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 | 4,346,505,972 | I_kwDOQb6kR88AAAABAxJq9A | 73,829 | https://github.com/openclaw/openclaw/issues/73829 | https://api.github.com/repos/openclaw/openclaw/issues/73829 | Auto-compaction produces empty-content assistant messages that hard-kill sessions (Venice HTTP 400) | ## Summary
When auto-compaction fails or is interrupted mid-run, it can write structurally invalid assistant messages with `content: []` (empty array) into the session JSONL file. Every subsequent turn then returns `HTTP 400 (no body)` from Venice — zero input tokens, zero output tokens — because the API rejects a con... | closed | completed | true | 1 | [] | [] | 2026-04-28T21:59:51Z | 2026-05-01T04:41:28Z | 2026-04-28T22:05:06Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | axonrelaybot | 275,827,808 | U_kgDOEHDMYA | User | false |
openclaw/openclaw | 4,337,272,900 | I_kwDOQb6kR88AAAABAoWIRA | 72,915 | https://github.com/openclaw/openclaw/issues/72915 | https://api.github.com/repos/openclaw/openclaw/issues/72915 | Bug: DeepSeek V4 reasoning_content silently dropped, causing 400 on multi-turn conversations | ## Bug Description
When using DeepSeek V4 models (`deepseek-v4-pro`, `deepseek-v4-flash`) with multi-turn conversations, the assistant's `reasoning_content` from previous turns is silently dropped when constructing follow-up API requests. This causes DeepSeek's API to return a 400 error:
> "The reasoning_content in t... | closed | completed | true | 2 | [] | [] | 2026-04-27T16:27:23Z | 2026-05-03T04:36:43Z | 2026-04-30T14:54:10Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | m8j8whq88q-hub | 279,378,010 | U_kgDOEKb4Wg | User | false |
openclaw/openclaw | 4,346,523,207 | I_kwDOQb6kR88AAAABAxKuRw | 73,831 | https://github.com/openclaw/openclaw/issues/73831 | https://api.github.com/repos/openclaw/openclaw/issues/73831 | [Bug]: undici HTTP/2 hang on Windows extends from Telegram polling into the LLM model dispatcher (related to #66885) | # [Bug]: undici HTTP/2 hang on Windows extends from Telegram polling into the LLM model dispatcher (related to #66885 / #10795 / #4847)
## Summary
On Windows running OpenClaw 2026.4.23 and 2026.4.26 with Node 24.13.0, all outbound `fetch`-based HTTP calls intermittently hang for 90–200 seconds before failing. This af... | closed | not_planned | true | 1 | [] | [] | 2026-04-28T22:02:53Z | 2026-05-03T04:36:30Z | 2026-04-30T14:59:40Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | joeywrightphoto | 114,463,115 | U_kgDOBtKRiw | User | false |
openclaw/openclaw | 4,346,491,239 | I_kwDOQb6kR88AAAABAxIxZw | 73,826 | https://github.com/openclaw/openclaw/issues/73826 | https://api.github.com/repos/openclaw/openclaw/issues/73826 | [Feature]: Webchat: TTS MEDIA: audio attachments render as download cards — no inline audio player | ### Summary
TTS is configured and working (messages.tts.auto: "always"). Audio files are generated correctly by any TTS provider (ElevenLabs, Chatterbox, OpenAI-compatible, etc.) and delivered as MEDIA: attachments. In webchat, they arrive as generic download cards — no inline audio player, no auto-play. This breaks a... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-28T21:57:14Z | 2026-05-01T04:41:27Z | 2026-04-28T22:08:04Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Jiminigrist | 184,861,194 | U_kgDOCwTCCg | User | false |
openclaw/openclaw | 4,110,738,161 | I_kwDOQb6kR871BOLx | 51,345 | https://github.com/openclaw/openclaw/issues/51345 | https://api.github.com/repos/openclaw/openclaw/issues/51345 | [Bug]: sessions_spawn(runtime="acp") hangs immediately — acpx never produces output, stalls for full relay timeout (6h) | ## Summary
`sessions_spawn(runtime: "acp", agentId: "claude", mode: "run")` silently hangs from the first turn. The acpx process appears to start (gateway writes the `start` system event to the stream log) but Claude Code produces zero output for the entire 6-hour relay timeout, then the stream relay times out.
This ... | closed | completed | true | 7 | [] | [] | 2026-03-21T00:24:22Z | 2026-05-01T04:41:26Z | 2026-04-28T22:10:36Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | mattgodbolt-molty | 260,133,431 | U_kgDOD4FSNw | User | false |
openclaw/openclaw | 4,111,247,537 | I_kwDOQb6kR871DKix | 51,404 | https://github.com/openclaw/openclaw/issues/51404 | https://api.github.com/repos/openclaw/openclaw/issues/51404 | 启动问候语显示错误：声称使用默认模型但实际使用 runtime 模型 | ## 问题描述
新会话启动时，问候语显示「当前使用默认模型 minimax-portal/MiniMax-M2.5」，但实际 runtime 模型是 。
## 重现步骤
1. 发送 /new 或 /reset 启动新会话
2. 查看问候语和 /status 显示的模型
## 预期 vs 实际
- **预期**：问候语应显示实际 runtime 模型（如 volcengine/ark-code-latest），或先切换到默认模型再显示
- **实际**：问候语显示默认模型名称，但 runtime 实际使用的是另一个模型
## 环境信息
- OpenClaw 版本：2026.3.13
- 模型配置：default_mode... | closed | completed | true | 2 | [] | [] | 2026-03-21T03:33:52Z | 2026-05-01T04:41:25Z | 2026-04-28T22:10:42Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | cyydark | 121,326,405 | U_kgDOBztLRQ | User | false |
openclaw/openclaw | 4,117,278,436 | I_kwDOQb6kR871aK7k | 52,499 | https://github.com/openclaw/openclaw/issues/52499 | https://api.github.com/repos/openclaw/openclaw/issues/52499 | feat: native support for coordinator + reporter multi-agent channel patterns | ## Summary
Multi-agent workflows increasingly need a "coordinator + reporter" pattern: one agent is the primary responder bound to a channel, while other agents post status updates or results to that same channel without being bound to it as a session. There is no clean, documented way to do this today.
## Current Be... | closed | completed | true | 2 | [] | [] | 2026-03-22T22:32:18Z | 2026-05-01T04:41:23Z | 2026-04-28T22:10:57Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,118,179,104 | I_kwDOQb6kR871dm0g | 52,632 | https://github.com/openclaw/openclaw/issues/52632 | https://api.github.com/repos/openclaw/openclaw/issues/52632 | Channel removal incomplete: WhatsApp credentials cause channel to reappear on restart | **Problem:** Removing WhatsApp via `openclaw configure channels` does not fully remove the channel. Gateway restart recreates the `channels.whatsapp` config.
**Reproduction steps:**
1. Run `openclaw configure channels` and remove WhatsApp channel
2. Verify `channels.whatsapp` is removed from openclaw.json
3. Restart t... | closed | completed | true | 3 | [] | [] | 2026-03-23T04:39:32Z | 2026-05-01T04:41:22Z | 2026-04-28T22:11:04Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | ZinkDifferent | 5,345,100 | MDQ6VXNlcjUzNDUxMDA= | User | false |
openclaw/openclaw | 4,120,570,874 | I_kwDOQb6kR871muv6 | 52,840 | https://github.com/openclaw/openclaw/issues/52840 | https://api.github.com/repos/openclaw/openclaw/issues/52840 | Security audit false-positive: lossless-claw engine.ts:829 flagged as potential-exfiltration | Hi OpenClaw team,
Reported by **Claw-II** on Android runtime.
## Summary
`openclaw security audit --deep` flags `lossless-claw` with `potential-exfiltration` at:
- `~/.openclaw/extensions/lossless-claw/src/engine.ts:829`
This appears to be a **false positive** based on static pattern matching.
## Why this looks fa... | closed | not_planned | true | 1 | [] | [] | 2026-03-23T12:42:23Z | 2026-05-01T04:41:21Z | 2026-04-28T22:11:10Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | thoitiettxl-cyber | 231,700,119 | U_kgDODc92lw | User | false |
openclaw/openclaw | 4,346,562,868 | I_kwDOQb6kR88AAAABAxNJNA | 73,832 | https://github.com/openclaw/openclaw/issues/73832 | https://api.github.com/repos/openclaw/openclaw/issues/73832 | Context token count shows 0 for Ollama provider despite usage data being returned | ## Bug Description
When using the Ollama provider (ollama/phi3.5:latest), the session status shows **0/197k tokens (0%)** even after active conversation. The context counter doesn't track token usage for Ollama models.
## Root Cause
Ollama's OpenAI-compatible API (http://localhost:11434/v1/chat/completions) **does ret... | closed | completed | true | 1 | [] | [] | 2026-04-28T22:10:55Z | 2026-05-01T04:41:20Z | 2026-04-28T22:15:39Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | vibhud | 28,487,841 | MDQ6VXNlcjI4NDg3ODQx | User | false |
openclaw/openclaw | 4,346,566,407 | I_kwDOQb6kR88AAAABAxNXBw | 73,833 | https://github.com/openclaw/openclaw/issues/73833 | https://api.github.com/repos/openclaw/openclaw/issues/73833 | [Bug]: Bundled ACPX extension missing plugin-local acpx/codex-acp binaries after 2026.4.26 update | ## Bug
After updating to OpenClaw `2026.4.26`, ACP/Codex runtime spawn can be unavailable because the bundled ACPX extension points at a plugin-local `codex-acp` binary that is not installed/staged.
## Local evidence
OpenClaw config points Codex ACP at:
```text
/opt/homebrew/lib/node_modules/openclaw/dist/extension... | closed | completed | true | 2 | [] | [] | 2026-04-28T22:11:41Z | 2026-05-01T04:41:19Z | 2026-04-28T22:16:34Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | only21mil | 263,261,067 | U_kgDOD7ELiw | User | false |
openclaw/openclaw | 4,288,745,150 | I_kwDOQb6kR87_oQ6- | 68,655 | https://github.com/openclaw/openclaw/issues/68655 | https://api.github.com/repos/openclaw/openclaw/issues/68655 | bug(web-ui): streaming markdown flickers — full DOM replacement on every token via unsafeHTML | ## Summary
- **Problem:** During streaming, every new token causes the full message container to flicker. Text visibly disappears and re-appears, especially noticeable with paragraphs and code fences.
