,757,531 | I_kwDOQb6kR88AAAABAgOZGw | 71,677 | https://github.com/openclaw/openclaw/issues/71677 | https://api.github.com/repos/openclaw/openclaw/issues/71677 | [Feature]: | ### Summary
Support instant card update via Feishu WebSocket callback response — allows buttons to disable within milliseconds of a click instead of waiting 3–5 seconds for the agent to finish processing.
### Problem to solve
Feishu's card.action.trigger WebSocket callback supports instant card updates by returning ... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-04-25T17:15:34Z | 2026-05-02T16:01:55Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | eddy1068 | 279,298,293 | U_kgDOEKXA9Q | User | false |
openclaw/openclaw | 4,178,564,263 | I_kwDOQb6kR875D9Sn | 58,355 | https://github.com/openclaw/openclaw/issues/58355 | https://api.github.com/repos/openclaw/openclaw/issues/58355 | [Bug]: Failed to set model: GatewayRequestError: model not allowed: ollama/qwen3.5:35b-a3b-coding-nvfp4 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Failed to set model: GatewayRequestError: model not allowed: ollama/qwen3.5:35b-a3b-coding-nvfp4
### Steps to reproduce
Failed to set model: GatewayRequestError: model not allowed: ollama/qwen3.5:35b-a3b-coding-nvfp4
##... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-03-31T11:59:15Z | 2026-04-28T04:55:06Z | 2026-04-25T17:46:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hufeng825 | 1,576,614 | MDQ6VXNlcjE1NzY2MTQ= | User | false |
openclaw/openclaw | 4,178,862,739 | I_kwDOQb6kR875FGKT | 58,377 | https://github.com/openclaw/openclaw/issues/58377 | https://api.github.com/repos/openclaw/openclaw/issues/58377 | cron: LiveSessionModelSwitchError treated as transient failure — triggers retry loop and Sonnet budget leak | ## Summary
When a cron job's isolated agentTurn run hits a `LiveSessionModelSwitchError`, the cron engine treats it as a transient failure and immediately re-queues the job. This causes a tight retry loop until the gateway restarts — burning fallback budget on every iteration.
## Behavior
1. Cron job fires (Haiku mo... | closed | completed | true | 2 | [] | [] | 2026-03-31T12:42:38Z | 2026-04-28T04:55:05Z | 2026-04-25T17:46:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rayrage213 | 261,946,476 | U_kgDOD5z8bA | User | false |
openclaw/openclaw | 4,179,387,154 | I_kwDOQb6kR875HGMS | 58,396 | https://github.com/openclaw/openclaw/issues/58396 | https://api.github.com/repos/openclaw/openclaw/issues/58396 | Plugin API: Add injectSystemNote() / injectContext() for before_dispatch hooks | ## Feature Request
### Problem
The `before_dispatch` hook (added in 2026.3.28) allows plugins to intercept incoming messages and load external context — but there is currently no way to **inject that context into the agent's session** synchronously.
When a plugin loads data in `before_dispatch`, it has no method to ... | closed | completed | true | 2 | [] | [] | 2026-03-31T13:49:24Z | 2026-04-28T04:55:05Z | 2026-04-25T17:46:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | agentmorpheus77 | 258,008,831 | U_kgDOD2Dm_w | User | false |
openclaw/openclaw | 4,179,530,097 | I_kwDOQb6kR875HpFx | 58,402 | https://github.com/openclaw/openclaw/issues/58402 | https://api.github.com/repos/openclaw/openclaw/issues/58402 | [Bug]: message tool "Operation aborted" false negative + WhatsApp reconnect storm + xai-auth log spam | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
The message tool returns "Operation aborted" to the model even though the gateway confirms successful Telegram delivery (sendMessage ok). 9 confirmed cases in one session. Additionally: WhatsApp reconnects endlessly with no ... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-31T14:05:39Z | 2026-04-28T04:55:04Z | 2026-04-25T17:46:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | odedkd | 91,198,021 | MDQ6VXNlcjkxMTk4MDIx | User | false |
openclaw/openclaw | 4,180,752,864 | I_kwDOQb6kR875MTng | 58,470 | https://github.com/openclaw/openclaw/issues/58470 | https://api.github.com/repos/openclaw/openclaw/issues/58470 | Gateway startup doesn't clear stale 'running' status in sessions.json | ## Bug
On startup, the gateway correctly removes stale `.lock` files (dead-pid check), but it does **not** reset stale `"status": "running"` entries in the session store (`sessions.json`).
