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 | 4,313,012,820 | I_kwDOQb6kR88AAAABARNaVA | 70,438 | https://github.com/openclaw/openclaw/issues/70438 | https://api.github.com/repos/openclaw/openclaw/issues/70438 | [Feature]: Make non-image webchat attachment handling configurable (PDF offload, text inline) | ## Ask
Add a `webchat.attachments` config (or per-tenant knob) controlling how non-image uploads are handled, rather than silently dropping them in `attachment-normalize`.
**Proposed defaults (configurable):**
- `image/*` → existing inline-or-offload path (unchanged)
- `application/pdf` → offload and inject `[media a... | closed | completed | true | 1 | [] | [] | 2026-04-23T02:24:55Z | 2026-04-29T04:34:01Z | 2026-04-26T21:31:38Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | faux123 | 598,478 | MDQ6VXNlcjU5ODQ3OA== | User | false |
openclaw/openclaw | 4,313,013,851 | I_kwDOQb6kR88AAAABARNeWw | 70,439 | https://github.com/openclaw/openclaw/issues/70439 | https://api.github.com/repos/openclaw/openclaw/issues/70439 | [Feature]: Signal channel — pluggable audio preflight (whisper auto-transcribe for voice notes) | ## Ask
Expose a pluggable audio-preflight hook for the Signal channel (and ideally a channel-agnostic base) so voice notes can be auto-transcribed before they reach the agent — parity with Telegram and Discord.
**Two flavors would work:**
1. **Built-in whisper integration**, mirroring what Telegram/Discord ship — a ... | open | null | false | 2 | [] | [] | 2026-04-23T02:25:10Z | 2026-05-02T12:54:08Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | faux123 | 598,478 | MDQ6VXNlcjU5ODQ3OA== | User | false |
openclaw/openclaw | 4,313,014,479 | I_kwDOQb6kR88AAAABARNgzw | 70,440 | https://github.com/openclaw/openclaw/issues/70440 | https://api.github.com/repos/openclaw/openclaw/issues/70440 | [Feature]: Config option to silently swallow empty-turn responses instead of delivering error text | ## Ask
Add a config option (global or per-channel) to silently swallow empty-turn responses server-side instead of delivering `"Agent couldn't generate a response"` (or similar) text to the channel.
**Proposed:**
```yaml
incomplete_turns:
deliver: silent # error | silent
```
- `error` (current default) — existin... | closed | completed | true | 2 | [] | [] | 2026-04-23T02:25:20Z | 2026-05-03T04:35:23Z | 2026-04-30T19:18:23Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | faux123 | 598,478 | MDQ6VXNlcjU5ODQ3OA== | User | false |
openclaw/openclaw | 4,313,022,905 | I_kwDOQb6kR88AAAABAROBuQ | 70,443 | https://github.com/openclaw/openclaw/issues/70443 | https://api.github.com/repos/openclaw/openclaw/issues/70443 | [Feature]: Add --offline flag to scripts/docker/setup.sh for airgapped re-runs with a pre-loaded image | ### Summary
Add a `--offline` flag to `scripts/docker/setup.sh` so operators on airgapped hosts can re-run the full setup routine after transferring the OpenClaw image out-of-band (e.g. via USB or an internal artifact bundle) and loading it with `docker load`.
### Problem to solve
`scripts/docker/setup.sh` unconditi... | open | null | false | 2 | [] | [] | 2026-04-23T02:27:28Z | 2026-05-02T17:31:46Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | safrano9999 | 240,768,512 | U_kgDODlnWAA | User | false |
openclaw/openclaw | 4,313,032,398 | I_kwDOQb6kR88AAAABAROmzg | 70,445 | https://github.com/openclaw/openclaw/issues/70445 | https://api.github.com/repos/openclaw/openclaw/issues/70445 | [Feishu] TTS voice bubble: MiniMax provider does not support voice-compatible output | ## Problem
When using the MiniMax TTS provider with Feishu as the target channel, voice messages are received as `.mp3` file attachments instead of native Feishu voice bubbles.
## Expected behavior
Feishu should receive the TTS audio as a playable voice bubble (`msg_type: "audio"`), not as a generic file attachment.... | closed | completed | true | 1 | [] | [] | 2026-04-23T02:29:53Z | 2026-04-27T04:34:54Z | 2026-04-25T03:53:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Lrs-GreenBottle | 253,340,193 | U_kgDODxmqIQ | User | false |
openclaw/openclaw | 4,313,081,796 | I_kwDOQb6kR88AAAABARRnxA | 70,447 | https://github.com/openclaw/openclaw/issues/70447 | https://api.github.com/repos/openclaw/openclaw/issues/70447 | [Bug]: Discord slash commands in thread hang on “thinking…” and crash with partial channel error | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Version: OpenClaw 2026.4.21
Channel: Discord
Scope: Slash commands (e.g. /models, /focus) when executed inside a Discord thread
**Summary**
After updating to 2026.4.21, when slash commands are run inside a Discord thread, t... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-23T02:40:38Z | 2026-04-26T04:30:14Z | 2026-04-23T16:37:02Z | NONE | null | 20260426T060034Z | 2026-04-26T06:00:34Z | Kalclavia | 13,138,605 | MDQ6VXNlcjEzMTM4NjA1 | User | false |
openclaw/openclaw | 4,079,241,015 | I_kwDOQb6kR87zJEc3 | 47,639 | https://github.com/openclaw/openclaw/issues/47639 | https://api.github.com/repos/openclaw/openclaw/issues/47639 | Bug: Token usage not recorded in session files, dashboard shows n/a (vLLM provider) | ### Bug type
Regression (worked before, now fails)
### Summary
**Bug Report: Token usage not being recorded in session files and dashboard shows "n/a"**
**Description**
Token usage statistics from LLM provider (vLLM) are not being saved to session files, causing the OpenClaw dashboard to display "n/a" in the Tokens... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-15T21:32:59Z | 2026-04-25T04:15:52Z | 2026-04-23T02:57:05Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | pi-null-mezon | 5,991,537 | MDQ6VXNlcjU5OTE1Mzc= | User | false |
openclaw/openclaw | 4,313,206,296 | I_kwDOQb6kR88AAAABARZOGA | 70,451 | https://github.com/openclaw/openclaw/issues/70451 | https://api.github.com/repos/openclaw/openclaw/issues/70451 | [Bug]: CLI hooks enable times out / SIGKILL on Windows | ## Bug Description
On Windows (OpenClaw 2026.4.21), running `openclaw hooks enable self-improvement` hangs and is killed by SIGKILL after ~60 seconds.
All CLI commands except `--help` and `--version` hang for 60+ seconds. This is consistent with issue #64306 (CLI commands hang in ensureCliPluginRegistryLoaded).
## E... | open | null | false | 1 | [] | [] | 2026-04-23T03:07:33Z | 2026-05-02T23:12:40Z | null | NONE | null | 20260503T000641Z | 2026-05-03T00:06:41Z | huydaobk | 253,844,076 | U_kgDODyFabA | User | false |
openclaw/openclaw | 4,138,302,462 | I_kwDOQb6kR872qXv- | 54,690 | https://github.com/openclaw/openclaw/issues/54690 | https://api.github.com/repos/openclaw/openclaw/issues/54690 | [Bug]: ACP sessions_spawn with streamTo:"parent" never delivers completion event to parent session | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Summary:
When spawning a Codex ACP session via sessions_spawn({ runtime: "acp", agentId: "codex", streamTo: "parent" }), the parent session receives the started event and no output for 60s stall warnings, but never receives the completion e... | closed | completed | true | 6 | [
"bug"
] | [
"vincentkoc"
] | 2026-03-25T20:27:14Z | 2026-04-28T04:46:01Z | 2026-04-25T21:11:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | andreas-oendig | 211,547,131 | U_kgDODJvz-w | User | false |
openclaw/openclaw | 4,313,226,664 | I_kwDOQb6kR88AAAABARadqA | 70,453 | https://github.com/openclaw/openclaw/issues/70453 | https://api.github.com/repos/openclaw/openclaw/issues/70453 | [Bug]: CLI and WebUI silent failure/hang when calling Gemini via HTTP proxy (v2026.4.21) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Setting http_proxy environment variables to bypass a local Fake IP SSRF block causes OpenClaw's CLI and WebUI to silently hang and fail without returning any errors when querying Gemini models.
### Steps to reproduce
1. Se... | closed | completed | false | 3 | [
"bug",
"regression"
] | [] | 2026-04-23T03:12:14Z | 2026-05-03T20:48:52Z | 2026-05-03T20:48:52Z | NONE | null | 20260503T210034Z | 2026-05-03T21:00:34Z | FanXiangTM | 48,204,825 | MDQ6VXNlcjQ4MjA0ODI1 | User | false |
openclaw/openclaw | 4,313,257,200 | I_kwDOQb6kR88AAAABARcU8A | 70,454 | https://github.com/openclaw/openclaw/issues/70454 | https://api.github.com/repos/openclaw/openclaw/issues/70454 | [Bug]: when installing v2026.4.21, on mac and liunx it keep on giving "Error: Cannot find module" when using openai-codex oauth | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
When installing v2026.4.21, on mac and liunx, installation is exiting and keep on giving "Error: Cannot find module *" when using openai-codex oauth, tried multiple time and with different model but same issue
### Steps to rep... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-23T03:20:25Z | 2026-04-26T04:30:08Z | 2026-04-23T16:47:15Z | NONE | null | 20260426T060034Z | 2026-04-26T06:00:34Z | patilak1234 | 5,363,239 | MDQ6VXNlcjUzNjMyMzk= | User | false |
openclaw/openclaw | 4,313,264,222 | I_kwDOQb6kR88AAAABARcwXg | 70,455 | https://github.com/openclaw/openclaw/issues/70455 | https://api.github.com/repos/openclaw/openclaw/issues/70455 | Vision image lag: model describes previous images instead of current one | ## Description
When using vision-capable models (e.g. ), images sent by users are consistently described with a 1-2 message lag — the model describes the **previous** image instead of the current one.