- **Why it matters:** Any skill that produces structured markdown output (code blocks, JSON fences, multi-paragraph re... | open | null | false | 1 | [] | [] | 2026-04-18T17:46:41Z | 2026-05-02T14:22:11Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | perlowjanv | 201,165,108 | U_kgDOC_2JNA | User | false |
openclaw/openclaw | 4,346,386,854 | I_kwDOQb6kR88AAAABAxCZpg | 73,814 | https://github.com/openclaw/openclaw/issues/73814 | https://api.github.com/repos/openclaw/openclaw/issues/73814 | [Bug]: Installer hangs and truncates function "warn_shell_path_missing_di" in install.sh presumably due to stdin consumption from "curl | bash" | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
Running the install command "curl -fsSL https://openclaw.ai/install.sh | bash" from the documentation causes the script to hang indefinitely (for more than 20 minutes, and multiple times) and throw a "command not found" error ... | open | null | false | 5 | [
"bug",
"bug:crash"
] | [] | 2026-04-28T21:36:30Z | 2026-05-03T03:37:46Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | ItsMeForLua | 171,505,882 | U_kgDOCjj42g | User | false |
openclaw/openclaw | 4,346,278,805 | I_kwDOQb6kR88AAAABAw7zlQ | 73,803 | https://github.com/openclaw/openclaw/issues/73803 | https://api.github.com/repos/openclaw/openclaw/issues/73803 | [Bug]: Telegram /new and /reset reset context but still trigger empty OpenAI Responses API error | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
In OpenClaw 2026.4.26, Telegram /new and /reset reset the context according to /status, but still return an empty OpenAI Responses API error to Telegram.
### Steps to reproduce
1. Run OpenClaw 2026.4.26 with ... | closed | completed | true | 6 | [
"bug",
"bug:behavior"
] | [] | 2026-04-28T21:16:43Z | 2026-05-03T04:37:08Z | 2026-04-30T14:08:13Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | bignyong | 280,142,767 | U_kgDOELKjrw | User | false |
openclaw/openclaw | 4,346,707,096 | I_kwDOQb6kR88AAAABAxV8mA | 73,845 | https://github.com/openclaw/openclaw/issues/73845 | https://api.github.com/repos/openclaw/openclaw/issues/73845 | Feature: onSessionEnd agent turn — let the agent run cleanup before session reset | ## Summary
When a session resets (via `/new`, idle timeout, or compaction), the agent has no opportunity to run a final turn before the session is destroyed. Session end protocol tasks (writing daily notes, saving context) can only be done via cron workarounds.
## Current Behavior
- `session_end` and `before_reset` ... | closed | not_planned | true | 2 | [] | [] | 2026-04-28T22:44:35Z | 2026-05-03T04:35:00Z | 2026-04-30T21:14:12Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | louria | 24,884,925 | MDQ6VXNlcjI0ODg0OTI1 | User | false |
openclaw/openclaw | 4,346,716,072 | I_kwDOQb6kR88AAAABAxWfqA | 73,846 | https://github.com/openclaw/openclaw/issues/73846 | https://api.github.com/repos/openclaw/openclaw/issues/73846 | Docker/Unraid: 2026.4.24 and 2026.4.25 gateway never becomes healthy; doctor --fix does not recover, 2026.4.26 doctor still hangs | ### Bug type
Crash / startup failure / doctor repair failure
### Summary
On Docker/Unraid, OpenClaw `2026.4.24` and `2026.4.25` never bring the gateway up. Docker reports the container as unhealthy, but this is not just a Docker healthcheck false positive: `/healthz` never reports healthy and the gateway does not be... | closed | completed | true | 1 | [] | [] | 2026-04-28T22:46:53Z | 2026-05-04T04:37:01Z | 2026-05-01T09:44:02Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | miguelarios | 24,581,058 | MDQ6VXNlcjI0NTgxMDU4 | User | false |
openclaw/openclaw | 4,089,096,050 | I_kwDOQb6kR87zuqdy | 49,080 | https://github.com/openclaw/openclaw/issues/49080 | https://api.github.com/repos/openclaw/openclaw/issues/49080 | Live channel sessions keep stale policy context after durable rule changes | ## Problem
Durable rule/policy changes (for example Discord repo whitelist / scope rules stored in files like `USER.md` / `MEMORY.md`) do not reliably become active in already-running channel sessions.
In practice, after changing a whitelist or similar rule, the operator may need to manually go into each relevant liv... | closed | completed | true | 4 | [] | [] | 2026-03-17T14:53:08Z | 2026-05-01T04:41:17Z | 2026-04-28T22:53:28Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | SorenOpenClaw | 267,805,828 | U_kgDOD_ZkhA | User | false |
openclaw/openclaw | 4,346,398,120 | I_kwDOQb6kR88AAAABAxDFqA | 73,816 | https://github.com/openclaw/openclaw/issues/73816 | https://api.github.com/repos/openclaw/openclaw/issues/73816 | [Bug] Cron job reports status 'ok' immediately after sessions_yield, before subagent completes | ## Summary
When a cron job with calls after spawning a subagent, the cron infrastructure immediately marks the job as and records a short (~25 seconds), even though the subagent continues running for several minutes and the parent session later resumes to process the result.
This is a **monitoring/observability b... | closed | not_planned | true | 1 | [] | [] | 2026-04-28T21:38:45Z | 2026-05-01T04:41:16Z | 2026-04-28T22:53:43Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | lunan-stellaraether | 266,393,738 | U_kgDOD-DYig | User | false |
openclaw/openclaw | 4,070,292,578 | I_kwDOQb6kR87ym7xi | 45,003 | https://github.com/openclaw/openclaw/issues/45003 | https://api.github.com/repos/openclaw/openclaw/issues/45003 | [Feature]: Script to restore session history archived by new daily-reset mechanism (introduced v2026.2+) | ### Summary
A utility script to automatically restore and re-index session transcript files
that were archived as `.reset.<timestamp>` by the new daily session-reset mechanism.
### Problem to solve
Starting with PR #14949 (Feb 13, 2026) and PR #35493 (Mar 6, 2026), OpenClaw
began archiving old transcript files as `<... | open | null | false | 5 | [
"enhancement"
] | [] | 2026-03-13T10:42:00Z | 2026-05-04T13:19:04Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | CadanHu | 39,733,381 | MDQ6VXNlcjM5NzMzMzgx | User | false |
openclaw/openclaw | 4,077,019,718 | I_kwDOQb6kR87zAmJG | 46,641 | https://github.com/openclaw/openclaw/issues/46641 | https://api.github.com/repos/openclaw/openclaw/issues/46641 | Feature Request: User-facing notification before auto-compaction | ## Problem
When working interactively with a user on complex, multi-step tasks (e.g., tax return preparation in a browser), auto-compaction fires silently and wipes active task context. The pre-compaction memory flush is model-only and silent — the user has no idea compaction is about to happen and no opportunity to t... | open | null | false | 1 | [] | [] | 2026-03-14T23:16:20Z | 2026-05-04T00:44:22Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | adminmoe527 | 266,539,591 | U_kgDOD-MSRw | User | false |
openclaw/openclaw | 4,077,041,050 | I_kwDOQb6kR87zArWa | 46,647 | https://github.com/openclaw/openclaw/issues/46647 | https://api.github.com/repos/openclaw/openclaw/issues/46647 | Slack system prompt incorrectly checks inlineButtons instead of interactiveReplies | ## Summary
The system prompt hint for Slack inline buttons checks `runtimeCapabilitiesLower.has("inlinebuttons")` (line 377 of `system-prompt.ts`), but the actual Slack interactive reply directive pipeline is gated on `channels.slack.capabilities.interactiveReplies` (via `isSlackInteractiveRepliesEnabled()` in `slack/... | closed | completed | true | 5 | [] | [] | 2026-03-14T23:25:45Z | 2026-05-02T04:26:33Z | 2026-04-30T02:42:04Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | jeremykoerber | 2,133,982 | MDQ6VXNlcjIxMzM5ODI= | User | false |
openclaw/openclaw | 4,077,474,831 | I_kwDOQb6kR87zCVQP | 46,810 | https://github.com/openclaw/openclaw/issues/46810 | https://api.github.com/repos/openclaw/openclaw/issues/46810 | Control UI: Exec approval modal with command highlighting | ## Feature Request
### Summary
Replace the current text-based exec approval flow with a proper modal dialog in the Control UI.
### Motivation
The current approval UX requires the user to copy an /approve code from chat and run it in a separate terminal or chat message. This is friction-heavy and error-prone, especial... | open | null | false | 1 | [] | [] | 2026-03-15T03:25:52Z | 2026-05-04T00:44:43Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | allisonshawai-ctrl | 262,603,479 | U_kgDOD6cC1w | User | false |
openclaw/openclaw | 4,077,502,313 | I_kwDOQb6kR87zCb9p | 46,830 | https://github.com/openclaw/openclaw/issues/46830 | https://api.github.com/repos/openclaw/openclaw/issues/46830 | [Feature]: Message Delivery Failure Notification | ### Summary
When sending messages to channels (Feishu, Telegram, etc.), if the API call fails (network issues, rate limits, API errors), the agent may believe the message was sent successfully, but the user never receives it.