### Impact
Sessions left in `running` state after a crash/restart silently block new inbound messages. For forum channels, a par... | closed | completed | true | 1 | [] | [] | 2026-03-31T16:51:17Z | 2026-04-28T04:55:03Z | 2026-04-25T17:46:43Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | biz-biz-Bot | 269,882,212 | U_kgDOEBYTZA | User | false |
openclaw/openclaw | 4,183,763,324 | I_kwDOQb6kR875Xyl8 | 58,704 | https://github.com/openclaw/openclaw/issues/58704 | https://api.github.com/repos/openclaw/openclaw/issues/58704 | WhatsApp bridge: assistant output can leak into user messages during /new or /reset session transition | ## Summary
In a same-phone WhatsApp direct chat, OpenClaw can misattribute assistant-originated or queued content as a new inbound user message across a `/new` or `/reset` boundary.
This is a message-integrity issue: text not authored by the user can be represented to the agent as authenticated user input.
## Securi... | closed | completed | true | 1 | [] | [] | 2026-04-01T03:19:55Z | 2026-04-28T04:55:02Z | 2026-04-25T17:47:05Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | apollostreetcompany | 171,311,035 | U_kgDOCjX_uw | User | false |
openclaw/openclaw | 4,328,827,254 | I_kwDOQb6kR88AAAABAgSpdg | 71,689 | https://github.com/openclaw/openclaw/issues/71689 | https://api.github.com/repos/openclaw/openclaw/issues/71689 | Bug: tasks registry restore fails on malformed SQLite image | **Reporting Agent**: GitHub Copilot
**Session ID**: platform-infra-2026-04-25
**Component**: tasks/registry
**Severity**: High
### Description
Gateway startup repeatedly failed to restore durable task registry due to SQLite corruption in `tasks/runs.sqlite` (`database disk image is malformed`). This undermines durable... | open | null | false | 2 | [] | [] | 2026-04-25T17:47:31Z | 2026-05-02T21:47:40Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | waynegault | 39,455,578 | MDQ6VXNlcjM5NDU1NTc4 | User | false |
openclaw/openclaw | 4,328,833,605 | I_kwDOQb6kR88AAAABAgTCRQ | 71,690 | https://github.com/openclaw/openclaw/issues/71690 | https://api.github.com/repos/openclaw/openclaw/issues/71690 | Gateway cold-start: bundled-plugin loading dominates startup; plugins.deny not honored at import-time | ## Summary
OpenClaw gateway cold-start time is dominated by loading bundled plugins that the operator doesn't actually use. On a fast workstation with warm disk cache, `openclaw gateway` reaches TCP-listening in ~4.3s; cold cache pushes that to ~10.8s. In containerized environments with slower per-vCPU and remote-moun... | closed | completed | true | 2 | [] | [] | 2026-04-25T17:50:32Z | 2026-04-29T04:38:08Z | 2026-04-26T09:53:18Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jarvis-ai-gregmoser | 258,473,251 | U_kgDOD2f9Iw | User | false |
openclaw/openclaw | 4,328,836,073 | I_kwDOQb6kR88AAAABAgTL6Q | 71,691 | https://github.com/openclaw/openclaw/issues/71691 | https://api.github.com/repos/openclaw/openclaw/issues/71691 | Session fork caused by heartbeat compaction loop losing conversation context | ## Description
When a session mixes direct user chat (Telegram) with background subagent work (heartbeats, sherlock runs), the compaction system enters a loop that:
1. Repeatedly compacts every few seconds
2. Generates "No goal established — the conversation is empty" summaries despite active conversation history
3. G... | closed | completed | true | 2 | [] | [] | 2026-04-25T17:51:49Z | 2026-04-29T04:38:07Z | 2026-04-26T09:53:22Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | Shymitzk | 89,461,234 | MDQ6VXNlcjg5NDYxMjM0 | User | false |
openclaw/openclaw | 4,201,838,476 | I_kwDOQb6kR876cveM | 60,490 | https://github.com/openclaw/openclaw/issues/60490 | https://api.github.com/repos/openclaw/openclaw/issues/60490 | [Bug]: Gateway not starting on Windows automatically (need to start manually) – Dashboard unreachable (127.0.0.1 refused to connect) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Gateway does not start on Windows, causing the dashboard at 127.0.0.1:18789 to refuse connection
### Steps to reproduce
Install OpenClaw on Windows
Uninstall
Reinstall OpenClaw on windows
Open PowerShell as ... | closed | completed | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-04-03T19:05:26Z | 2026-05-01T04:42:33Z | 2026-04-28T17:42:30Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | alexandre-leng | 72,550,137 | MDQ6VXNlcjcyNTUwMTM3 | User | false |
openclaw/openclaw | 4,328,826,678 | I_kwDOQb6kR88AAAABAgSnNg | 71,688 | https://github.com/openclaw/openclaw/issues/71688 | https://api.github.com/repos/openclaw/openclaw/issues/71688 | Bug: hal-startup hook timers can block lifecycle closure | **Reporting Agent**: GitHub Copilot