## Steps to Reproduce
1. Configure a vision-capable model (e.g. ) with `input: ["text", "image"]` in the provider ca... | closed | completed | true | 1 | [] | [] | 2026-04-23T03:22:39Z | 2026-04-29T04:42:05Z | 2026-04-26T07:14:07Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | samurai-bot | 52,017,349 | MDQ6VXNlcjUyMDE3MzQ5 | User | false |
openclaw/openclaw | 4,313,273,590 | I_kwDOQb6kR88AAAABARdU9g | 70,458 | https://github.com/openclaw/openclaw/issues/70458 | https://api.github.com/repos/openclaw/openclaw/issues/70458 | Control UI exposes internal-only async exec completion system messages to users | ### Summary
Control UI currently exposes async exec completion system messages on the user surface, even when the event is explicitly marked for internal handling only.
This creates noisy and confusing UX:
- the user sees a raw system event like `Exec completed (...)`
- the agent is then instructed to handle the resu... | closed | completed | true | 1 | [] | [] | 2026-04-23T03:25:39Z | 2026-04-28T04:47:53Z | 2026-04-25T19:23:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wrnmjb2023-art | 262,927,172 | U_kgDOD6vzRA | User | false |
openclaw/openclaw | 4,313,283,928 | I_kwDOQb6kR88AAAABARd9WA | 70,460 | https://github.com/openclaw/openclaw/issues/70460 | https://api.github.com/repos/openclaw/openclaw/issues/70460 | [Performance] memory_search cold start ~5-7s — needs module pre-warming | ## Problem Description
memory_search tool has a significant cold start latency (~5-7 seconds) on first invocation after gateway starts. Subsequent calls are fast (~50ms).
**Root Cause Analysis:**
Each memory_search call goes through a dynamic import chain:
- createMemorySearchTool
- loadMemoryToolRuntime() [has m... | closed | not_planned | true | 1 | [] | [] | 2026-04-23T03:28:43Z | 2026-04-29T04:37:00Z | 2026-04-26T10:51:11Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | bthaha123-Carl | 161,189,858 | U_kgDOCZuP4g | User | false |
openclaw/openclaw | 4,313,295,554 | I_kwDOQb6kR88AAAABAReqwg | 70,463 | https://github.com/openclaw/openclaw/issues/70463 | https://api.github.com/repos/openclaw/openclaw/issues/70463 | WhatsApp: health-monitor race condition causes multi-hour outages | # OpenClaw WhatsApp Issue: Health monitor race condition with Baileys 30-minute disconnections
## Problem Summary
WhatsApp Web gateway experiences a race condition that causes ~7-hour outages:
1. **Normal operation:** WhatsApp server disconnects every ~30 minutes (status 499)
2. **Baileys retry mechanism:** Automati... | closed | not_planned | true | 1 | [] | [] | 2026-04-23T03:32:34Z | 2026-04-29T04:36:59Z | 2026-04-26T10:51:15Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | YYChien | 86,012,800 | MDQ6VXNlcjg2MDEyODAw | User | false |
openclaw/openclaw | 4,215,027,231 | I_kwDOQb6kR877PDYf | 62,243 | https://github.com/openclaw/openclaw/issues/62243 | https://api.github.com/repos/openclaw/openclaw/issues/62243 | [Bug] Update to 2026.4.5 fails: Cannot find module '@buape/carbon' | null | closed | duplicate | true | 7 | [] | [] | 2026-04-07T02:14:37Z | 2026-04-27T04:40:25Z | 2026-04-25T02:27:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rose131-max | 208,792,308 | U_kgDODHHq9A | User | false |
openclaw/openclaw | 4,313,374,938 | I_kwDOQb6kR88AAAABARjg2g | 70,467 | https://github.com/openclaw/openclaw/issues/70467 | https://api.github.com/repos/openclaw/openclaw/issues/70467 | [Enhancement] OpenClaw News Collector: Add more technical blog RSS sources | ## Overview
The OpenClaw News Collector (`collect_all.py`) currently has limited RSS sources in `extended_sources.json`. This issue tracks adding more high-quality technical blog sources.
## Current Sources (summary)
- Priority 1-7 across: HN RSS, TechCrunch, The Verge, Ars Technica, Reddit, Zenn, Qiita, dev.to, Mediu... | closed | not_planned | true | 2 | [] | [] | 2026-04-23T03:55:47Z | 2026-04-28T04:47:53Z | 2026-04-25T19:23:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Keis828 | 9,313,642 | MDQ6VXNlcjkzMTM2NDI= | User | false |
openclaw/openclaw | 4,313,267,241 | I_kwDOQb6kR88AAAABARc8KQ | 70,456 | https://github.com/openclaw/openclaw/issues/70456 | https://api.github.com/repos/openclaw/openclaw/issues/70456 | [Bug] Agent tool-use regressed for all providers between v2026.4.20 and main — 0/6 models invoke tools on a cold-start prompt; multiple model families converge on fabricated "Exec is blocked on this session" phrase | # [Bug] Agent tool-use regressed for all providers between v2026.4.20 and main — 0/6 models invoke tools on a cold-start prompt; multiple model families converge on fabricated "Exec is blocked on this session" phrase
## Summary
Between `v2026.4.20` (working) and current `origin/main` (`1c91f3f`, broken), OpenClaw's a... | closed | completed | true | 6 | [] | [] | 2026-04-23T03:23:43Z | 2026-04-25T04:15:46Z | 2026-04-23T04:06:51Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | perlowja | 162,375,972 | U_kgDOCa2pJA | User | false |
openclaw/openclaw | 4,313,403,577 | I_kwDOQb6kR88AAAABARlQuQ | 70,470 | https://github.com/openclaw/openclaw/issues/70470 | https://api.github.com/repos/openclaw/openclaw/issues/70470 | [Bug]: WhatsApp behavior seems to depend on read receipts being enabled | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
n my tests, OpenClaw behaved inconsistently depending on whether the WhatsApp account had read receipts enabled.
### Steps to reproduce
Set up OpenClaw with WhatsApp linked and working
Configure a working Goo... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-23T04:03:22Z | 2026-05-01T04:40:57Z | 2026-04-29T01:32:17Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | hadnet | 13,828,833 | MDQ6VXNlcjEzODI4ODMz | User | false |
openclaw/openclaw | 4,038,712,786 | I_kwDOQb6kR87wud3S | 38,961 | https://github.com/openclaw/openclaw/issues/38961 | https://api.github.com/repos/openclaw/openclaw/issues/38961 | [Bug]: [WebChat] 回复重复发送两次 (delivery-mirror + original) | ### Bug type
Regression (worked before, now fails)
### Summary
通过 WebChat (Control UI) 发送消息时，AI 的回复会重复显示两条，内容完全相同。
### Steps to reproduce
1. 通过 WebChat (Control UI) 发送任意消息
2. 收到两条相同的 AI 回复
### Expected behavior
从会话历史可以看到，同一条回复被记录了两次：
1. `provider: "minimax-coding"`, `model: "MiniMax-M2.5"` - 原始回复
2. `provider: "... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T15:26:17Z | 2026-04-27T05:00:41Z | 2026-04-24T20:13:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Lyndoncthia | 38,835,833 | MDQ6VXNlcjM4ODM1ODMz | User | false |
openclaw/openclaw | 4,038,726,299 | I_kwDOQb6kR87wuhKb | 38,963 | https://github.com/openclaw/openclaw/issues/38963 | https://api.github.com/repos/openclaw/openclaw/issues/38963 | Feature Request: Display Resolved Sender Name in Dashboard Session List | # Feature Request: Display Resolved Sender Name in Dashboard Session List
## Problem
In the OpenClaw Gateway Dashboard's session list, conversations are currently identified by raw peer IDs (e.g., `ou_9ca4d34a47794278dd1d906d7bfdac66` for Feishu). This makes it difficult to quickly identify which user or conversation... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T15:33:34Z | 2026-04-27T05:00:40Z | 2026-04-24T20:13:12Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Robinzhh | 260,084,079 | U_kgDOD4CRbw | User | false |
openclaw/openclaw | 4,038,732,554 | I_kwDOQb6kR87wuisK | 38,966 | https://github.com/openclaw/openclaw/issues/38966 | https://api.github.com/repos/openclaw/openclaw/issues/38966 | TUI does not refresh when session is reset from another client (e.g. Telegram /new) | ## Summary
When a session is reset via `/new` or `/reset` from a non-TUI client (e.g. Telegram, Discord, WhatsApp), the TUI does not reflect the change. The chat log continues to show the old conversation, and the user has no indication that the session was reset.
## Steps to Reproduce
1. Open the TUI (`openclaw tui... | open | null | false | 4 | [] | [] | 2026-03-07T15:37:25Z | 2026-05-04T06:59:27Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | yizhanzjz | 11,608,878 | MDQ6VXNlcjExNjA4ODc4 | User | false |
openclaw/openclaw | 4,038,745,434 | I_kwDOQb6kR87wul1a | 38,971 | https://github.com/openclaw/openclaw/issues/38971 | https://api.github.com/repos/openclaw/openclaw/issues/38971 | [Bug]: macOS app Settings → Cron tab shows "Error: cron.list: The data couldn't be read because it is missing." | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Crons are working fine — heartbeats firing, jobs executing normally on the gateway. Issue is purely in the macOS app's Cron tab UI. Likely a Swift JSON deserialization mismatch between what the gateway returns for `cron.list` and what the m... | closed | completed | true | 7 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T15:43:21Z | 2026-04-27T05:00:39Z | 2026-04-24T20:13:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wikithoughts | 7,737,137 | MDQ6VXNlcjc3MzcxMzc= | User | false |
openclaw/openclaw | 4,038,758,118 | I_kwDOQb6kR87wuo7m | 38,974 | https://github.com/openclaw/openclaw/issues/38974 | https://api.github.com/repos/openclaw/openclaw/issues/38974 | [Bug]: clawhub fetch failed all the time | ### Bug type
Regression (worked before, now fails)
### Summary
npx clawhub login
Opening browser: https://clawhub.ai/cli/auth?
✖ fetch failed
Error: fetch failed
### Steps to reproduce
just npx clawhub login
### Expected behavior
close
### Actual behavior
open
### OpenClaw v... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T15:49:48Z | 2026-04-27T05:00:38Z | 2026-04-24T20:13:19Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lmxiaonuo | 8,564,887 | MDQ6VXNlcjg1NjQ4ODc= | User | false |
openclaw/openclaw | 4,038,774,957 | I_kwDOQb6kR87wutCt | 38,980 | https://github.com/openclaw/openclaw/issues/38980 | https://api.github.com/repos/openclaw/openclaw/issues/38980 | Feishu DM: assistant emits text but dispatch completes with queuedFinal=false, replies=0 | ## Summary
Feishu direct-message replies can be dropped with `replies=0` even when the assistant produced normal text.