**Observed behavior (from logs):**
```
feishu[default] final reply failed: AxiosError: Reque... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-15T03:48:33Z | 2026-05-04T03:59:37Z | null | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | sunxzycn | 104,638,670 | U_kgDOBjyozg | User | false |
openclaw/openclaw | 4,077,622,311 | I_kwDOQb6kR87zC5Qn | 46,899 | https://github.com/openclaw/openclaw/issues/46899 | https://api.github.com/repos/openclaw/openclaw/issues/46899 | [Bug]: Cron Job Delivery Channel Mismatch - Telegram Configured but Feishu Used | ### Bug type
Regression (worked before, now fails)
### Summary
Cron job configured for Telegram delivery uses Feishu channel with invalid group ID
### Steps to reproduce
1. Configure a cron job with delivery.channel = "telegram"
2. Set delivery.to to a valid Telegram user ID
3. Wait for cron job to execute at sche... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-03-15T05:06:09Z | 2026-05-04T00:33:17Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | Augusta-waveryu | 255,800,970 | U_kgDODz82ig | User | false |
openclaw/openclaw | 4,077,647,535 | I_kwDOQb6kR87zC_av | 46,919 | https://github.com/openclaw/openclaw/issues/46919 | https://api.github.com/repos/openclaw/openclaw/issues/46919 | [Feature]: Token URL Disabled Via Config | ### Summary
The token URL is kind of dumb. Replace with a simple config flag. Currently, I have a fork with this as a perm-feature to rip out of all updates. It's a good workflow, removed NO_REPLY, replaced session memory with full lancedb full photographic memory, not summaries, and not when it feels like it, all ... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-15T05:26:47Z | 2026-05-04T00:33:21Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | NewCoffee7477 | 114,102,862 | U_kgDOBs0STg | User | false |
openclaw/openclaw | 4,077,679,150 | I_kwDOQb6kR87zDHIu | 46,935 | https://github.com/openclaw/openclaw/issues/46935 | https://api.github.com/repos/openclaw/openclaw/issues/46935 | /model available values are not round-trippable; bare model names get prefixed with current provider | ## Summary
`/model` shows bare model names in `Available:`, but those values are not reliably accepted by `/model <value>`.
When the current model provider is `openai-codex`, entering a bare name like:
```text
/model claude-opus-4-6
```
is interpreted as:
```text
openai-codex/claude-opus-4-6
```
which then fails ... | open | null | false | 2 | [] | [] | 2026-03-15T05:50:56Z | 2026-05-04T01:32:16Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | EdwardOptimization | 248,197,022 | U_kgDODssvng | User | false |
openclaw/openclaw | 4,077,681,599 | I_kwDOQb6kR87zDHu_ | 46,938 | https://github.com/openclaw/openclaw/issues/46938 | https://api.github.com/repos/openclaw/openclaw/issues/46938 | 模型超时导致记忆维护失败，建议增加 session 临时存储机制 | ## 问题描述
当模型连续超时时，AI 无法执行记忆维护（写入 memory 文件），导致对话记录丢失。
### 实际案例
今天凌晨 00:00-01:00 期间，GLM-5 模型连续超时 6 次（每次 10 分钟），导致：
1. AI 无法响应用户消息
2. 没有执行记忆维护，没有创建 `memory/2026-03-15.md`
3. 用户发送 "怎么样了？"、"总结一下吧" 都没有收到响应
Session 文件存在：
```
~/.openclaw/agents/main/sessions/fde6169e-1566-48c3-839e-c2534fa05eec.jsonl.reset.2026-03-14T20-01... | open | null | false | 1 | [] | [] | 2026-03-15T05:53:00Z | 2026-05-04T02:38:26Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | tociyou1995 | 150,312,743 | U_kgDOCPWXJw | User | false |
openclaw/openclaw | 4,078,070,215 | I_kwDOQb6kR87zEmnH | 47,167 | https://github.com/openclaw/openclaw/issues/47167 | https://api.github.com/repos/openclaw/openclaw/issues/47167 | Skipped group messages behind requireMention do not reach message hooks, which blocks full-history archive plugins | ## Summary
I’m building an OpenClaw plugin for full conversation history retrieval.
The goal is not to replace curated memory (`MEMORY.md`, daily notes, etc.), but to add a searchable raw-message layer that users can fall back to when memory does not contain enough detail.
Examples:
- exact wording
- chronology / t... | open | null | false | 3 | [] | [] | 2026-03-15T10:01:24Z | 2026-05-04T02:38:29Z | null | CONTRIBUTOR | null | 20260504T030031Z | 2026-05-04T03:00:31Z | dashhuang | 125,997,726 | U_kgDOB4KSng | User | false |
openclaw/openclaw | 4,078,628,741 | I_kwDOQb6kR87zGu-F | 47,444 | https://github.com/openclaw/openclaw/issues/47444 | https://api.github.com/repos/openclaw/openclaw/issues/47444 | Workflow responses leak internal <function_calls> scaffolding into user-visible chat | ## Summary
In workflow-style sessions, user-visible replies sometimes include internal tool-call scaffolding instead of only the clean final result.
## Example
User sees leaked content like:
```xml
<function_calls>
<invoke name="exec">
<parameter name="command">cd /Users/mikeroberts/.openclaw/workspace && node script... | open | null | false | 1 | [] | [] | 2026-03-15T15:48:20Z | 2026-05-04T01:21:25Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | 5toCode | 70,207,997 | MDQ6VXNlcjcwMjA3OTk3 | User | false |
openclaw/openclaw | 4,078,745,914 | I_kwDOQb6kR87zHLk6 | 47,496 | https://github.com/openclaw/openclaw/issues/47496 | https://api.github.com/repos/openclaw/openclaw/issues/47496 | [Bug]: { "status": "error", "tool": "read", "error": "ENOENT: no such file or directory, access '/root/.openclaw/workspace/components/ui/label.tsx'" } | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Deployed on the Durban system, frequent UI issues occur in the chat interface, including incomplete chat loading and frequent tool-related prompts.
<img width="1150" height="1149" alt="Image" src="https://github.com/user-attachments/assets... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T16:53:38Z | 2026-05-04T02:26:43Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | AD696 | 127,026,717 | U_kgDOB5JGHQ | User | false |
openclaw/openclaw | 4,078,910,508 | I_kwDOQb6kR87zHzws | 47,561 | https://github.com/openclaw/openclaw/issues/47561 | https://api.github.com/repos/openclaw/openclaw/issues/47561 | Bug: /new session can be stranded by routing-mode changes and resume an older session | ## Summary
`/new` can successfully create a fresh session, but that fresh session can later be abandoned if the session-routing mode changes (for example reverting from per-channel-peer back to unified), causing the conversation to resume on an older session with stale cached state.
## Why this is a bug
From the ope... | open | null | false | 4 | [] | [] | 2026-03-15T18:18:48Z | 2026-05-04T01:33:03Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,079,025,078 | I_kwDOQb6kR87zIPu2 | 47,591 | https://github.com/openclaw/openclaw/issues/47591 | https://api.github.com/repos/openclaw/openclaw/issues/47591 | Feature: Silence Watchdog + Response Delivery Confirmation — runtime-enforced operator visibility | ## Problem
Agent-level (model-layer) progress beacons and delivery tracking are structurally unreliable. During sustained tool-call chains, the model cannot pause to evaluate its own silence duration — it proceeds to the next tool call. This means:
1. **Mid-turn silence:** The agent is actively working (making tool c... | open | null | false | 2 | [] | [] | 2026-03-15T19:24:09Z | 2026-05-04T01:33:10Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | CaiusPrime | 262,411,762 | U_kgDOD6QV8g | User | false |
openclaw/openclaw | 4,079,843,648 | I_kwDOQb6kR87zLXlA | 47,805 | https://github.com/openclaw/openclaw/issues/47805 | https://api.github.com/repos/openclaw/openclaw/issues/47805 | [Feature Request] Allow importing .jsonl session files to appear in session menu | ### Summary
Allow users to import .jsonl session backup files so they appear in the session menu for easy browsing and retrieval.
### Problem to solve
Currently, when users manually copy .jsonl session files to the sessions/ directory, these sessions do not appear in the OpenClaw session menu. The session menu only ... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-16T02:20:47Z | 2026-05-04T14:25:50Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | pekywang | 239,189,800 | U_kgDODkG_KA | User | false |
openclaw/openclaw | 4,079,852,984 | I_kwDOQb6kR87zLZ24 | 47,811 | https://github.com/openclaw/openclaw/issues/47811 | https://api.github.com/repos/openclaw/openclaw/issues/47811 | Multi-profile profile targeting/config truth is ambiguous with ~/.openclaw-* homes on macOS | # OpenClaw multi-profile config / targeting issue (2026-03-15)
## Why this is being filed
We are finishing OpenClawBrain v2 on a real macOS host with multiple live OpenClaw profiles. The current blocker is not that basic LCM/lossless-claw is broken — that is working on Tern right now — but that the surrounding profile... | open | null | false | 1 | [] | [] | 2026-03-16T02:24:46Z | 2026-05-04T14:02:21Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | jonathangu | 2,887,205 | MDQ6VXNlcjI4ODcyMDU= | User | false |
openclaw/openclaw | 4,080,011,759 | I_kwDOQb6kR87zMAnv | 47,849 | https://github.com/openclaw/openclaw/issues/47849 | https://api.github.com/repos/openclaw/openclaw/issues/47849 | [Feature]: openclaw webui ,实现多用户登录 | ### Summary
OpenClaw 的认证模式只有这几种：
none - 无认证
token - 单 token（所有人共享）
password - 单密码（所有人用同一个密码）
trusted-proxy - 依赖反向代理认证
没有内置的多用户系统。
这意味着所有人用同一个 token ，或密码共享同一组会话。 提出实现多用户登录的建议，让会话隔离。实现多用户操作，创建news session 不会看到别人的session记录
### Problem to solve
想用实现的效果：webui 提供多用户登录，让会话隔离。实现多用户操作，创建news session 不会看到别人的session记录
### Pr... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-16T03:30:05Z | 2026-05-03T04:34:46Z | 2026-04-30T22:51:42Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | caijx168 | 54,843,188 | MDQ6VXNlcjU0ODQzMTg4 | User | false |
openclaw/openclaw | 4,080,024,201 | I_kwDOQb6kR87zMDqJ | 47,852 | https://github.com/openclaw/openclaw/issues/47852 | https://api.github.com/repos/openclaw/openclaw/issues/47852 | Feature: Per-agent generation parameter configuration (temperature, top_p, top_k, penalties, max_tokens) | Currently OpenClaw uses fixed defaults for LLM generation parameters across all agents. The underlying code already accepts `temperature` as a parameter, but there's no way to configure generation params per-agent in `openclaw.json`.