**Session ID**: platform-infra-2026-04-25
**Component**: gateway/hooks
**Severity**: High
### Description
The internal startup hook `hal-startup` scheduled delayed and periodic timers without calling `unref()`. In lifecycle paths where completion depends on event-loop quiescence, th... | closed | not_planned | true | 1 | [] | [] | 2026-04-25T17:47:18Z | 2026-04-28T04:54:37Z | 2026-04-25T17:52:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | waynegault | 39,455,578 | MDQ6VXNlcjM5NDU1NTc4 | User | false |
openclaw/openclaw | 4,328,842,567 | I_kwDOQb6kR88AAAABAgTlRw | 71,693 | https://github.com/openclaw/openclaw/issues/71693 | https://api.github.com/repos/openclaw/openclaw/issues/71693 | BUG Powershell or Command prompt popping up and disappearing every 30 seconds | ### Summary
Powershell or Command prompt popping up and disappearing every 30 seconds
### Problem to solve
Powershell or Command prompt popping up and disappearing every 30 seconds
### Proposed solution
No popups
### Alternatives considered
_No response_
### Impact
Every 30 seconds
### Evidence/examples
_No ... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-04-25T17:55:05Z | 2026-04-29T04:35:44Z | 2026-04-26T14:56:49Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | athuljayaram | 141,337,513 | U_kgDOCGyjqQ | User | false |
openclaw/openclaw | 4,184,623,941 | I_kwDOQb6kR875bEtF | 58,807 | https://github.com/openclaw/openclaw/issues/58807 | https://api.github.com/repos/openclaw/openclaw/issues/58807 | [Bug]: falling back to HTTP. error=Error: connect ENETUNREACH | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Openclaw websocket connect to ws api.openapi.com doesn't respect HTTP_PROXY, HTTPS_PROXY, look like it direct connect to the server
openclaw-gateway-1 | 2026-03-31T23:49:35.300-07:00 [agent/embedded] [ws-stream] WebSocket ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-01T06:57:59Z | 2026-04-28T04:54:03Z | 2026-04-25T17:58:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | minhnv02acb | 175,083,321 | U_kgDOCm-POQ | User | false |
openclaw/openclaw | 4,185,836,198 | I_kwDOQb6kR875fsqm | 58,941 | https://github.com/openclaw/openclaw/issues/58941 | https://api.github.com/repos/openclaw/openclaw/issues/58941 | Regression: Discord exec approvals stop working on 2026.3.31 (works on 2026.3.28) | ## Summary
Updating OpenClaw from `2026.3.28` to `2026.3.31` caused Discord exec approvals to stop working in a previously working setup. Rolling back to `2026.3.28` restored expected behavior.
## Good / bad versions
- **Works:** `2026.3.28`
- **Breaks:** `2026.3.31`
## Error shown
When running a command that require... | closed | completed | true | 5 | [] | [] | 2026-04-01T10:24:23Z | 2026-04-28T04:54:03Z | 2026-04-25T17:58:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jonathanadams96 | 254,273,217 | U_kgDODyfmwQ | User | false |
openclaw/openclaw | 4,186,650,939 | I_kwDOQb6kR875izk7 | 59,004 | https://github.com/openclaw/openclaw/issues/59004 | https://api.github.com/repos/openclaw/openclaw/issues/59004 | [macOS] camera.snap command not available on paired node | ## 问题描述
macOS 节点 (v2026.3.31) 配对后，`camera.snap` 和 `camera.clip` 命令不可用。
## 环境
- Gateway: OpenClaw 2026.3.31
- macOS 节点: 曾蔚的MacBook Pro, macOS 26.3.0, app v2026.3.31
- 节点权限: camera=yes, 已在应用设置中开启 Allow Camera
## 复现步骤
1. 在 macOS 设备上配对 OpenClaw 应用为节点
2. 尝试执行 `openclaw nodes camera snap --node <node>`
3. 返回错误: `node co... | closed | completed | true | 1 | [] | [] | 2026-04-01T12:44:47Z | 2026-04-28T04:54:02Z | 2026-04-25T17:58:24Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nightq | 3,429,433 | MDQ6VXNlcjM0Mjk0MzM= | User | false |
openclaw/openclaw | 4,186,692,898 | I_kwDOQb6kR875i90i | 59,010 | https://github.com/openclaw/openclaw/issues/59010 | https://api.github.com/repos/openclaw/openclaw/issues/59010 | [Bug]: Exec approval popups cannot be disabled — flood of popups with no way to turn them off | ### Summary
After updating to 2026.4.1, the web UI is flooded with "Exec approval needed" popups that **cannot be disabled**. Setting `tools.exec.ask = "off"` has no effect on existing sessions. The popups show `Ask: off` but still appear. Clicking "Always allow" or "Deny" just clears that one popup — the next exec tr... | closed | completed | true | 3 | [] | [] | 2026-04-01T12:51:40Z | 2026-04-28T04:54:01Z | 2026-04-25T17:58:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fwends | 4,113,483 | MDQ6VXNlcjQxMTM0ODM= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.