In affected sessions, gateway logs show:
- `dispatching to agent (session=agent:main:feishu:direct:...)`
- `dispatch complete (queuedFinal=false, replies=0)`
But session transcript contains assistant... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T15:57:40Z | 2026-04-27T05:00:38Z | 2026-04-24T20:13:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | nevergow | 50,450,991 | MDQ6VXNlcjUwNDUwOTkx | User | false |
openclaw/openclaw | 4,038,810,695 | I_kwDOQb6kR87wu1xH | 38,986 | https://github.com/openclaw/openclaw/issues/38986 | https://api.github.com/repos/openclaw/openclaw/issues/38986 | web_fetch fails when using VPN TUN mode (DNS hijacked to 198.18.x.x) | ## Description
When using a VPN in TUN mode (e.g., ShadowRocket on macOS), the system DNS is hijacked to 198.18.0.2, which returns fake IPs in the 198.18.x.x range for all domain lookups.
The web_fetch tool fails with:
```
Blocked: resolves to private/internal/special-use IP address
```
This happens because OpenClaw... | closed | completed | true | 6 | [
"stale"
] | [] | 2026-03-07T16:11:03Z | 2026-04-27T05:00:37Z | 2026-04-24T20:13:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kikkishook | 258,604,130 | U_kgDOD2n8Yg | User | false |
openclaw/openclaw | 4,038,813,980 | I_kwDOQb6kR87wu2kc | 38,987 | https://github.com/openclaw/openclaw/issues/38987 | https://api.github.com/repos/openclaw/openclaw/issues/38987 | [Bug]: Tools broken after first exec use Thread_ID: 1479868353546948780 & 1479809707693576263 | ### Bug type
Regression (worked before, now fails)
### Summary
Tools broken after first exec use
### Steps to reproduce
After starting a new session.
Bot tries exec command and get stuck.
See for Details the following to Discord Thread_IDs 1479868353546948780 & 1479809707693576263
### Expected behavior
exec... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T16:12:21Z | 2026-04-27T05:00:36Z | 2026-04-24T20:13:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | imheretorage | 264,646,979 | U_kgDOD8YxQw | User | false |
openclaw/openclaw | 4,038,815,490 | I_kwDOQb6kR87wu28C | 38,988 | https://github.com/openclaw/openclaw/issues/38988 | https://api.github.com/repos/openclaw/openclaw/issues/38988 | infra: Sentry監視環境のセットアップ | ## 概要
AgentOS cron タスク (agentos-sentry-poll) で Sentry 監視を実行するための環境セットアップが必要
## 現状
- sentry-cli がインストールされていない
- SENTRY_AUTH_TOKEN 等の認証情報が未設定
- .sentryclirc 設定ファイルが存在しない
## 必要な作業
1. sentry-cli のインストール
```bash
npm install -g @sentry/cli
# or
brew install getsentry/tools/sentry-cli
```
2. 認証トークンの設定
- S... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T16:12:53Z | 2026-04-27T05:00:35Z | 2026-04-24T20:19:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rnrnstar2 | 39,724,729 | MDQ6VXNlcjM5NzI0NzI5 | User | false |
openclaw/openclaw | 4,038,838,827 | I_kwDOQb6kR87wu8or | 38,995 | https://github.com/openclaw/openclaw/issues/38995 | https://api.github.com/repos/openclaw/openclaw/issues/38995 | Gemini 2.5 Flash produces duplicate response content in single turn | ## Environment
- OpenClaw version: 2026.3.2
- Model: google/gemini-2.5-flash
- Channel: Telegram (webchat surface)
- OS: Ubuntu 24.04 WSL2
## Description
When using Gemini 2.5 Flash as the primary model, responses are frequently duplicated 2-4x within a single turn. The exact same text block appears multiple times in ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T16:20:42Z | 2026-04-27T05:00:34Z | 2026-04-24T20:19:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Shawnski45 | 213,051,355 | U_kgDODLLn2w | User | false |
openclaw/openclaw | 4,038,843,869 | I_kwDOQb6kR87wu93d | 38,997 | https://github.com/openclaw/openclaw/issues/38997 | https://api.github.com/repos/openclaw/openclaw/issues/38997 | Signal: read receipts/typing indicators surfaced as <media:unknown> messages | ## Description
When using Signal channel, read receipts, typing indicators, and emoji reactions are delivered to the agent as `<media:unknown>` messages instead of being silently filtered or properly parsed.
## Environment
- OpenClaw: 2026.3.2
- signal-cli: 0.14.0 (also reproduced on 0.13.24)
- OS: macOS (arm64)
- C... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T16:22:51Z | 2026-04-27T05:00:34Z | 2026-04-24T20:19:16Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | nicolasgrasset | 347,560 | MDQ6VXNlcjM0NzU2MA== | User | false |
openclaw/openclaw | 4,038,845,226 | I_kwDOQb6kR87wu-Mq | 38,998 | https://github.com/openclaw/openclaw/issues/38998 | https://api.github.com/repos/openclaw/openclaw/issues/38998 | Feature: pairing:request hook event for proactive notifications | ## Summary
When a channel with `dmPolicy: "pairing"` receives a message from an unknown sender, it auto-generates a pairing code and replies — but the agent is never notified. The owner only discovers pending pairing requests by manually running `clawdbot pairing list <channel>`.
## Problem
There's no proactive way ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T16:23:23Z | 2026-04-28T04:42:41Z | 2026-04-26T02:42:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | TheHubOfGitMayPolitelySuckMyAss | 255,424,812 | U_kgDODzl5LA | User | false |
openclaw/openclaw | 4,038,858,115 | I_kwDOQb6kR87wvBWD | 39,000 | https://github.com/openclaw/openclaw/issues/39000 | https://api.github.com/repos/openclaw/openclaw/issues/39000 | Feature request: api.callGatewayMethod() for inter-plugin communication | ## Summary
Plugins can register gateway methods via `api.registerGatewayMethod()`, but there is no corresponding `api.callGatewayMethod()` to invoke them from within another plugin. This makes cross-plugin communication impossible through the intended API surface.
## Problem
OpenClaw's plugin API is scoped per-plugi... | open | null | false | 2 | [] | [] | 2026-03-07T16:27:37Z | 2026-05-04T06:59:35Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | coolmanns | 117,685,070 | U_kgDOBwO7Tg | User | false |
openclaw/openclaw | 4,038,868,241 | I_kwDOQb6kR87wvD0R | 39,007 | https://github.com/openclaw/openclaw/issues/39007 | https://api.github.com/repos/openclaw/openclaw/issues/39007 | [Bug]: Browser tool times out reaching browser control service though service is listening | ### Bug type
Regression (worked before, now fails)
### Summary
Direct browser profile (`openclaw`) consistently fails with timeout: "Can't reach the OpenClaw browser control service" even though browser control server is listening on 127.0.0.1:18791 and returns 401 Unauthorized. Browser Relay works.
### Steps to re... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T16:32:20Z | 2026-04-27T05:00:33Z | 2026-04-24T20:19:20Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | stanleyau-xx | 3,599,587 | MDQ6VXNlcjM1OTk1ODc= | User | false |
openclaw/openclaw | 4,038,882,719 | I_kwDOQb6kR87wvHWf | 39,010 | https://github.com/openclaw/openclaw/issues/39010 | https://api.github.com/repos/openclaw/openclaw/issues/39010 | Critical: Telegram message replay storm after upgrade to 2026.3.2 | ## Summary
After upgrading from 2026.2.23 to 2026.3.2, a **critical message replay storm** occurs where already-processed Telegram messages are repeatedly re-delivered to agent sessions in an infinite loop (~30 min cycle). Rolling back to 2026.2.23 resolves the issue.
## Severity: Critical
- Messages replayed **indef... | closed | completed | true | 5 | [
"stale"
] | [] | 2026-03-07T16:37:55Z | 2026-04-27T05:00:32Z | 2026-04-24T20:19:23Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 4ier | 5,648,066 | MDQ6VXNlcjU2NDgwNjY= | User | false |
openclaw/openclaw | 4,038,892,443 | I_kwDOQb6kR87wvJub | 39,013 | https://github.com/openclaw/openclaw/issues/39013 | https://api.github.com/repos/openclaw/openclaw/issues/39013 | Hook-injected virtual files show as `undefined` in `/context list`, and session memory keeps recalling hook content as if it were real session history | Two related bugs I ran into while investigating some weird context bloat. They're separate issues but they interact in a confusing way, so filing together.
## 1. Virtual bootstrap files show as `undefined` in `/context list`
When a hook injects a virtual file via `event.context.bootstrapFiles.push({ path: 'FILENAME.m... | closed | not_planned | true | 4 | [] | [] | 2026-03-07T16:42:36Z | 2026-05-01T04:40:37Z | 2026-04-29T03:21:18Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | far-ouq | 125,839,498 | U_kgDOB4Aoig | User | false |
openclaw/openclaw | 4,038,925,856 | I_kwDOQb6kR87wvR4g | 39,022 | https://github.com/openclaw/openclaw/issues/39022 | https://api.github.com/repos/openclaw/openclaw/issues/39022 | Feature Request: Per-model turn limits with auto-revert | ## Summary
Add support for `/model <name> --turns N` that automatically reverts to the session default model after N turns.
## Use Case
Teams sharing a bot where users can access premium models (GPT-5, Claude) but should not leave them on accidentally for long projects or cron jobs. A turn limit ensures expensive mo... | open | null | false | 2 | [] | [] | 2026-03-07T16:57:25Z | 2026-05-04T06:59:40Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | krmesser | 30,294,918 | MDQ6VXNlcjMwMjk0OTE4 | User | false |
openclaw/openclaw | 4,038,964,173 | I_kwDOQb6kR87wvbPN | 39,031 | https://github.com/openclaw/openclaw/issues/39031 | https://api.github.com/repos/openclaw/openclaw/issues/39031 | Feature: Forward iMessage tapback reactions to agent | ## Summary
iMessage tapback reactions (👍❤️😂 etc.) are captured in the Messages database and visible via `imsg history`, but the iMessage channel plugin doesn't forward them to the agent as events.