## Proposed change
Expose generation parameters as optional fields in each agent's c... | open | null | false | 2 | [] | [] | 2026-03-16T03:35:59Z | 2026-05-04T14:02:27Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ThalynLabs | 38,703,365 | MDQ6VXNlcjM4NzAzMzY1 | User | false |
openclaw/openclaw | 4,080,076,462 | I_kwDOQb6kR87zMQau | 47,865 | https://github.com/openclaw/openclaw/issues/47865 | https://api.github.com/repos/openclaw/openclaw/issues/47865 | 系统消息被错误推送到用户聊天界面 | ## Bug 描述
系统内部消息（System Event）被错误地推送到用户聊天界面，尽管配置了 `delivery: { mode: "none" }`。
## 复现步骤
1. 创建 cron 任务，配置如下：
```json
{
"name": "Pulse-Heartbeat",
"schedule": { "kind": "every", "everyMs": 300000 },
"sessionTarget": "main",
"payload": { "kind": "systemEvent", "text": "Pulse heartbeat: ..." },
"delivery": { "... | open | null | false | 2 | [] | [] | 2026-03-16T03:59:48Z | 2026-04-30T21:58:10Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | xlfc-yu | 267,224,686 | U_kgDOD-2Gbg | User | false |
openclaw/openclaw | 4,080,147,740 | I_kwDOQb6kR87zMh0c | 47,884 | https://github.com/openclaw/openclaw/issues/47884 | https://api.github.com/repos/openclaw/openclaw/issues/47884 | [Bug]: memory_search tool fails with "fetch failed" despite embedding provider configured | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
memory_search tool returns "fetch failed" error despite embedding provider correctly configured
### Steps to reproduce
1. Configure embedding provider in openclaw.json:
{
"bailian-embedding": {
"baseUrl": "https://dashscope... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-16T04:26:59Z | 2026-05-04T14:53:50Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | AllenSupermanxiaodingdang | 266,199,897 | U_kgDOD93jWQ | User | false |
openclaw/openclaw | 4,080,151,408 | I_kwDOQb6kR87zMitw | 47,887 | https://github.com/openclaw/openclaw/issues/47887 | https://api.github.com/repos/openclaw/openclaw/issues/47887 | [Bug]: iOS LAN gateway connects are forced to wss, and bootstrap-only setup codes fail onboarding | ### Bug type
Regression (worked before, now fails)
### Summary
I tracked a cluster of iOS gateway onboarding failures to four related issues.
### Steps to reproduce
- Manual LAN connect to `ws://192.168.x.x:18789` or `ws://openclaw.local:18789` hangs/fails before auth.
- The iOS app discovers the gateway and HTT... | open | null | false | 2 | [
"bug",
"regression"
] | [
"BunsDev",
"Copilot"
] | 2026-03-16T04:28:19Z | 2026-05-05T06:45:30Z | null | NONE | null | 20260505T090045Z | 2026-05-05T09:00:45Z | styles01 | 11,393,400 | MDQ6VXNlcjExMzkzNDAw | User | false |
openclaw/openclaw | 4,080,433,565 | I_kwDOQb6kR87zNnmd | 47,941 | https://github.com/openclaw/openclaw/issues/47941 | https://api.github.com/repos/openclaw/openclaw/issues/47941 | Context breakdown can label virtual bootstrap files as 'undefined' instead of falling back to path/basename | ## Summary
`/context breakdown` (or similar context reporting) can show a virtual bootstrap file as `undefined` even when the injected content is present and counted.
In the observed case, a bootstrap hook injected a virtual file with:
- `path: 'SELF_IMPROVEMENT_REMINDER.md'`
- content present
- `virtual: true`
The c... | open | null | false | 3 | [] | [] | 2026-03-16T06:06:43Z | 2026-05-04T14:02:47Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | i0ivi0i | 105,140,938 | U_kgDOBkRSyg | User | false |
openclaw/openclaw | 4,080,483,756 | I_kwDOQb6kR87zNz2s | 47,951 | https://github.com/openclaw/openclaw/issues/47951 | https://api.github.com/repos/openclaw/openclaw/issues/47951 | ACP Codex long tasks can be accepted but never start (empty child session transcript) | ## Summary
ACP Codex sessions can execute short/medium tasks, but some longer implementation-style tasks are accepted and never actually start. The child session is created, but its transcript remains empty even after waiting 30 seconds.
## Environment
- OpenClaw with bundled `acpx` plugin loaded
- ACP backend: `acp... | open | null | false | 4 | [] | [] | 2026-03-16T06:19:48Z | 2026-05-04T14:02:49Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | YeTianXingShi | 42,585,691 | MDQ6VXNlcjQyNTg1Njkx | User | false |
openclaw/openclaw | 4,080,650,023 | I_kwDOQb6kR87zOccn | 47,979 | https://github.com/openclaw/openclaw/issues/47979 | https://api.github.com/repos/openclaw/openclaw/issues/47979 | Control UI Dashboard v2 freezes on Chrome 146 (works fine in Firefox) | ## Description
The Control UI (Dashboard v2, introduced in 2026.3.12 #41503) completely freezes in Chrome 146 — the page loads and the WebSocket connects, but the UI is unresponsive (cannot click on chats, sidebar does not render chat list, no RPC calls are sent). Firefox 148 works perfectly with the same gateway.
##... | open | null | false | 3 | [] | [] | 2026-03-16T07:05:43Z | 2026-05-04T14:02:53Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | tavomed | 29,103,177 | MDQ6VXNlcjI5MTAzMTc3 | User | false |
openclaw/openclaw | 4,081,014,603 | I_kwDOQb6kR87zP1dL | 48,042 | https://github.com/openclaw/openclaw/issues/48042 | https://api.github.com/repos/openclaw/openclaw/issues/48042 | browser.cdpUrl config ignored — tool still looks for DevToolsActivePort file | ## Bug
When `browser.cdpUrl` is set in config (e.g. `http://localhost:9222`), the browser tool ignores it and still looks for the `DevToolsActivePort` file at the default Chrome profile path.
## Config
```json
{
"browser": {
"enabled": true,
"cdpUrl": "http://localhost:9222",
"attachOnly": false,
"d... | open | null | false | 3 | [] | [] | 2026-03-16T08:32:54Z | 2026-04-30T21:58:12Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Max-Resilient | 64,356,730 | MDQ6VXNlcjY0MzU2NzMw | User | false |
openclaw/openclaw | 4,081,224,146 | I_kwDOQb6kR87zQonS | 48,082 | https://github.com/openclaw/openclaw/issues/48082 | https://api.github.com/repos/openclaw/openclaw/issues/48082 | [Bug]: exec tool SIGSEGV on Feiniu NAS (custom kernel 6.12.18-trim) /exec 工具在飞牛 NAS 上崩溃 (SIGSEGV，定制内核 6.12.18-trim) | ### Bug type
Crash (process/app exits or hangs)
### Summary
## Summary
The `exec` tool crashes with SIGSEGV on Feiniu NAS environment, while Node's native `spawn` and manual `node-pty` tests work correctly.
## Environment
- **OS:** Debian 12 (bookworm) on Feiniu NAS
- **Kernel:** 6.12.18-trim (custom kernel)
- **... | open | null | false | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-16T09:17:57Z | 2026-05-04T15:47:20Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ltw55568 | 265,176,022 | U_kgDOD85D1g | User | false |
openclaw/openclaw | 4,081,392,464 | I_kwDOQb6kR87zRRtQ | 48,104 | https://github.com/openclaw/openclaw/issues/48104 | https://api.github.com/repos/openclaw/openclaw/issues/48104 | Model safety/alignment can block explicitly authorized operational tasks (e.g. SSH diagnostics) | ## Summary
OpenClaw sometimes refuses or avoids performing clearly authorized operational tasks because the underlying model applies overly broad safety / moral boundaries.
This is especially noticeable in admin / operator workflows such as:
- connecting to a server via SSH
- checking service health
- reading logs
-... | open | null | false | 3 | [] | [] | 2026-03-16T09:48:53Z | 2026-05-04T14:25:58Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Lulu-Grant | 141,903,813 | U_kgDOCHVHxQ | User | false |
openclaw/openclaw | 4,081,855,101 | I_kwDOQb6kR87zTCp9 | 48,159 | https://github.com/openclaw/openclaw/issues/48159 | https://api.github.com/repos/openclaw/openclaw/issues/48159 | [Feature] Ephemeral / Secure Chat Mode — per-session privacy with no persistence | ## Summary
OpenClaw currently persists all session data (transcripts, LCM summaries, memory flushes, compaction) with no way to opt out on a per-session basis. Users need a "secure chat" mode — analogous to Telegram Secret Chats or WhatsApp disappearing messages — where conversations leave zero trace in the agent's lo... | open | null | false | 3 | [] | [] | 2026-03-16T11:18:03Z | 2026-04-30T21:58:14Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Fato07 | 40,800,448 | MDQ6VXNlcjQwODAwNDQ4 | User | false |
openclaw/openclaw | 4,082,041,523 | I_kwDOQb6kR87zTwKz | 48,179 | https://github.com/openclaw/openclaw/issues/48179 | https://api.github.com/repos/openclaw/openclaw/issues/48179 | enhancement: Optimize voice message processing pipeline for Telegram | ## Summary
Improve voice message handling latency and throughput in Telegram group chats.
## Problem Statement
Current voice message processing in Telegram sessions exhibits suboptimal latency, particularly for:
- Audio attachment detection and download
- Transcription pipeline (faster-whisper integration)
- Text-to... | open | null | false | 2 | [] | [] | 2026-03-16T11:59:03Z | 2026-05-04T14:26:01Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Moltmanisgreat | 258,667,831 | U_kgDOD2r1Nw | User | false |
openclaw/openclaw | 4,082,434,519 | I_kwDOQb6kR87zVQHX | 48,220 | https://github.com/openclaw/openclaw/issues/48220 | https://api.github.com/repos/openclaw/openclaw/issues/48220 | [Bug]: Looks like a gui bug in openclaw Dashboard. | ### Bug type
Regression (worked before, now fails)
### Summary
after upgradeing to 2026.3.12 i update daily so dunno when it started to fail, but i see pending device in pending.json but not in openclaw cli so need to add manualy.
looks like its a warning bug? see the warning say 53K / 41K
but status shows 31k/41K
... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-16T13:16:52Z | 2026-05-01T04:41:00Z | 2026-04-29T01:07:01Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | StianNOR | 82,333,847 | MDQ6VXNlcjgyMzMzODQ3 | User | false |
openclaw/openclaw | 4,082,491,358 | I_kwDOQb6kR87zVd_e | 48,229 | https://github.com/openclaw/openclaw/issues/48229 | https://api.github.com/repos/openclaw/openclaw/issues/48229 | operator.write scope not granted for WebSocket connections when gateway.bind is 'lan' with token auth | ## Description
When running OpenClaw in Docker containers with `gateway.bind: "lan"` and `gateway.auth.mode: "token"`, the A2A gateway plugin (connecting as `role: "operator"` with `scopes: ["operator.admin", "operator.write"]`) fails with:
```
missing scope: operator.write
```
The same plugin works correctly when `... | open | null | false | 4 | [] | [] | 2026-03-16T13:27:16Z | 2026-05-04T14:03:25Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | lishixiang0705 | 220,044,982 | U_kgDODR2etg | User | false |
openclaw/openclaw | 4,082,563,629 | I_kwDOQb6kR87zVvot | 48,241 | https://github.com/openclaw/openclaw/issues/48241 | https://api.github.com/repos/openclaw/openclaw/issues/48241 | Aborted agent run emits garbage/unrelated content instead of discarding response | # Aborted agent run emits garbage/unrelated content instead of discarding response
## Description
When multiple user messages arrive in quick succession for the same agent session, the gateway occasionally aborts an in-progress run (`openclaw:prompt-error` / `"error": "aborted"`) but **still writes an assistant messa... | open | null | false | 2 | [] | [] | 2026-03-16T13:40:47Z | 2026-05-04T14:03:29Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | zeeyuu222 | 267,768,244 | U_kgDOD_XRtA | User | false |
openclaw/openclaw | 4,082,700,307 | I_kwDOQb6kR87zWRAT | 48,253 | https://github.com/openclaw/openclaw/issues/48253 | https://api.github.com/repos/openclaw/openclaw/issues/48253 | Feature Request: Abstract file scanning pipeline as Memory Plugin API | ## Feature Request
### Problem
The built-in `memorySearch` engine has a robust file scanning pipeline:
1. **Startup full scan** — indexes `MEMORY.md` + `memory/` + `extraPaths` directories
2. **Real-time file watching** — chokidar monitors for add/change/delete events
3. **Incremental sync** — hash-based change dete... | open | null | false | 1 | [] | [] | 2026-03-16T14:05:45Z | 2026-05-04T15:05:32Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | aidyfeng | 13,725,371 | MDQ6VXNlcjEzNzI1Mzcx | User | false |
openclaw/openclaw | 4,082,884,055 | I_kwDOQb6kR87zW93X | 48,279 | https://github.com/openclaw/openclaw/issues/48279 | https://api.github.com/repos/openclaw/openclaw/issues/48279 | [Bug]: agentclientprotocol SDK v0.16.x breaks ACP listSessions functionality | ### Bug type
Regression (worked before, now fails)
### Summary
As per [ACP TS SDK v0.16.0 release notes](https://github.com/agentclientprotocol/typescript-sdk/releases/tag/v0.16.0) ([related diff](https://github.com/agentclientprotocol/typescript-sdk/pull/84/changes#diff-1cc1b51c3835b557f892e733708be4074bf9237c1c017... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-03-16T14:36:30Z | 2026-04-30T21:58:15Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | thezzisu | 21,094,314 | MDQ6VXNlcjIxMDk0MzE0 | User | false |
openclaw/openclaw | 4,083,120,695 | I_kwDOQb6kR87zX3o3 | 48,304 | https://github.com/openclaw/openclaw/issues/48304 | https://api.github.com/repos/openclaw/openclaw/issues/48304 | [Feature]: Tool-level authorization policy — per-action, per-sender validation for message tool | ### Summary
In multi-user setups where each team member has a personal assistant agent, there's no reliable way to prevent non-owners from instructing an assistant to take protected actions (post in channels, send DMs to others, send emails). Prompt-based authorization rules are consistently ignored by models (tested ... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-16T15:18:56Z | 2026-05-04T15:05:34Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | gbgeka | 121,305,411 | U_kgDOBzr5Qw | User | false |
openclaw/openclaw | 4,083,133,717 | I_kwDOQb6kR87zX60V | 48,306 | https://github.com/openclaw/openclaw/issues/48306 | https://api.github.com/repos/openclaw/openclaw/issues/48306 | Bug: macOS app updater fails because appcast points to missing asset, leaving app/gateway split-brained | ## Summary
On macOS, the built-in **OpenClaw.app** updater can fail with:
> Update error! An error occurred while downloading the update. Please try again later.
In the current state, this leaves operators stuck in a bad split-version condition:
- the human cannot update the app from the app UI because the download... | open | null | false | 4 | [] | [] | 2026-03-16T15:21:05Z | 2026-05-04T14:03:45Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,083,314,815 | I_kwDOQb6kR87zYnB_ | 48,321 | https://github.com/openclaw/openclaw/issues/48321 | https://api.github.com/repos/openclaw/openclaw/issues/48321 | Feature: Inject timestamp + position metadata on images in thread history | ### Summary
Problem:
When an agent replies within a threaded conversation (Slack, Discord, etc.), thread history includes all images from every prior message. The model receives these images with no metadata about when they were sent or where they appear in the conversation flow.
This causes a specific, recurring fai... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-16T15:51:14Z | 2026-04-30T21:58:17Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | ilike4movies | 201,359,153 | U_kgDODAB_MQ | User | false |
openclaw/openclaw | 4,083,398,194 | I_kwDOQb6kR87zY7Yy | 48,338 | https://github.com/openclaw/openclaw/issues/48338 | https://api.github.com/repos/openclaw/openclaw/issues/48338 | Custom / third-party model APIs often fall back to an unclear 16K context window UX | ## Problem
When users integrate their own third-party model API through OpenClaw's custom provider / OpenAI-compatible provider flow, the effective context window often ends up behaving like **16K** unless they manually configure it correctly.
For experienced users this is annoying but manageable. For less technical ... | open | null | false | 5 | [] | [] | 2026-03-16T16:06:09Z | 2026-05-04T14:04:02Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | quqi1599 | 51,637,638 | MDQ6VXNlcjUxNjM3NjM4 | User | false |
openclaw/openclaw | 4,083,701,658 | I_kwDOQb6kR87zaFea | 48,379 | https://github.com/openclaw/openclaw/issues/48379 | https://api.github.com/repos/openclaw/openclaw/issues/48379 | [Feature]: Per-agent X-Title header for OpenRouter (agent attribution in API logs) | ### Summary
Allow per-agent X-Title header for OpenRouter so each agent is individually identifiable in API activity logs and cost tracking.
### Problem to solve
When running multiple OpenClaw agents (e.g. separate agents for backend, frontend, scraping, architecture), all requests to OpenRouter appear with X-Title:... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-16T17:05:24Z | 2026-05-04T14:53:59Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | oleg-khymchenko-kh | 18,165,522 | MDQ6VXNlcjE4MTY1NTIy | User | false |
openclaw/openclaw | 4,083,896,278 | I_kwDOQb6kR87za0_W | 48,404 | https://github.com/openclaw/openclaw/issues/48404 | https://api.github.com/repos/openclaw/openclaw/issues/48404 | [Bug]: OpenClaw v2026.3.13 fails to start on the target system due to a **segmentation fault** | ### Bug type
Regression (worked before, now fails)
### Summary
### Investigation Details
#### 1. System Environment
- **OS**: ImmortalWrt (OpenWrt fork) 2026
- **Kernel**: Linux 6.12.74 x86_64
- **Shell**: BusyBox ash (limited toolset)
- **Node.js**: v22.22.1 (meets OpenClaw's minimum requirement of v22.12+)
- **M... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-16T17:45:27Z | 2026-05-01T04:40:56Z | 2026-04-29T01:38:05Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | qiuzi | 1,689,784 | MDQ6VXNlcjE2ODk3ODQ= | User | false |
openclaw/openclaw | 4,084,187,438 | I_kwDOQb6kR87zb8Eu | 48,435 | https://github.com/openclaw/openclaw/issues/48435 | https://api.github.com/repos/openclaw/openclaw/issues/48435 | [Feature]: Slack: add unfurl_links/unfurl_media config option for bot messages | ### Summary
When the OpenClaw Slack plugin sends messages containing URLs
(e.g., LinkedIn profile links in recruiter output), Slack
auto-unfurls them into large preview cards that clutter the output.
The Slack Web API supports `unfurl_links: false` and
`unfurl_media: false` parameters on chat.postMessage.
Featur... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-16T18:47:52Z | 2026-04-30T21:58:18Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | esegev1 | 4,525,779 | MDQ6VXNlcjQ1MjU3Nzk= | User | false |
openclaw/openclaw | 4,084,295,408 | I_kwDOQb6kR87zcWbw | 48,450 | https://github.com/openclaw/openclaw/issues/48450 | https://api.github.com/repos/openclaw/openclaw/issues/48450 | Discord: thread-create returns error but thread is created (ghost duplicate threads) | ## Description
When using the `message` tool with `action=thread-create`, the tool occasionally returns an error to the agent, but Discord has actually created the thread successfully. The agent retries, creating a duplicate thread — one empty "ghost" and one with the intended content.
## Observed Behavior
1. Agent ... | closed | completed | true | 4 | [] | [] | 2026-03-16T19:11:07Z | 2026-05-04T04:33:53Z | 2026-05-02T04:22:41Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | dahifi | 8,903,067 | MDQ6VXNlcjg5MDMwNjc= | User | false |
openclaw/openclaw | 4,084,592,501 | I_kwDOQb6kR87zde91 | 48,471 | https://github.com/openclaw/openclaw/issues/48471 | https://api.github.com/repos/openclaw/openclaw/issues/48471 | [Feature Request]: one-line local bootstrap for daemon + dashboard auth + Telegram owner setup | ## Summary
OpenClaw has a strong interactive installer/onboarding flow, but there is still no single shipped one-line path that fully automates a common local setup:
1. install OpenClaw,
2. install/start or refresh the local gateway service,
3. bootstrap dashboard / Control UI auth on localhost,
4. configure a Telegr... | open | null | false | 2 | [] | [] | 2026-03-16T20:19:02Z | 2026-04-30T21:58:19Z | null | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | OpenCodeEngineer | 261,470,075 | U_kgDOD5W3ew | User | false |
openclaw/openclaw | 4,084,770,496 | I_kwDOQb6kR87zeKbA | 48,488 | https://github.com/openclaw/openclaw/issues/48488 | https://api.github.com/repos/openclaw/openclaw/issues/48488 | Lane queue has no task-level timeout — hung promises permanently block session lanes | ## Summary
Session lanes in the gateway's command queue (`src/process/command-queue.ts`) have no task-level timeout. If an enqueued task's promise never settles, the lane is permanently jammed with no automatic recovery. This affects all messaging channels and cron.