## Use Case: Lightweight Affordances
Two-way tapbacks could serve as lightweight affordances:
1. **Agent → User**: Ag... | open | null | false | 2 | [] | [] | 2026-03-07T17:14:56Z | 2026-05-04T06:59:42Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | TheHubOfGitMayPolitelySuckMyAss | 255,424,812 | U_kgDODzl5LA | User | false |
openclaw/openclaw | 4,038,965,611 | I_kwDOQb6kR87wvblr | 39,032 | https://github.com/openclaw/openclaw/issues/39032 | https://api.github.com/repos/openclaw/openclaw/issues/39032 | Subagent completion output leaks internal tool-failure reasoning to requester session | ## Summary
When a subagent encounters a tool failure (e.g. calling `message` with a session key instead of a Telegram chat ID), the agent's internal reasoning about the failure leaks into the auto-announce completion output delivered to the requester session.
## Steps to reproduce
1. Spawn a subagent with `sessions_... | open | null | false | 4 | [] | [] | 2026-03-07T17:15:28Z | 2026-05-04T06:59:44Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | cryptopoly | 31,970,407 | MDQ6VXNlcjMxOTcwNDA3 | User | false |
openclaw/openclaw | 4,038,983,021 | I_kwDOQb6kR87wvf1t | 39,035 | https://github.com/openclaw/openclaw/issues/39035 | https://api.github.com/repos/openclaw/openclaw/issues/39035 | [Bug]: openai-codex OAuth succeeds on macOS, but all Codex requests fall back to Gemini | ### Bug type
Regression (worked before, now fails)
### Summary
Summary
On macOS, OpenClaw can complete `openai-codex` OAuth successfully and enumerate Codex models, but actual requests to Codex models never complete. The agent always falls back to Gemini.
Environment
- OpenClaw: `2026.3.7` (GitHub source build)
-... | closed | completed | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T17:21:27Z | 2026-04-27T05:00:31Z | 2026-04-24T20:19:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | fangjing2021 | 91,580,531 | U_kgDOBXVocw | User | false |
openclaw/openclaw | 4,038,987,117 | I_kwDOQb6kR87wvg1t | 39,037 | https://github.com/openclaw/openclaw/issues/39037 | https://api.github.com/repos/openclaw/openclaw/issues/39037 | WhatsApp channel: expose ctwa_clid from Click-to-WhatsApp ads in session metadata | ## Problem
When a lead arrives from a Click-to-WhatsApp (CTWA) Meta ad, the WhatsApp webhook includes a `referral` object in the message's `contextInfo` with the `ctwaClid` (Click-to-WhatsApp Click ID).
This click ID is required by Meta's Conversions API (`business_messaging` action source) to attribute conversions ba... | open | null | false | 3 | [] | [] | 2026-03-07T17:22:41Z | 2026-05-04T06:59:46Z | null | CONTRIBUTOR | null | 20260504T090105Z | 2026-05-04T09:01:05Z | zomars | 3,504,472 | MDQ6VXNlcjM1MDQ0NzI= | User | false |
openclaw/openclaw | 4,038,990,174 | I_kwDOQb6kR87wvhle | 39,038 | https://github.com/openclaw/openclaw/issues/39038 | https://api.github.com/repos/openclaw/openclaw/issues/39038 | [Bug]: OpenClaw节点程序在Windows 11 24H2上启动后卡在PATH信息，无法连接Gateway | ### Bug type
Crash (process/app exits or hangs)
### Summary
"节点程序在Windows 11 24H2启动后卡在PATH信息
### Steps to reproduce
1.确保Gateway运行正常（openclaw gateway start）
2.运行 openclaw node run 或 openclaw node run --host [127.0.0.1](http://127.0.0.1/) --port 18789
3.观察：显示PATH信息后程序卡住，无进一步输出
### Expected behavior
1.显示PATH信息
2.连... | open | null | false | 8 | [
"bug",
"bug:crash"
] | [] | 2026-03-07T17:23:42Z | 2026-05-04T09:03:17Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Joper89 | 266,336,986 | U_kgDOD9_62g | User | false |
openclaw/openclaw | 4,039,030,201 | I_kwDOQb6kR87wvrW5 | 39,057 | https://github.com/openclaw/openclaw/issues/39057 | https://api.github.com/repos/openclaw/openclaw/issues/39057 | [Bug]: `openclaw node status` reports "stopped" on German Windows while task is running | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Version
OpenClaw 2026.3.2 (85377a2)
## Platform
- Node host: Windows 10/11, German locale (de-DE)
- Gateway: Ubuntu Linux (noble)
## Description
`openclaw node status` on the Windows node host reports the service as "stopped", even tho... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T17:42:45Z | 2026-04-27T05:00:30Z | 2026-04-24T20:19:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | a-p-l | 56,471,516 | MDQ6VXNlcjU2NDcxNTE2 | User | false |
openclaw/openclaw | 4,039,030,948 | I_kwDOQb6kR87wvrik | 39,058 | https://github.com/openclaw/openclaw/issues/39058 | https://api.github.com/repos/openclaw/openclaw/issues/39058 | [Bug]: Session delete button too easy to misclick - accidental deletion without confirmation | ## Bug Description
In the ClawX desktop app, the session delete button (X) in the left sidebar is too easy to accidentally click when trying to switch between sessions. This results in unintended deletion of conversation history.
## Current Behavior
- Clicking on a session in the left sidebar sometimes triggers the de... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T17:43:08Z | 2026-04-27T05:00:29Z | 2026-04-24T20:19:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | longxia89532 | 266,341,167 | U_kgDOD-ALLw | User | false |
openclaw/openclaw | 4,039,045,624 | I_kwDOQb6kR87wvvH4 | 39,066 | https://github.com/openclaw/openclaw/issues/39066 | https://api.github.com/repos/openclaw/openclaw/issues/39066 | [BUG]: delivery-recovery races ahead of channel startup, causing permanent message loss | ## Problem
In `src/gateway/server.impl.ts`, `recoverPendingDeliveries` is launched as a fire-and-forget async at line 760:
```typescript
void (async () => {
const { recoverPendingDeliveries } = await import("../infra/outbound/delivery-queue.js");
const { deliverOutboundPayloads } = await import("../infra/outbound... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T17:50:16Z | 2026-04-27T05:00:28Z | 2026-04-24T20:19:46Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kenkamwaileung | 266,336,690 | U_kgDOD9_5sg | User | false |
openclaw/openclaw | 4,039,049,799 | I_kwDOQb6kR87wvwJH | 39,068 | https://github.com/openclaw/openclaw/issues/39068 | https://api.github.com/repos/openclaw/openclaw/issues/39068 | [Bug]: on new install it is not possible to enable tools on main agent | ### Bug type
Regression (worked before, now fails)
### Summary
Trying to install OpenClaw using standard installer. Main agent locked down to messaging tools only, impossible to enable any other tool.
### Steps to reproduce
1. Use the script to install
2. go through the wizard, configured model and everything
3. ... | closed | completed | true | 6 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T17:52:16Z | 2026-04-27T05:00:27Z | 2026-04-24T20:19:49Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | francip | 49,422 | MDQ6VXNlcjQ5NDIy | User | false |
openclaw/openclaw | 4,039,108,650 | I_kwDOQb6kR87wv-gq | 39,082 | https://github.com/openclaw/openclaw/issues/39082 | https://api.github.com/repos/openclaw/openclaw/issues/39082 | Flush pending chat delta before emitting tool events in server-chat.ts | ## Problem
When the model produces text followed by a tool call, the Control UI receives the tool event before the final text delta has been delivered. This causes text preceding a tool card to appear truncated during streaming.
### Root Cause
In `server-chat.ts`, `emitChatDelta` throttles text broadcasts at 150ms i... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T18:20:41Z | 2026-04-27T05:00:26Z | 2026-04-24T20:25:15Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jakepresent | 31,285,242 | MDQ6VXNlcjMxMjg1MjQy | User | false |
openclaw/openclaw | 4,039,115,826 | I_kwDOQb6kR87wwAQy | 39,086 | https://github.com/openclaw/openclaw/issues/39086 | https://api.github.com/repos/openclaw/openclaw/issues/39086 | [Bug]: `store: false` is still sent in embedded agent runs despite `compat.supportsStore: false`, and custom headers are dropped | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using Google AI Studio (Gemini) via an OpenAI-compatible endpoint (like Cloudflare AI Gateway), OpenClaw throws `400 Bad Request` or `401 Unauthorized` errors.
After deep debugging via a local proxy, I found two distinct issues in Op... | closed | completed | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T18:23:49Z | 2026-04-27T05:00:25Z | 2026-04-24T20:25:19Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wonderful259 | 21,329,339 | MDQ6VXNlcjIxMzI5MzM5 | User | false |
openclaw/openclaw | 4,039,148,021 | I_kwDOQb6kR87wwIH1 | 39,093 | https://github.com/openclaw/openclaw/issues/39093 | https://api.github.com/repos/openclaw/openclaw/issues/39093 | [Bug]: Compact context can be timeout during long-time-span tasks, also those long tasks just stuck without any retry. | ### Bug type
Regression (worked before, now fails)
### Summary
ver: 2026.3.2
desc: very easy to get stuck during compacting context. which would result in long task block all night.
### Steps to reproduce
1. Make a long task that takes hours to finish.
2. Context become very long due to it uses browser to operat... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T18:38:31Z | 2026-04-27T05:00:24Z | 2026-04-24T20:25:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | AntiMoron | 6,587,734 | MDQ6VXNlcjY1ODc3MzQ= | User | false |
openclaw/openclaw | 4,039,150,527 | I_kwDOQb6kR87wwIu_ | 39,094 | https://github.com/openclaw/openclaw/issues/39094 | https://api.github.com/repos/openclaw/openclaw/issues/39094 | Feature Request: Add reasoningDefault config option | ## Summary
Add a `reasoningDefault` config option to `agents.defaults`, mirroring `thinkingDefault` and `verboseDefault` patterns.
## Motivation
Currently `/reasoning on|off|stream` defaults to `off` with no way to change via config. Every session requires manual `/reasoning on`.