## Symptom
Webchat session stops responding perman... | open | null | false | 4 | [] | [] | 2026-03-16T21:00:49Z | 2026-04-29T00:15:31Z | null | NONE | null | 20260429T030016Z | 2026-04-29T03:00:16Z | kyletabor | 6,003,044 | MDQ6VXNlcjYwMDMwNDQ= | User | false |
openclaw/openclaw | 4,084,958,897 | I_kwDOQb6kR87ze4ax | 48,503 | https://github.com/openclaw/openclaw/issues/48503 | https://api.github.com/repos/openclaw/openclaw/issues/48503 | [Feature]: Enrich before_tool_call event with action classification and input provenance |
### Background
We implemented a POC for CaMeL style prompt injection defense (DeepMind, 2024) as a native TypeScript port for OpenClaw. The full implementation touches too many core systems to land as a single PR: taint tracking, policy enforcement, approval routing, session isolation, and provenance propagation all... | open | null | false | 2 | [] | [] | 2026-03-16T21:49:41Z | 2026-04-30T22:39:12Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,085,013,329 | I_kwDOQb6kR87zfFtR | 48,514 | https://github.com/openclaw/openclaw/issues/48514 | https://api.github.com/repos/openclaw/openclaw/issues/48514 | WebChat UI freezes and becomes unresponsive with accumulated session history | ## Problem Description
The WebChat UI becomes extremely slow and unresponsive when session history accumulates over time.
### Symptoms
- **Page freezing**: UI becomes sluggish and takes a long time to respond
- **Slow loading**: Chat messages take a long time to appear
- **Cannot interact**: Unable to click dialog bo... | open | null | false | 3 | [] | [] | 2026-03-16T22:03:31Z | 2026-04-30T22:39:13Z | null | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | luzhidong | 15,848,762 | MDQ6VXNlcjE1ODQ4NzYy | User | false |
openclaw/openclaw | 4,085,020,557 | I_kwDOQb6kR87zfHeN | 48,516 | https://github.com/openclaw/openclaw/issues/48516 | https://api.github.com/repos/openclaw/openclaw/issues/48516 | Android node notification forwarding causes cross-session replies (WhatsApp duplicate sends to wrong group) | ## Bug Description
When a WhatsApp message arrives, the gateway processes it correctly via the native Baileys WA provider. However, the Android companion app **also** receives the same message as a push notification and forwards it as a `notifications.changed` node event. This creates a duplicate processing path where... | open | null | false | 3 | [] | [] | 2026-03-16T22:05:16Z | 2026-05-04T14:04:38Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | dyoung522 | 422,900 | MDQ6VXNlcjQyMjkwMA== | User | false |
openclaw/openclaw | 4,085,022,336 | I_kwDOQb6kR87zfH6A | 48,517 | https://github.com/openclaw/openclaw/issues/48517 | https://api.github.com/repos/openclaw/openclaw/issues/48517 | [Feature]: Support plugin-requestable session isolation for trust boundaries |
### Problem
Plugins that process untrusted data alongside trusted context have no way to isolate the untrusted processing. A before_tool_call hook can block or allow, but it cannot say "run this in a sandbox and give me the result with metadata about what happened."
OpenClaw already has sub-agent infrastructure (s... | open | null | false | 2 | [] | [] | 2026-03-16T22:05:42Z | 2026-05-04T15:05:36Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,085,026,126 | I_kwDOQb6kR87zfI1O | 48,519 | https://github.com/openclaw/openclaw/issues/48519 | https://api.github.com/repos/openclaw/openclaw/issues/48519 | [Feature]: Track data provenance across session boundaries for persistent state |
### Problem
When an agent writes to persistent storage (memory, files, stored preferences, scheduled tasks), the provenance of the data being written is lost. A future session that reads this stored state treats it as trusted context, regardless of whether it originated from a verified user message, an inter-session... | open | null | false | 2 | [] | [] | 2026-03-16T22:06:40Z | 2026-04-30T21:58:21Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,085,133,860 | I_kwDOQb6kR87zfjIk | 48,532 | https://github.com/openclaw/openclaw/issues/48532 | https://api.github.com/repos/openclaw/openclaw/issues/48532 | Security by Intent: Progressive Permission Pattern Generalization to Eliminate Privilege Escalation and Approval Fatigue | ## Security by Intent: Progressive Permission Pattern Generalization
> **Keywords:** security, permissions, allowlist, access control, least privilege, privilege escalation prevention, approval fatigue, pattern inference, adaptive security, exec approval, policy mining
### The Problem: "Always Allow" Grants Far More ... | open | null | false | 4 | [] | [] | 2026-03-16T22:34:14Z | 2026-05-04T14:26:19Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | xinbenlv | 640,325 | MDQ6VXNlcjY0MDMyNQ== | User | false |
openclaw/openclaw | 4,085,154,694 | I_kwDOQb6kR87zfoOG | 48,534 | https://github.com/openclaw/openclaw/issues/48534 | https://api.github.com/repos/openclaw/openclaw/issues/48534 | [Bug]: Plugin before_agent_start hook hang permanently blocks all agent runs with zero diagnostic output | ### Bug type
Regression (worked before, now fails)
### Summary
If a plugin's before_agent_start hook awaits a network call that never resolves, the entire agent run pipeline blocks permanently with no timeout, no error logged, and no recovery path short of restarting the gateway.
### Steps to reproduce
1. Install ... | open | null | false | 5 | [
"bug",
"regression"
] | [] | 2026-03-16T22:38:38Z | 2026-05-04T14:26:21Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | VideoScape | 1,013,207 | MDQ6VXNlcjEwMTMyMDc= | User | false |
openclaw/openclaw | 4,085,434,875 | I_kwDOQb6kR87zgsn7 | 48,573 | https://github.com/openclaw/openclaw/issues/48573 | https://api.github.com/repos/openclaw/openclaw/issues/48573 | [Bug]: Embedded-run session state leak - zombie agents persist after parent agent termination | ## Summary
Embedded-run subagents spawned with the parameter may leave zombie state in the session store after the parent agent terminates, causing subsequent runs to encounter stale execution contexts.
## Bug type
Behavior bug (incorrect output/state without crash)
## Steps to Reproduce
1. Start an OpenClaw agen... | open | null | false | 10 | [] | [] | 2026-03-17T00:07:44Z | 2026-05-04T15:05:38Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,085,481,026 | I_kwDOQb6kR87zg35C | 48,579 | https://github.com/openclaw/openclaw/issues/48579 | https://api.github.com/repos/openclaw/openclaw/issues/48579 | Context pruning mode 'off' not preventing compactions (Anthropic API key, cache-ttl smart defaults) | ## Summary
Context pruning set to `mode: "off"` with `ttl: "999h"` and `keepLastAssistants: 999`, but compactions continue to fire aggressively (~82 compactions in a single day) at only 12-16% context usage (120-160k of 1M tokens).
## Environment
- OpenClaw version: 2026.2.17 (4134875)
- Model: `anthropic/claude-opu... | open | null | false | 4 | [] | [] | 2026-03-17T00:21:15Z | 2026-04-30T23:21:19Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | adeydas | 1,383,020 | MDQ6VXNlcjEzODMwMjA= | User | false |
openclaw/openclaw | 4,085,482,341 | I_kwDOQb6kR87zg4Nl | 48,580 | https://github.com/openclaw/openclaw/issues/48580 | https://api.github.com/repos/openclaw/openclaw/issues/48580 | Bug: acpx codex sessions 创建的会话立即退出 - stdin is not a terminal | ## Bug 描述
`acpx codex sessions new` 创建的 Codex 会话立即退出，进程 PID 存在但实际未运行。
## 复现步骤
1. 运行 `acpx codex sessions new --name test`
2. 检查会话状态：`.closed = false`，但进程已退出
## 错误信息
Codex CLI 需要 TTY，但 acpx 启动时未正确分配终端。
## 环境信息
- OpenClaw: v2026.3.13
- acpx: 0.1.15
- codex-cli: 0.114.0
- macOS: Darwin 24.6.0
## 日志
```
[acpx] s... | open | null | false | 2 | [] | [] | 2026-03-17T00:21:44Z | 2026-04-30T22:39:16Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | lonelybeanz | 34,127,680 | MDQ6VXNlcjM0MTI3Njgw | User | false |
openclaw/openclaw | 4,085,687,879 | I_kwDOQb6kR87zhqZH | 48,620 | https://github.com/openclaw/openclaw/issues/48620 | https://api.github.com/repos/openclaw/openclaw/issues/48620 | [Feature]: No way to view/manage paired contacts & permissions | ### Summary
As an admin, there's no command to list currently paired/connected WhatsApp numbers
and their associated permissions in one place.
The only way to manage permissions is by manually editing openclaw.json,
which is not intuitive for non-technical users or even for quick admin tasks.
Something intuitive ... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-17T01:32:24Z | 2026-05-04T15:05:46Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ch-arslanahmad | 218,025,646 | U_kgDODP7Org | User | false |
openclaw/openclaw | 4,085,700,509 | I_kwDOQb6kR87zhted | 48,623 | https://github.com/openclaw/openclaw/issues/48623 | https://api.github.com/repos/openclaw/openclaw/issues/48623 | Auth profile failover: lastGood not respected for ordering + shared timeout budget causes cascade failures | ## Problem
When the primary auth profile hits a rate limit (429), the failover system correctly decides `rotate_profile` to the backup. However, two bugs cause the backup to also fail:
### 1. `lastGood` not used for profile ordering
`auth-profiles.json` correctly tracks `lastGood: "anthropic:backup"` after the defau... | open | null | false | 5 | [] | [] | 2026-03-17T01:37:03Z | 2026-04-30T21:58:22Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,085,714,100 | I_kwDOQb6kR87zhwy0 | 48,628 | https://github.com/openclaw/openclaw/issues/48628 | https://api.github.com/repos/openclaw/openclaw/issues/48628 | Feature: Pass raw audio to multimodal models instead of (or alongside) STT transcription | # Feature: Pass raw audio to multimodal models instead of (or alongside) STT transcription
## Problem
Currently, when a user sends a voice/audio message via Telegram, Signal, etc., OpenClaw's pipeline always runs STT first and only passes the transcript text to the model. The original audio is discarded before reachi... | open | null | false | 1 | [] | [] | 2026-03-17T01:41:58Z | 2026-04-30T23:21:20Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | chenzefeng09 | 7,556,537 | MDQ6VXNlcjc1NTY1Mzc= | User | false |
openclaw/openclaw | 4,085,790,478 | I_kwDOQb6kR87ziDcO | 48,641 | https://github.com/openclaw/openclaw/issues/48641 | https://api.github.com/repos/openclaw/openclaw/issues/48641 | Discord DMs: inbound messages silently dropped (outbound works) | ## Bug Description
Discord inbound DMs from an allowlisted user are silently dropped by the gateway. Outbound DMs (bot → user) work correctly. Guild channel messages work correctly.
## Environment
- **OpenClaw version:** 2026.3.13 (61d171a)
- **OS:** macOS 26.3.1 (arm64)
- **Node:** v25.8.0
- **Discord bot count:** ... | open | null | false | 3 | [] | [] | 2026-03-17T02:10:04Z | 2026-05-04T15:05:50Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ngmeyer | 3,868,701 | MDQ6VXNlcjM4Njg3MDE= | User | false |
openclaw/openclaw | 4,085,913,378 | I_kwDOQb6kR87zihci | 48,680 | https://github.com/openclaw/openclaw/issues/48680 | https://api.github.com/repos/openclaw/openclaw/issues/48680 | [Bug] Model fallback treats HTTP 403 business rejection as 'candidate_succeeded', skipping remaining fallback models | ## Bug Description
When a model provider returns a business-level rejection (HTTP 200 with error payload, or non-timeout error), the fallback mechanism incorrectly marks the candidate as `candidate_succeeded` instead of `candidate_failed`, preventing the remaining fallback models from being tried.
## Steps to Reprodu... | open | null | false | 4 | [] | [] | 2026-03-17T02:58:56Z | 2026-05-04T14:54:49Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | lovensky1992-wk | 264,908,128 | U_kgDOD8otYA | User | false |
openclaw/openclaw | 4,086,035,337 | I_kwDOQb6kR87zi_OJ | 48,708 | https://github.com/openclaw/openclaw/issues/48708 | https://api.github.com/repos/openclaw/openclaw/issues/48708 | Feature: TUI chat message timestamps | ## Summary
Add an option to display timestamps on each message (user and assistant) in the TUI chat interface.
## Problem
The TUI currently shows messages without any time indication. When reviewing a conversation, there's no way to tell when a question was asked or when the response arrived. This makes it hard to:
... | open | null | false | 1 | [] | [] | 2026-03-17T03:47:59Z | 2026-05-01T00:41:50Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jimgifw-wang | 18,047,894 | MDQ6VXNlcjE4MDQ3ODk0 | User | false |
openclaw/openclaw | 4,086,054,217 | I_kwDOQb6kR87zjD1J | 48,711 | https://github.com/openclaw/openclaw/issues/48711 | https://api.github.com/repos/openclaw/openclaw/issues/48711 | [Bug]: Task and Memory Recall Is Often Too Weak to Trust Across Real Conversations | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw can remember too little in realistic task and memory workflows, especially after fresh sessions, state mutations, and broad natural-language recall prompts. In practice this means users cannot reliably trust it to:
[openclaw-memor... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-17T03:56:24Z | 2026-05-04T14:55:12Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | WarforgeTech | 221,437,236 | U_kgDODTLdNA | User | false |
openclaw/openclaw | 4,086,535,648 | I_kwDOQb6kR87zk5Xg | 48,788 | https://github.com/openclaw/openclaw/issues/48788 | https://api.github.com/repos/openclaw/openclaw/issues/48788 | feat: centralized filename encoding utility for multi-encoding Content-Disposition handling | ## Context
PR #48578 fixes the most common case (UTF-8 misread as Latin-1) for Feishu Chinese filenames. However, as noted by @Artyomkun, a proper architectural solution should handle multiple encodings (Shift-JIS, EUC-KR, GB18030, etc.) across all channel adapters.
## Proposal
Create a central filename utility that... | open | null | false | 17 | [] | [] | 2026-03-17T06:37:04Z | 2026-05-04T14:55:24Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | alex-xuweilong | 267,811,734 | U_kgDOD_Z7lg | User | false |
openclaw/openclaw | 4,086,661,328 | I_kwDOQb6kR87zlYDQ | 48,810 | https://github.com/openclaw/openclaw/issues/48810 | https://api.github.com/repos/openclaw/openclaw/issues/48810 | [Bug]: Compaction retry creates orphan fork in parentId chain (dead-end branch breaks chain reconstruction) | ### Bug type
Regression (worked before, now fails)
### Summary
---
Problem / 问题描述
When analyzing user session event chains via the parentId linked list in JSONL storage, we discovered that
compaction retry can produce a fork under the same parentId — one compaction event becomes a dead-end (no
descendants),... | open | null | false | 5 | [
"bug",
"regression"
] | [] | 2026-03-17T07:11:02Z | 2026-04-30T21:58:23Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | taohaowei | 23,160,436 | MDQ6VXNlcjIzMTYwNDM2 | User | false |
openclaw/openclaw | 4,087,107,124 | I_kwDOQb6kR87znE40 | 48,874 | https://github.com/openclaw/openclaw/issues/48874 | https://api.github.com/repos/openclaw/openclaw/issues/48874 | [RFC] Multi-Session Architecture: Shared LLM + Isolated Sessions + Public Knowledge Base | # [RFC] Multi-Session Architecture: Shared LLM + Isolated Sessions + Public Knowledge Base
## Summary
Proposing a new Session management architecture:
- **Shared LLM Layer**: All sessions share the same model inference instance, saving resources
- **Isolated Session Layer**: Each channel/task has independent context ... | open | null | false | 5 | [] | [] | 2026-03-17T08:55:55Z | 2026-04-30T21:58:25Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | KilimiaoSix | 39,623,866 | MDQ6VXNlcjM5NjIzODY2 | User | false |
openclaw/openclaw | 4,087,308,092 | I_kwDOQb6kR87zn188 | 48,897 | https://github.com/openclaw/openclaw/issues/48897 | https://api.github.com/repos/openclaw/openclaw/issues/48897 | [Bug] TUI input breaks on German keyboard layout: duplicated printable characters and broken AltGr (@, €) | ## Summary
In the TUI / gateway-client, keyboard input appears to break for a German keyboard layout:
- some characters are duplicated when typed
- AltGr-based printable characters are not entered correctly
This looks TUI/input-layer specific and does **not** reproduce outside the TUI.
## Observed behavior
Examples ... | open | null | false | 2 | [] | [] | 2026-03-17T09:33:59Z | 2026-05-04T14:55:53Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | jemmel | 153,074 | MDQ6VXNlcjE1MzA3NA== | User | false |
openclaw/openclaw | 4,087,512,392 | I_kwDOQb6kR87zon1I | 48,915 | https://github.com/openclaw/openclaw/issues/48915 | https://api.github.com/repos/openclaw/openclaw/issues/48915 | [Bug]: isRawToolCallBlock misses snake_case tool call types (tool_use, tool_call), causing corrupted session history errors with Anthropic | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The function isRawToolCallBlock in the session history repair logic only recognizes camelCase tool call block types (toolUse, toolCall, functionCall), but does NOT handle the snake_case variants used by Anthropic and some other providers (t... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-17T10:10:16Z | 2026-05-04T15:06:14Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | shrekcg | 19,178,584 | MDQ6VXNlcjE5MTc4NTg0 | User | false |
openclaw/openclaw | 4,087,531,241 | I_kwDOQb6kR87zosbp | 48,918 | https://github.com/openclaw/openclaw/issues/48918 | https://api.github.com/repos/openclaw/openclaw/issues/48918 | Feature Request: User-Level Skill Preferences/Conventions Support | # Feature Request: User-Level Skill Preferences/Conventions Support
## Problem Statement
Currently, OpenClaw lacks a standardized mechanism for users to define **personal preferences or conventions** for specific skills without:
1. **Overriding the entire SKILL.md** (which requires copying system skill content)
2. *... | open | null | false | 3 | [] | [] | 2026-03-17T10:13:52Z | 2026-04-30T21:58:26Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | davidfuzju | 3,890,843 | MDQ6VXNlcjM4OTA4NDM= | User | false |
openclaw/openclaw | 4,087,808,007 | I_kwDOQb6kR87zpwAH | 48,947 | https://github.com/openclaw/openclaw/issues/48947 | https://api.github.com/repos/openclaw/openclaw/issues/48947 | [Bug]: Openclaw is stupid again: 200000 tool calls, no feedback, interruptions | ### Bug type
Regression (worked before, now fails)
### Summary
Jesus.... you can't do a release with breaking anything eh?
After [#35077 #35077](https://github.com/openclaw/openclaw/issues/35077)
we're at it again......
<img width="1029" height="651" alt="Image" src="https://github.com/user-attachments/assets/e007... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-03-17T11:04:03Z | 2026-05-01T01:22:34Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | paciox | 6,546,735 | MDQ6VXNlcjY1NDY3MzU= | User | false |
openclaw/openclaw | 4,087,841,473 | I_kwDOQb6kR87zp4LB | 48,949 | https://github.com/openclaw/openclaw/issues/48949 | https://api.github.com/repos/openclaw/openclaw/issues/48949 | 🐛 Feishu channel fails with tenant_access_token error when HTTP proxy is configured | # 🐛 Bug: Feishu channel fails with `tenant_access_token` error when HTTP proxy is configured
## Summary
Feishu (飞书) channel cannot send messages when system HTTP proxy is configured. Error: `Cannot destructure property 'tenant_access_token' of '(intermediate value)' as it is undefined.`
**Root cause:** The `@larksu... | open | null | false | 4 | [] | [] | 2026-03-17T11:09:31Z | 2026-05-04T15:06:16Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | AndrzejXIAO | 144,600,026 | U_kgDOCJ5r2g | User | false |
openclaw/openclaw | 4,088,166,182 | I_kwDOQb6kR87zrHcm | 48,979 | https://github.com/openclaw/openclaw/issues/48979 | https://api.github.com/repos/openclaw/openclaw/issues/48979 | Telegram channel cannot send images via read tool | ## Bug Description
When using the `read` tool to read an image file and trying to send it through Telegram channel, the image is not delivered to the user. The read tool successfully reads the image (returns image metadata), but the image attachment is not included in the Telegram message.