`thinkingDefault`, `verboseDefault`... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T18:39:51Z | 2026-04-29T10:47:20Z | 2026-04-26T01:49:22Z | NONE | null | 20260429T120040Z | 2026-04-29T12:00:40Z | wilsoniu | 29,446,464 | MDQ6VXNlcjI5NDQ2NDY0 | User | false |
openclaw/openclaw | 4,039,152,896 | I_kwDOQb6kR87wwJUA | 39,096 | https://github.com/openclaw/openclaw/issues/39096 | https://api.github.com/repos/openclaw/openclaw/issues/39096 | Discord: health-monitor 'stuck' detection triggers excessive false-positive reconnects | ## Description
The built-in Discord health-monitor fires 'stuck' reconnects at an extremely high rate even when the gateway is otherwise healthy — cron jobs running, messages flowing, no WS death spiral in progress.
## Environment
- OpenClaw: v2026.3.2
- OS: Windows 10 x64
- Discord channel, single guild
## Observed... | closed | completed | true | 5 | [
"stale"
] | [] | 2026-03-07T18:40:55Z | 2026-04-27T05:00:24Z | 2026-04-24T20:25:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Stache73 | 7,196,384 | MDQ6VXNlcjcxOTYzODQ= | User | false |
openclaw/openclaw | 4,039,155,890 | I_kwDOQb6kR87wwKCy | 39,097 | https://github.com/openclaw/openclaw/issues/39097 | https://api.github.com/repos/openclaw/openclaw/issues/39097 | Browser: support blocked URL paths / domain denylist | ## Problem
When the browser navigates to a site protected by bot detection (Incapsula, Cloudflare, etc.), it fails silently or returns an unhelpful "Request unsuccessful" page. The agent has no way to know in advance which domains will block it, and ends up wasting many turns retrying or trying workarounds.
## Propos... | open | null | false | 2 | [] | [] | 2026-03-07T18:41:58Z | 2026-05-04T06:59:52Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | openclaw-dok[bot] | 260,007,003 | BOT_kgDOD39kWw | Bot | false |
openclaw/openclaw | 4,039,174,110 | I_kwDOQb6kR87wwOfe | 39,101 | https://github.com/openclaw/openclaw/issues/39101 | https://api.github.com/repos/openclaw/openclaw/issues/39101 | voice-call: staleCallReaperSeconds missing from openclaw.plugin.json configSchema | ## Bug
`staleCallReaperSeconds` is defined in the Zod schema in `extensions/voice-call/src/config.ts` (line ~297) but is **not present** in the JSON Schema in `openclaw.plugin.json`.
Since the JSON Schema has `"additionalProperties": false`, adding `staleCallReaperSeconds` to the config causes:
```
Invalid config at ... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T18:50:13Z | 2026-04-27T05:00:23Z | 2026-04-24T20:25:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | afalk42 | 6,906,332 | MDQ6VXNlcjY5MDYzMzI= | User | false |
openclaw/openclaw | 4,039,236,048 | I_kwDOQb6kR87wwdnQ | 39,110 | https://github.com/openclaw/openclaw/issues/39110 | https://api.github.com/repos/openclaw/openclaw/issues/39110 | Docs drift: session-tool says subagents cannot spawn, but nested orchestration is supported with maxSpawnDepth>=2 | ## Summary
`docs/concepts/session-tool.md` currently states that sub-agents cannot call `sessions_spawn`, but `docs/tools/subagents.md` documents nested orchestration where depth-1 sub-agents can spawn depth-2 workers when `maxSpawnDepth >= 2`.
This causes confusion for operators implementing orchestrator-worker patte... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T19:17:13Z | 2026-04-27T05:00:22Z | 2026-04-24T20:25:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | metzmitchell | 170,978,957 | U_kgDOCjDujQ | User | false |
openclaw/openclaw | 4,039,262,406 | I_kwDOQb6kR87wwkDG | 39,125 | https://github.com/openclaw/openclaw/issues/39125 | https://api.github.com/repos/openclaw/openclaw/issues/39125 | bug: Spanish (es) locale missing from published npm package build artifacts | ## Summary
The Spanish (`es`) locale is documented as supported and fully implemented in source (`ui/src/i18n/locales/es.ts`, merged in b3fb881a), but the compiled chunk is missing from the published npm package (`2026.3.2`).
Users installing the current release get no Spanish UI — the browser language detection sile... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T19:28:37Z | 2026-04-27T05:00:21Z | 2026-04-24T20:25:37Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | RodolfoMiguelLopez | 2,590,663 | MDQ6VXNlcjI1OTA2NjM= | User | false |
openclaw/openclaw | 4,039,295,529 | I_kwDOQb6kR87wwsIp | 39,134 | https://github.com/openclaw/openclaw/issues/39134 | https://api.github.com/repos/openclaw/openclaw/issues/39134 | [Bug]: Anthropic API validation error: tools missing required 'function' field in cron job sessions | ### Bug type
Regression (worked before, now fails)
### Summary
lastTouchedVersion: "2026.2.15"
lastRunVersion: "2026.2.15"
Last updated: 2026-03-07
What it means:
• The OpenClaw gateway's tool definitions are missing a required field called 'function'
• When the system tries to invoke tools (like read, kimi_search... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T19:42:51Z | 2026-04-27T05:00:21Z | 2026-04-24T20:25:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | fli-rpx | 1,309,970 | MDQ6VXNlcjEzMDk5NzA= | User | false |
openclaw/openclaw | 4,039,301,944 | I_kwDOQb6kR87wwts4 | 39,138 | https://github.com/openclaw/openclaw/issues/39138 | https://api.github.com/repos/openclaw/openclaw/issues/39138 | Gateway 重启后 delivery recovery 重新发送大量旧消息（包括 15 小时前的） | **问题描述**
Gateway 重启后，delivery recovery 机制重新发送了 110+ 条旧消息，包括 15 小时前的消息，导致用户收到大量不相关的消息。
**时间线**
1. 用户发送新任务（18:50）
2. 我回复了 3 条消息（18:51）
3. **Gateway 收到 SIGTERM 信号，重启了（18:54:40）**
4. delivery recovery 开始恢复 110 条待发送的消息（18:54:45）
5. 用户收到旧消息（15 小时前的消息也被发送）
**根本原因**
1. Gateway 重启时，delivery recovery 机制把之前的消息都重新发送了
2. **没有消... | closed | not_planned | true | 3 | [] | [] | 2026-03-07T19:45:28Z | 2026-05-02T04:28:41Z | 2026-04-29T09:10:05Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | ksxh0524 | 40,232,908 | MDQ6VXNlcjQwMjMyOTA4 | User | false |
openclaw/openclaw | 4,039,322,363 | I_kwDOQb6kR87wwyr7 | 39,142 | https://github.com/openclaw/openclaw/issues/39142 | https://api.github.com/repos/openclaw/openclaw/issues/39142 | Configurable user-facing message when all models are exhausted | ## Problem
When all configured models hit their API limits (daily cap, billing, rate limit), the user sees a raw error:
```
Agent failed before reply: All models failed (2): openrouter/google/gemini-2.5-pro: 403 Key limit exceeded (daily limit) | openrouter/google/gemini-2.5-flash: 403 Key limit exceeded (daily limit... | open | null | false | 2 | [] | [] | 2026-03-07T19:53:03Z | 2026-05-04T07:00:21Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Sokanon | 53,493,508 | MDQ6VXNlcjUzNDkzNTA4 | User | false |
openclaw/openclaw | 4,039,322,786 | I_kwDOQb6kR87wwyyi | 39,143 | https://github.com/openclaw/openclaw/issues/39143 | https://api.github.com/repos/openclaw/openclaw/issues/39143 | Set Discord bot presence to idle/DND when agent is in degraded state | ## Problem
When all configured models are exhausted (daily API limit, billing errors), the Discord bot stays "online" even though it can't respond to anything. Users keep messaging it expecting replies, not knowing it's effectively down until limits reset.
## Proposed solution
Update the Discord bot's presence/activ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T19:53:12Z | 2026-04-27T05:00:20Z | 2026-04-24T20:25:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Sokanon | 53,493,508 | MDQ6VXNlcjUzNDkzNTA4 | User | false |
openclaw/openclaw | 4,039,380,791 | I_kwDOQb6kR87wxA83 | 39,160 | https://github.com/openclaw/openclaw/issues/39160 | https://api.github.com/repos/openclaw/openclaw/issues/39160 | RFC: CaMeL Prompt Injection Defense for OpenClaw | # RFC: CaMeL Prompt Injection Defense for OpenClaw
## Summary
Add opt-in CaMeL-style (CApabilities for MachinE Learning) prompt injection defense to OpenClaw, implementing data provenance tracking and capability-based security policies at the tool-call boundary. This is a native TypeScript implementation inspired by ... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-07T20:17:21Z | 2026-04-28T04:43:12Z | 2026-04-26T01:59:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,039,412,290 | I_kwDOQb6kR87wxIpC | 39,164 | https://github.com/openclaw/openclaw/issues/39164 | https://api.github.com/repos/openclaw/openclaw/issues/39164 | Feature: Auto-detect rtk for 60-90% exec output token savings | ## Problem
Every `exec` tool call returns raw shell output into the agent's context window. Verbose commands like `git status`, `ls -la`, `grep`, and test runners consume thousands of tokens on output that could be compressed 60-90% without losing actionable information.
## Proposal
Integrate [rtk](https://github.co... | closed | completed | true | 4 | [] | [] | 2026-03-07T20:31:31Z | 2026-04-30T04:50:17Z | 2026-04-25T05:38:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | TheAgentWire | 260,305,314 | U_kgDOD4Pxog | User | false |
openclaw/openclaw | 4,039,448,682 | I_kwDOQb6kR87wxRhq | 39,170 | https://github.com/openclaw/openclaw/issues/39170 | https://api.github.com/repos/openclaw/openclaw/issues/39170 | voice-call: conversation mode drops after 1s — maybeSpeakInitialMessageOnAnswered race condition | ## Bug
Conversation mode calls drop after ~1 second. Notify mode works fine.
## Root Cause
`maybeSpeakInitialMessageOnAnswered()` fires on the `call.answered` webhook — **before** the Twilio media stream WebSocket connects. With no stream available, `playTts()` falls back to a TwiML `<Say>` REST API call (`PUT /Call... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T20:48:07Z | 2026-04-27T05:00:18Z | 2026-04-24T20:28:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | zurbrick | 24,554,955 | MDQ6VXNlcjI0NTU0OTU1 | User | false |
openclaw/openclaw | 4,039,460,900 | I_kwDOQb6kR87wxUgk | 39,173 | https://github.com/openclaw/openclaw/issues/39173 | https://api.github.com/repos/openclaw/openclaw/issues/39173 | Heartbeat runner never re-fires when a run exceeds the 'every' interval | ## Description
When a heartbeat run takes longer than (or close to) the configured `every` interval, the next heartbeat never fires. The runner becomes permanently stuck until the gateway is restarted.
## Root cause
In `health-*.js` (`startHeartbeatRunner`), when a heartbeat run is in progress and the interval timer... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T20:52:21Z | 2026-04-27T04:54:58Z | 2026-04-24T22:05:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jean1190 | 125,493,381 | U_kgDOB3rghQ | User | false |
openclaw/openclaw | 4,039,472,261 | I_kwDOQb6kR87wxXSF | 39,176 | https://github.com/openclaw/openclaw/issues/39176 | https://api.github.com/repos/openclaw/openclaw/issues/39176 | SQLite busy_timeout should be set on every new connection (not persistent across restarts) | ## Problem
`PRAGMA busy_timeout` is a per-connection setting — it resets to 0 on every new SQLite connection. With 10+ concurrent processes (gateway, cron agents, subagents) hitting `main.sqlite`, a `busy_timeout=0` means any concurrent write attempt instantly fails with `SQLITE_BUSY` instead of retrying.
## Context
... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T20:56:24Z | 2026-04-27T04:54:57Z | 2026-04-24T22:05:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | nishantkabra77 | 194,645,679 | U_kgDOC5oOrw | User | false |
openclaw/openclaw | 4,039,511,725 | I_kwDOQb6kR87wxg6t | 39,179 | https://github.com/openclaw/openclaw/issues/39179 | https://api.github.com/repos/openclaw/openclaw/issues/39179 | Moltbook skill: threaded comment replies not supported (parent_comment_id rejected) | ## Summary
When using the Moltbook API to reply to a specific comment, passing `parent_comment_id` in the POST body to `/api/v1/posts/{id}/comments` returns a 400 error:
```json
{
"statusCode": 400,
"message": ["property parent_comment_id should not exist"],
"error": "Bad Request"
}
```
The workaround is to po... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-07T21:08:33Z | 2026-04-27T05:00:18Z | 2026-04-24T20:28:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ocagent-merlin | 264,257,253 | U_kgDOD8A-5Q | User | false |
openclaw/openclaw | 4,039,520,782 | I_kwDOQb6kR87wxjIO | 39,180 | https://github.com/openclaw/openclaw/issues/39180 | https://api.github.com/repos/openclaw/openclaw/issues/39180 | Document uploads (XLS/XLSX) blow out context window — need content size limit or save-to-disk mode | When a user uploads an Excel file (.xls/.xlsx) via Discord or Telegram, OpenClaw converts the binary file to text and injects the full contents into the user message. A 493KB XLS file (8,336 rows × 2 columns) expanded to **1.16 million characters** of text in the context window, instantly exceeding the context limit an... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T21:11:35Z | 2026-04-27T05:00:17Z | 2026-04-24T20:28:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rharlan-ve | 266,372,111 | U_kgDOD-CEDw | User | false |
openclaw/openclaw | 4,039,546,477 | I_kwDOQb6kR87wxpZt | 39,185 | https://github.com/openclaw/openclaw/issues/39185 | https://api.github.com/repos/openclaw/openclaw/issues/39185 | [Bug]: Heartbeat subsystem ignores HEARTBEAT.md configuration, defaults to HEARTBEAT_OK instead of service-specific status reports | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Description,
The heartbeat subsystem starts successfully but fails to generate the required service-specific reports as defined in HEARTBEAT.md. Instead of following the configured report format, it defaults to sending HEARTBEAT_OK, which v... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T21:20:41Z | 2026-04-27T05:00:16Z | 2026-04-24T20:28:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Nemesis851 | 17,552,992 | MDQ6VXNlcjE3NTUyOTky | User | false |
openclaw/openclaw | 4,039,549,412 | I_kwDOQb6kR87wxqHk | 39,187 | https://github.com/openclaw/openclaw/issues/39187 | https://api.github.com/repos/openclaw/openclaw/issues/39187 | [Bug]: After upgrade from original Clawdbot to OpenClaw features stopped working | ### Bug type
Regression (worked before, now fails)
### Summary
I have upgraded from original clawdbot to openclaw now, everything seems to be working fine, no gateway issues, telegram chats all work but occasionally if i ask it to do something via TG like scrape this site it just goes silent, not information that it... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T21:21:45Z | 2026-04-27T05:00:15Z | 2026-04-24T20:28:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | REPTILEHAUS | 4,680,621 | MDQ6VXNlcjQ2ODA2MjE= | User | false |
openclaw/openclaw | 4,039,556,939 | I_kwDOQb6kR87wxr9L | 39,188 | https://github.com/openclaw/openclaw/issues/39188 | https://api.github.com/repos/openclaw/openclaw/issues/39188 | bug: parseLaunchctlPrint accepts PID 0 (inconsistent with systemd parser) | ## Bug
`parseLaunchctlPrint()` in `src/daemon/launchd.ts:131` only checks `Number.isFinite(pid)` before storing the PID, but the equivalent systemd parser at `src/daemon/systemd.ts:122` correctly checks `Number.isFinite(pid) && pid > 0`.
When `launchctl print` reports `pid = 0` (meaning no running process), the launc... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-07T21:24:24Z | 2026-05-01T04:40:20Z | 2026-04-29T04:29:07Z | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | scoootscooob | 167,050,519 | U_kgDOCfT9Fw | User | false |
openclaw/openclaw | 4,039,557,204 | I_kwDOQb6kR87wxsBU | 39,190 | https://github.com/openclaw/openclaw/issues/39190 | https://api.github.com/repos/openclaw/openclaw/issues/39190 | bug: voice-call STT vadThreshold: 0 and silenceDurationMs: 0 silently ignored | ## Bug
In `extensions/voice-call/src/providers/stt-openai-realtime.ts:65-66`, the constructor uses `||` (logical OR) instead of `??` (nullish coalescing) for default values:
```typescript
this.silenceDurationMs = config.silenceDurationMs || 800;
this.vadThreshold = config.vadThreshold || 0.5;
```
This drops valid `0... | closed | completed | true | 5 | [
"stale"
] | [] | 2026-03-07T21:24:35Z | 2026-04-27T04:39:58Z | 2026-04-25T02:35:02Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | scoootscooob | 167,050,519 | U_kgDOCfT9Fw | User | false |
openclaw/openclaw | 4,039,557,486 | I_kwDOQb6kR87wxsFu | 39,191 | https://github.com/openclaw/openclaw/issues/39191 | https://api.github.com/repos/openclaw/openclaw/issues/39191 | bug: synology-chat ignores SYNOLOGY_RATE_LIMIT=0 (parseInt || 30 truthy check) | ## Bug
In `extensions/synology-chat/src/accounts.ts:83`, the rate limit parsing uses `||` instead of `??`:
```typescript
rateLimitPerMinute:
accountOverride.rateLimitPerMinute ??
channelCfg.rateLimitPerMinute ??
(envRateLimit ? parseInt(envRateLimit, 10) || 30 : 30),
```
When `SYNOLOGY_RATE_LIMIT=0`, `parseInt... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-07T21:24:41Z | 2026-05-01T04:40:19Z | 2026-04-29T04:29:09Z | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | scoootscooob | 167,050,519 | U_kgDOCfT9Fw | User | false |
openclaw/openclaw | 4,039,561,663 | I_kwDOQb6kR87wxtG_ | 39,194 | https://github.com/openclaw/openclaw/issues/39194 | https://api.github.com/repos/openclaw/openclaw/issues/39194 | [Bug]: TTS Feature in Chinese language duplicate replies. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Hi Genius !
This is a bug report from a normal user and I don't know any programming language.
I am currently running Version: 2026.3.2 on a vps and I use telegram channel.
I noticed if I use /tts in telegram chatbox, I got a dashboard and ... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T21:26:13Z | 2026-04-27T05:00:15Z | 2026-04-24T20:28:51Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | funbien | 70,850,499 | MDQ6VXNlcjcwODUwNDk5 | User | false |
openclaw/openclaw | 4,039,623,032 | I_kwDOQb6kR87wx8F4 | 39,217 | https://github.com/openclaw/openclaw/issues/39217 | https://api.github.com/repos/openclaw/openclaw/issues/39217 | [Bug] TUI: /new session replies broadcast to all connected TUI clients for the same agent | ### Bug type
Regression (worked before, now fails)
### Summary
When two TUI windows are connected to the same agent, creating a new session with /new in one window causes the agent's replies to be broadcast to all connected TUI clients, not just the one that initiated the session.
### Steps to reproduce
1. Open te... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T21:52:33Z | 2026-04-27T05:00:14Z | 2026-04-24T20:28:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | cyq1017 | 61,975,706 | MDQ6VXNlcjYxOTc1NzA2 | User | false |
openclaw/openclaw | 4,039,630,211 | I_kwDOQb6kR87wx92D | 39,218 | https://github.com/openclaw/openclaw/issues/39218 | https://api.github.com/repos/openclaw/openclaw/issues/39218 | [Docs] Cron delivery: document announce mode behavior and common pitfalls for multi-channel agents | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The "announce" delivery mode for Cron jobs lacks documentation on critical behaviors: multi-channel selection priority, invalid targeting fields (channel/to), and recovery from exponential backoff (consecutiveErrors in jobs.json).
### Step... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T21:55:15Z | 2026-04-27T05:00:13Z | 2026-04-24T20:28:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | cyq1017 | 61,975,706 | MDQ6VXNlcjYxOTc1NzA2 | User | false |
openclaw/openclaw | 4,039,637,688 | I_kwDOQb6kR87wx_q4 | 39,221 | https://github.com/openclaw/openclaw/issues/39221 | https://api.github.com/repos/openclaw/openclaw/issues/39221 | Bug: Image attachments not forwarded from web dashboard | **Channel:** Web Dashboard
**Severity:** Medium
**Description:**
Images pasted into the web dashboard chat are not being forwarded to the agent. Text arrives but the image attachment doesn't.