## Workaround
Directly cal... | open | null | false | 4 | [] | [] | 2026-03-17T12:13:57Z | 2026-05-04T14:56:25Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | claw-no-308 | 268,619,436 | U_kgDOEALOrA | User | false |
openclaw/openclaw | 4,088,391,709 | I_kwDOQb6kR87zr-gd | 49,010 | https://github.com/openclaw/openclaw/issues/49010 | https://api.github.com/repos/openclaw/openclaw/issues/49010 | Add opt-in message-level signing for mcporter stdio transport | cc @leondz
OpenClaw's mcporter provides a clean stdio transport for MCP servers. Currently, messages pass through unsigned -- which is fine for local development but becomes a risk in production deployments where:
- Multiple agents share MCP server pools
- Tool calls cross trust boundaries
- Replay attacks can trigge... | open | null | false | 2 | [] | [] | 2026-03-17T12:57:53Z | 2026-05-04T15:06:21Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | razashariff | 36,504,271 | MDQ6VXNlcjM2NTA0Mjcx | User | false |
openclaw/openclaw | 4,088,411,746 | I_kwDOQb6kR87zsDZi | 49,012 | https://github.com/openclaw/openclaw/issues/49012 | https://api.github.com/repos/openclaw/openclaw/issues/49012 | [Bug]: The OpenClaw gateway is "frozen" and experiencing "port conflicts," preventing web pages from loading. | ### Bug type
Regression (worked before, now fails)
### Summary
Win11 vm ubuntu 24.04.1; node -v v22.22.1 openclaw -v OpenClaw 2026.3.13 (61d171a) After completely uninstalling and then reinstalling OpenClaw, the issue still persists. 🦞 OpenClaw 2026.3.13 (61d171a) — If you're lost, run doctor; if you're b... | open | null | false | 2 | [
"bug",
"regression"
] | [
"vincentkoc"
] | 2026-03-17T13:01:56Z | 2026-05-01T23:59:41Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | namekong8-gmail | 228,416,831 | U_kgDODZ1dPw | User | false |
openclaw/openclaw | 4,088,455,224 | I_kwDOQb6kR87zsOA4 | 49,019 | https://github.com/openclaw/openclaw/issues/49019 | https://api.github.com/repos/openclaw/openclaw/issues/49019 | [Feature]: Add an optional result-return mode for /hooks/agent | ### Summary
Add an opt-in, machine-friendly result-return mode for `POST /hooks/agent`, so external automations can submit work and receive a deterministic result without relying on `sessions_history` polling or reopening `sessions_send` over HTTP.
### Problem to solve
OpenClaw currently has two nearby primitives, b... | open | null | false | 2 | [] | [] | 2026-03-17T13:09:48Z | 2026-05-04T20:22:29Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ronliu014 | 6,783,435 | MDQ6VXNlcjY3ODM0MzU= | User | false |
openclaw/openclaw | 4,088,632,151 | I_kwDOQb6kR87zs5NX | 49,036 | https://github.com/openclaw/openclaw/issues/49036 | https://api.github.com/repos/openclaw/openclaw/issues/49036 | openclaw doctor should detect stale gateway lock files | ## Summary
`openclaw doctor` validates session lock files (checking PID liveness and staleness) but does not check the gateway lock file at `$TMPDIR/openclaw-$UID/gateway.*.lock`.
## Problem
When the gateway lock file references a dead PID, `doctor` reports no issues, but the gateway cannot start. The user has no indi... | open | null | false | 2 | [] | [] | 2026-03-17T13:38:13Z | 2026-04-30T22:39:19Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | brandontomlin2 | 100,727,150 | U_kgDOBgD5bg | User | false |
openclaw/openclaw | 4,088,789,401 | I_kwDOQb6kR87ztfmZ | 49,046 | https://github.com/openclaw/openclaw/issues/49046 | https://api.github.com/repos/openclaw/openclaw/issues/49046 | What's the usage log format for model.usage events? | ### Summary
Looking to read OpenClaw's local inference logs to track usage alongside paid AI subscription costs — just need the field names.
### Problem to solve
I'm building Subarb (v0-subarb.vercel.app), a tool that tracks AI spend across paid platforms (Claude, ChatGPT, Cursor, etc.). I want to include OpenClaw l... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-17T14:04:57Z | 2026-05-04T14:56:37Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | silesc01 | 36,805,553 | MDQ6VXNlcjM2ODA1NTUz | User | false |
openclaw/openclaw | 4,088,872,968 | I_kwDOQb6kR87zt0AI | 49,055 | https://github.com/openclaw/openclaw/issues/49055 | https://api.github.com/repos/openclaw/openclaw/issues/49055 | Silent delivery drop after overloaded_error recovery | ## Description
When an Anthropic `overloaded_error` occurs on the first LLM attempt but the session recovers on retry and produces a valid response (`stop=stop`), the Discord message delivery silently fails. The response is recorded in the session transcript but never sent to Discord.
## Reproduction
1. A thread-bou... | open | null | false | 5 | [] | [] | 2026-03-17T14:17:59Z | 2026-04-30T21:58:27Z | null | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | sercada | 24,389,792 | MDQ6VXNlcjI0Mzg5Nzky | User | false |
openclaw/openclaw | 4,089,079,052 | I_kwDOQb6kR87zumUM | 49,077 | https://github.com/openclaw/openclaw/issues/49077 | https://api.github.com/repos/openclaw/openclaw/issues/49077 | Feature Request: Add configurable delay between API requests | ### Summary
I suggest adding a setting to configure the delay between consecutive API requests.
### Problem to solve
Since some API providers enforce strict rate limits, sending requests too rapidly triggers denial of service, significantly degrading the user experience.
### Proposed solution
Add a new configurat... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-17T14:50:34Z | 2026-04-30T21:58:29Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | 2203476 | 30,993,560 | MDQ6VXNlcjMwOTkzNTYw | User | false |
openclaw/openclaw | 4,089,082,119 | I_kwDOQb6kR87zunEH | 49,078 | https://github.com/openclaw/openclaw/issues/49078 | https://api.github.com/repos/openclaw/openclaw/issues/49078 | [Feature]: Cron job output routing — hard-coded destination for job results (file, channel, session) | ### Summary
Allow cron jobs to declare a deterministic output target (file path, Telegram chat, session) in the job config, so results are guaranteed to land somewhere regardless of whether the agent prompt correctly instructs output routing.
### Problem to solve
Currently, cron job output routing relies entirely on... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-17T14:51:02Z | 2026-04-30T21:58:30Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | agenthyjack | 268,943,245 | U_kgDOEAe_jQ | User | false |
openclaw/openclaw | 4,089,165,603 | I_kwDOQb6kR87zu7cj | 49,086 | https://github.com/openclaw/openclaw/issues/49086 | https://api.github.com/repos/openclaw/openclaw/issues/49086 | [Bug/Template]: LLM hallucinates Unix timestamps when writing heartbeat-state.json — causes skipped or over-triggered checks | ## Description
The AGENTS.md template recommends tracking heartbeat checks in `memory/heartbeat-state.json` with Unix timestamps. However, LLMs frequently **hallucinate incorrect epoch values** when writing these timestamps, causing heartbeat checks to be skipped entirely or over-triggered.
## Root Cause
When an LLM... | open | null | false | 2 | [] | [] | 2026-03-17T15:02:39Z | 2026-05-04T14:56:56Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | lunanova-core | 258,436,214 | U_kgDOD2dsdg | User | false |
openclaw/openclaw | 4,089,928,716 | I_kwDOQb6kR87zx1wM | 49,157 | https://github.com/openclaw/openclaw/issues/49157 | https://api.github.com/repos/openclaw/openclaw/issues/49157 | Session write locks leak when Gateway encounters unhandled promise rejections / streams, causing >30min deadlocks | ## Description
There is a fundamental flaw in how cquireSessionWriteLock and inspectLockPayload handle orphaned locks held by a continuously running process (e.g., the Gateway daemon).
## Steps to Reproduce (Conceptual)
1. A long-running Gateway process (PID A) acquires a session write lock (*.jsonl.lock) to append... | open | null | false | 7 | [] | [] | 2026-03-17T17:09:04Z | 2026-05-02T00:38:15Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | zhangxue1985122219 | 54,266,548 | MDQ6VXNlcjU0MjY2NTQ4 | User | false |
openclaw/openclaw | 4,090,041,163 | I_kwDOQb6kR87zyRNL | 49,162 | https://github.com/openclaw/openclaw/issues/49162 | https://api.github.com/repos/openclaw/openclaw/issues/49162 | Schema-backed memory search reliability controls: timeout, query cache TTL, and retry policy | # Draft GitHub Issue / Enhancement Proposal
## Title
Schema-backed memory search reliability controls: timeout, query cache TTL, and retry policy
## Summary
OpenClaw's memory search runtime currently benefits from timeout, retry, and query-result caching behavior, but these controls are not exposed as schema-supporte... | open | null | false | 2 | [] | [] | 2026-03-17T17:27:33Z | 2026-04-30T21:58:32Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | u4ma-kev | 45,077,960 | MDQ6VXNlcjQ1MDc3OTYw | User | false |
openclaw/openclaw | 4,090,186,153 | I_kwDOQb6kR87zy0mp | 49,175 | https://github.com/openclaw/openclaw/issues/49175 | https://api.github.com/repos/openclaw/openclaw/issues/49175 | cron.runs with scope=all leaks job names across sessions | ## Summary
Follow-up from #35447 (per-session isolation for cron operations). While implementing session-scoped filtering for `cron.list`, `cron.remove`, `cron.update`, and `cron.run`, we identified that `cron.runs` with `scope === "all"` still leaks job names from other agents/sessions.
## Current Behavior
In `src/... | open | null | false | 2 | [] | [] | 2026-03-17T17:52:59Z | 2026-04-30T23:21:23Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | antonio-mello-ai | 186,876,424 | U_kgDOCyOCCA | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.