**Reproduction:**
1. Open web dashboard
2. Paste image into chat input
3. Send message with text
4. Agent receives text but NO... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-07T21:57:50Z | 2026-04-27T05:00:12Z | 2026-04-24T20:29:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | SPRIC76 | 202,198,248 | U_kgDODA1M6A | User | false |
openclaw/openclaw | 4,039,649,234 | I_kwDOQb6kR87wyCfS | 39,222 | https://github.com/openclaw/openclaw/issues/39222 | https://api.github.com/repos/openclaw/openclaw/issues/39222 | CLI can leave orphan openclaw-gateway processes on Linux/systemd when gateway.mode=local | ## Summary
`openclaw` CLI commands can leave behind orphan `openclaw` / `openclaw-gateway` processes on Linux when `gateway.mode` is `local` and a systemd-managed gateway is already running.
This can lead to:
- duplicate gateway processes
- port `18789` conflicts
- `Gateway already running locally` / `another gatewa... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-07T22:01:18Z | 2026-04-27T05:00:11Z | 2026-04-24T20:29:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ottopan | 18,224,971 | MDQ6VXNlcjE4MjI0OTcx | User | false |
openclaw/openclaw | 4,039,652,869 | I_kwDOQb6kR87wyDYF | 39,223 | https://github.com/openclaw/openclaw/issues/39223 | https://api.github.com/repos/openclaw/openclaw/issues/39223 | [Bug]: Configure wizard hangs at Gateway selection - blocks OAuth auth flow | ## Bug Description
The `openclaw configure` wizard hangs/crashes at the 'Where will the Gateway run?' step, preventing users from completing the OAuth authentication flow for Anthropic.
## Environment
- **OpenClaw Version:** 2026.3.2 (b817600)
- **OS:** Linux (Docker container)
- **Node.js:** v22.22.0
## Steps to R... | open | null | false | 6 | [] | [] | 2026-03-07T22:02:44Z | 2026-05-04T07:00:28Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | LevityLeads | 194,811,850 | U_kgDOC5yXyg | User | false |
openclaw/openclaw | 4,039,653,425 | I_kwDOQb6kR87wyDgx | 39,224 | https://github.com/openclaw/openclaw/issues/39224 | https://api.github.com/repos/openclaw/openclaw/issues/39224 | [Bug] memory index/search breaks in FTS-only mode (provider=none) / FTS-only 模式下 memory 索引与检索异常 | # [Bug] memory index/search breaks in FTS-only mode (`provider=none`) / FTS-only 模式下 memory 索引与检索异常
## English
### Summary
When memory search runs in FTS-only mode (no embedding provider configured), indexing and retrieval become inconsistent in `v2026.3.2`.
Observed behavior:
- `openclaw memory status` can show scan... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T22:03:02Z | 2026-04-27T05:00:11Z | 2026-04-24T20:29:12Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | thrytku | 7,896,461 | MDQ6VXNlcjc4OTY0NjE= | User | false |
openclaw/openclaw | 4,039,676,257 | I_kwDOQb6kR87wyJFh | 39,228 | https://github.com/openclaw/openclaw/issues/39228 | https://api.github.com/repos/openclaw/openclaw/issues/39228 | bug: Docker gateway crash-loop when --bind lan is CLI-only (not in config file) | ## Summary
The Docker gateway container crash-loops with:
```
Gateway failed to start: Error: non-loopback Control UI requires gateway.controlUi.allowedOrigins (set explicit origins), or set gateway.controlUi.dangerouslyAllowHostHeaderOriginFallback=true to use Host-header origin fallback mode
```
This happens when ... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T22:12:18Z | 2026-04-27T05:00:10Z | 2026-04-24T20:29:15Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tosin2013 | 1,975,599 | MDQ6VXNlcjE5NzU1OTk= | User | false |
openclaw/openclaw | 4,039,692,158 | I_kwDOQb6kR87wyM9- | 39,232 | https://github.com/openclaw/openclaw/issues/39232 | https://api.github.com/repos/openclaw/openclaw/issues/39232 | feat: on_raw_inbound plugin hook for passive message observation | ## Problem
There is no plugin hook that fires **before** channel-level filtering (allowlist, mention gating, sender policy). This means plugins can only observe messages that pass all filters and enter the agent pipeline.
**Use case:** A daemon that builds ambient awareness from ALL incoming messages across channels ... | open | null | false | 2 | [] | [] | 2026-03-07T22:19:47Z | 2026-05-04T07:00:32Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | Brotherinlaw-13 | 78,745,996 | MDQ6VXNlcjc4NzQ1OTk2 | User | false |
openclaw/openclaw | 4,039,781,239 | I_kwDOQb6kR87wyit3 | 39,248 | https://github.com/openclaw/openclaw/issues/39248 | https://api.github.com/repos/openclaw/openclaw/issues/39248 | Bug: sandbox.mode: "non-main" silently breaks sessions_spawn subagent initialization | ## Summary
When `agents.defaults.sandbox.mode` is set to `"non-main"`, sub-agents spawned via `sessions_spawn` fail to initialize. Sessions are accepted (return valid `childSessionKey` and `runId`) but never start executing — zero tokens, zero messages, no transcript files created. **No errors appear in gateway logs.*... | open | null | false | 3 | [] | [] | 2026-03-07T22:54:35Z | 2026-05-04T07:00:38Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | alexputici | 222,088,397 | U_kgDODTzMzQ | User | false |
openclaw/openclaw | 4,039,781,726 | I_kwDOQb6kR87wyi1e | 39,249 | https://github.com/openclaw/openclaw/issues/39249 | https://api.github.com/repos/openclaw/openclaw/issues/39249 | [Housekeeping] Cleanup 806 orphan transcript files in duplex agent | ## Overnight Audit Finding
**Date:** 2026-03-07 22:52 UTC
**Agent:** sysadmin (Infrastructure Scout cron)
### Issue
`openclaw doctor` detected 806 orphan transcript files in `~/.openclaw/agents/duplex/sessions/` consuming ~97MB of disk space.
These files are not referenced by `sessions.json` and accumulate over time.... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T22:54:54Z | 2026-04-27T05:00:09Z | 2026-04-24T20:29:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,039,781,920 | I_kwDOQb6kR87wyi4g | 39,250 | https://github.com/openclaw/openclaw/issues/39250 | https://api.github.com/repos/openclaw/openclaw/issues/39250 | [Config] Migrate legacy gateway.bind to new bind mode format | ## Overnight Audit Finding
**Date:** 2026-03-07 22:52 UTC
**Agent:** sysadmin (Infrastructure Scout cron)
### Issue
Config file `/root/.openclaw/openclaw.json` uses legacy `gateway.bind: "0.0.0.0"` format.
`openclaw doctor` reports:
> gateway.bind host aliases (for example 0.0.0.0/localhost) are legacy; use bind mode... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T22:54:56Z | 2026-04-27T05:00:08Z | 2026-04-24T20:29:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,039,800,307 | I_kwDOQb6kR87wynXz | 39,253 | https://github.com/openclaw/openclaw/issues/39253 | https://api.github.com/repos/openclaw/openclaw/issues/39253 | Gemini OpenAI-compatible API regression: provider/model format sent to Google instead of bare model ID | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to 2026.3.2, Gemini 2.0 Flash model fails with 400 error when using OpenAI-compatible API. OpenClaw sends `gemini/gemini-2.0-flash` to Google's API instead of just `gemini-2.0-flash`.
### Steps to reproduce
1. Configure Gemini provider ... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-07T23:03:18Z | 2026-04-27T05:00:08Z | 2026-04-24T20:29:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | weekdaysailor | 47,041,985 | MDQ6VXNlcjQ3MDQxOTg1 | User | false |
openclaw/openclaw | 4,039,801,385 | I_kwDOQb6kR87wynop | 39,254 | https://github.com/openclaw/openclaw/issues/39254 | https://api.github.com/repos/openclaw/openclaw/issues/39254 | fix(ui): agent config page resets to default agent after saving changes | ## Problem
When editing an agent's settings in the web UI (`/agents`), saving any change causes the page to reset back to the default agent instead of staying on the agent you were just editing. This is extremely confusing and error-prone.
## Steps to reproduce
1. Go to `/agents` in the web UI
2. Select a non-defaul... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-07T23:03:54Z | 2026-04-27T05:00:07Z | 2026-04-24T20:29:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | fwends | 4,113,483 | MDQ6VXNlcjQxMTM0ODM= | User | false |
openclaw/openclaw | 4,039,887,888 | I_kwDOQb6kR87wy8wQ | 39,265 | https://github.com/openclaw/openclaw/issues/39265 | https://api.github.com/repos/openclaw/openclaw/issues/39265 | Feature request: compaction.suffix — inject custom text into compaction summaries | ## Problem
When auto-compaction occurs mid-conversation, agents lose context and may respond based on the compaction summary alone without re-reading their memory files. The compaction summary gives agents just enough confidence to think they know what's happening — but they don't.
Currently there's no way to inject ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-07T23:38:50Z | 2026-04-27T05:00:05Z | 2026-04-24T20:29:41Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | clawbert-ct | 262,684,597 | U_kgDOD6g_tQ | User | false |
openclaw/openclaw | 4,039,901,329 | I_kwDOQb6kR87wzACR | 39,268 | https://github.com/openclaw/openclaw/issues/39268 | https://api.github.com/repos/openclaw/openclaw/issues/39268 | [Bug]: Cannot save agent changes | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
If you toggle any setting at all on the agent page in the gateway GUI, the 'save' button stays in a grayed out state and you can't interact with it to save the config.
### Steps to reproduce
1. Change any toggle in agents page.
2. Try and... | closed | completed | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-07T23:45:24Z | 2026-04-27T04:54:56Z | 2026-04-24T22:05:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | linnyexe | 264,726,992 | U_kgDOD8dp0A | User | false |
openclaw/openclaw | 4,039,947,534 | I_kwDOQb6kR87wzLUO | 39,274 | https://github.com/openclaw/openclaw/issues/39274 | https://api.github.com/repos/openclaw/openclaw/issues/39274 | Discord: MESSAGE_CREATE listener times out (80+ seconds) | Environment: OpenClaw 2026.3.2, macOS 25.3.0, Node 25.7.0
Issue: Discord MESSAGE_CREATE events reach the gateway but take 80+ seconds to process, causing timeout.
Steps: Set up Discord bot, mention in guild channel, observe slow listener (86 seconds).
Config has intents.presence and intents.guildMembers enabled. Tel... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-08T00:06:07Z | 2026-04-27T05:00:04Z | 2026-04-24T20:29:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Thesurgicaledit | 265,808,874 | U_kgDOD9fr6g | User | false |
openclaw/openclaw | 4,039,972,795 | I_kwDOQb6kR87wzRe7 | 39,285 | https://github.com/openclaw/openclaw/issues/39285 | https://api.github.com/repos/openclaw/openclaw/issues/39285 | bug: voice-call TTS speed config uses || instead of ?? (same class as #39191) | ## Bug
In `extensions/voice-call/src/providers/tts-openai.ts:85`, the speed default uses `||` instead of `??`:
```typescript
this.speed = config.speed || 1.0;
```
Same class of bug as #39191 and #39190 — `||` treats `0` as falsy, so any falsy numeric value silently becomes the default.
While the current OpenAI API ... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-08T00:19:22Z | 2026-04-27T05:00:02Z | 2026-04-24T20:29:51Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mvanhorn | 455,140 | MDQ6VXNlcjQ1NTE0MA== | User | false |
openclaw/openclaw | 4,039,987,289 | I_kwDOQb6kR87wzVBZ | 39,288 | https://github.com/openclaw/openclaw/issues/39288 | https://api.github.com/repos/openclaw/openclaw/issues/39288 | Discord WebSocket drops every ~15 minutes with code 1006, resume fails with 1005 | ## Bug Report
**OpenClaw version:** 2026.3.2
**OS:** macOS 26.2 (arm64)
**Gateway mode:** local, loopback (127.0.0.1:18789)
## Description
The Discord provider WebSocket connection drops every ~15 minutes throughout the day. After each drop, it attempts a resume with 1000ms backoff but the resume consistently fails ... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-08T00:27:16Z | 2026-04-27T05:00:01Z | 2026-04-24T20:29:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | p19dixon | 153,921,518 | U_kgDOCSyn7g | User | false |
openclaw/openclaw | 4,040,006,976 | I_kwDOQb6kR87wzZ1A | 39,296 | https://github.com/openclaw/openclaw/issues/39296 | https://api.github.com/repos/openclaw/openclaw/issues/39296 | Feature Request: Native Heartbeat Pre-Check / Gatekeeper Pattern | ## Problem
Heartbeat-driven agents waste significant API tokens on empty checks. In a typical setup with 30-minute heartbeats, **80-90% of LLM calls return "nothing to report"** — each costing $1-3+ depending on the model and context window size. At 48 calls/day on a model like Claude Opus, this can reach **$96/day (~... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-08T00:38:15Z | 2026-04-27T05:00:00Z | 2026-04-24T20:30:00Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bo-blue | 257,455,029 | U_kgDOD1hztQ | User | false |
openclaw/openclaw | 4,040,007,713 | I_kwDOQb6kR87wzaAh | 39,297 | https://github.com/openclaw/openclaw/issues/39297 | https://api.github.com/repos/openclaw/openclaw/issues/39297 | Feature Request: Built-in Agent Cost & Usage Analytics | ## Problem
There is no native way to answer basic operational questions:
- How many LLM calls fired today?
- What did they cost?
- Which cron jobs are the most expensive?
- What's my daily/weekly/monthly spend trend?
Agent operators managing costs (especially with expensive models) have to build custom log parsers to... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-08T00:38:37Z | 2026-04-28T04:43:11Z | 2026-04-26T01:59:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bo-blue | 257,455,029 | U_kgDOD1hztQ | User | false |
openclaw/openclaw | 4,040,008,285 | I_kwDOQb6kR87wzaJd | 39,298 | https://github.com/openclaw/openclaw/issues/39298 | https://api.github.com/repos/openclaw/openclaw/issues/39298 | Docs: Document `openclaw system event` as the programmatic agent wake mechanism | When building external scripts that need to programmatically trigger an agent action (e.g., a bash gatekeeper that detects work and needs to wake the LLM), the correct command is:
```bash
openclaw system event --mode now --text "Work detected: 3 unread urgent emails"
```
This works perfectly, but it took significant ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T00:39:00Z | 2026-04-27T04:59:59Z | 2026-04-24T20:30:04Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bo-blue | 257,455,029 | U_kgDOD1hztQ | User | false |
openclaw/openclaw | 4,040,030,171 | I_kwDOQb6kR87wzffb | 39,303 | https://github.com/openclaw/openclaw/issues/39303 | https://api.github.com/repos/openclaw/openclaw/issues/39303 | [Bug]: health-monitor stale-socket breaks webhook mode with self-signed certs | ### Bug type
Regression (worked before, now fails)
### Summary
Version: 2026.3.2 (stable), Node v22.22.1, arm64
Setup: 3 Telegram bots using webhook mode with self-signed cert (direct IP, nginx reverse proxy on port 8443)
Problem: health-monitor fires stale-socket every ~35 min even in webhook mode (no long-pollin... | closed | completed | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-08T00:47:17Z | 2026-04-27T04:59:58Z | 2026-04-24T20:30:08Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wanggold | 2,307,322 | MDQ6VXNlcjIzMDczMjI= | User | false |
openclaw/openclaw | 4,040,038,535 | I_kwDOQb6kR87wzhiH | 39,305 | https://github.com/openclaw/openclaw/issues/39305 | https://api.github.com/repos/openclaw/openclaw/issues/39305 | Feature request: Escalating stall recovery for sub-agents (nudge → kill) | ### Problem
Sub-agents can stall indefinitely while remaining in "running" status. Today the only automated protection is `runTimeoutSeconds` (wall-clock cap), which can't distinguish active work from a stalled agent. Manual detection requires checking `sessions_history` timestamps — unreliable since LLMs have no inter... | open | reopened | false | 2 | [] | [] | 2026-03-08T00:50:29Z | 2026-05-04T07:00:43Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | emcc24444 | 204,263,091 | U_kgDODCzOsw | User | false |
openclaw/openclaw | 4,040,047,013 | I_kwDOQb6kR87wzjml | 39,307 | https://github.com/openclaw/openclaw/issues/39307 | https://api.github.com/repos/openclaw/openclaw/issues/39307 | feat: surface mid-thread context reset/compaction to users | ## Summary
Surface mid-thread compaction/context reset to users and make post-compaction recovery behavior explicit.
## Problem Statement
When an OpenClaw session compacts mid-conversation, the assistant can continue from reduced/refreshed context without any visible signal to the user. On high-stakes analytical threa... | open | null | false | 2 | [] | [] | 2026-03-08T00:53:58Z | 2026-05-04T07:00:45Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | ToneLoke | 10,687,151 | MDQ6VXNlcjEwNjg3MTUx | User | false |
openclaw/openclaw | 4,040,068,896 | I_kwDOQb6kR87wzo8g | 39,316 | https://github.com/openclaw/openclaw/issues/39316 | https://api.github.com/repos/openclaw/openclaw/issues/39316 | [Bug]: codex-cli backend defaults to read-only sandbox, preventing file writes in subagents | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw's built-in `codex-cli` backend default hardcodes Codex to `--sandbox read-only`, which prevents file creation/editing for Codex-backed subagents unless users manually override the backend config.
### Steps to reproduce
1. Configu... | closed | completed | true | 5 | [
"stale"
] | [] | 2026-03-08T01:04:01Z | 2026-04-27T04:59:57Z | 2026-04-24T20:30:11Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sebastianlang84 | 24,494,724 | MDQ6VXNlcjI0NDk0NzI0 | User | false |
openclaw/openclaw | 4,040,068,920 | I_kwDOQb6kR87wzo84 | 39,317 | https://github.com/openclaw/openclaw/issues/39317 | https://api.github.com/repos/openclaw/openclaw/issues/39317 | [Bug]: OpenClaw treats codex exec sandbox-denied runs as success because codex exits 0 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw can treat failed `codex exec --json` runs as successful because Codex may emit sandbox/write-denied failure information in JSONL agent output while still exiting with status code `0`.
### Steps to reproduce
1. Use OpenClaw `2026.... | open | null | false | 5 | [] | [] | 2026-03-08T01:04:02Z | 2026-05-04T07:00:47Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | sebastianlang84 | 24,494,724 | MDQ6VXNlcjI0NDk0NzI0 | User | false |
openclaw/openclaw | 4,040,084,284 | I_kwDOQb6kR87wzss8 | 39,327 | https://github.com/openclaw/openclaw/issues/39327 | https://api.github.com/repos/openclaw/openclaw/issues/39327 | [Bug]: opencode-go/glm-5 returns tool call arguments as string instead of dict | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The opencode-go/glm-5 model intermittently returns tool call arguments as a JSON-encoded string instead of a dict/object. This causes validation errors: `400 Expected dict in tool call arguments, got <class 'str'>`
### Steps to reproduce
1. ... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-08T01:10:47Z | 2026-04-27T04:59:57Z | 2026-04-24T20:30:19Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kent-ai-dev | 229,848,600 | U_kgDODbM2GA | User | false |
openclaw/openclaw | 4,040,100,627 | I_kwDOQb6kR87wzwsT | 39,335 | https://github.com/openclaw/openclaw/issues/39335 | https://api.github.com/repos/openclaw/openclaw/issues/39335 | [BUG] NO_REPLY fragment leaks to webchat UI | ## Bug
When the agent responds with `NO_REPLY` (e.g., after a memory flush prompt), a partial fragment ("NO") sometimes appears in the webchat UI on the *next* user message's response.
## Reproduction Steps
1. System sends a memory flush prompt to agent
2. Agent appends to memory, responds with `NO_REPLY`
3. User se... | closed | completed | true | 6 | [
"stale"
] | [] | 2026-03-08T01:16:54Z | 2026-04-27T04:59:56Z | 2026-04-24T20:30:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | emdash-dot | 265,377,198 | U_kgDOD9FVrg | User | false |
openclaw/openclaw | 4,040,118,918 | I_kwDOQb6kR87wz1KG | 39,341 | https://github.com/openclaw/openclaw/issues/39341 | https://api.github.com/repos/openclaw/openclaw/issues/39341 | [Bug]: Security audit doesn't check top-level channels.discord.allowFrom for slash commands | ## Bug type
Behavior bug
## Summary
Security audit reports "Discord slash commands have no allowlists" warning even when `channels.discord.allowFrom` is configured with a valid user ID.
## Steps to reproduce
1. Configure `channels.discord.allowFrom` with a Discord user ID (e.g., `"allowFrom": ["1231565465132"]`)
2. R... | open | null | false | 4 | [] | [] | 2026-03-08T01:24:57Z | 2026-05-04T07:00:55Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | kalydrae | 8,475,318 | MDQ6VXNlcjg0NzUzMTg= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.