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 | 4,015,684,503 | I_kwDOQb6kR87vWnuX | 32,853 | https://github.com/openclaw/openclaw/issues/32853 | https://api.github.com/repos/openclaw/openclaw/issues/32853 | feat(feishu): skip typing indicator for non-mention messages in groups | ## Problem
When `requireMention: false` is configured for Feishu groups, the bot receives all messages (including those not mentioning the bot). Currently, the typing indicator (emoji reaction) is added to **every** incoming message, even those the bot will ultimately ignore (reply with NO_REPLY/HEARTBEAT_OK).
This c... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T09:42:58Z | 2026-04-17T04:23:35Z | 2026-04-14T04:24:18Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | miaciswang | 5,661,723 | MDQ6VXNlcjU2NjE3MjM= | User | false |
openclaw/openclaw | 4,015,681,894 | I_kwDOQb6kR87vWnFm | 32,852 | https://github.com/openclaw/openclaw/issues/32852 | https://api.github.com/repos/openclaw/openclaw/issues/32852 | [Bug]: Telegram: sendMessageDraft preview renders LTR for RTL languages (Hebrew/Arabic) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Bug Report
**OpenClaw version:** 2026.3.2
**Channel:** Telegram DM
**Language:** Hebrew (likely affects Arabic and other RTL languages too)
### What happened
Since `sendMessageDraft` was introduced for DM streaming in 2026.3.2, streami... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T09:42:25Z | 2026-04-17T04:23:35Z | 2026-04-14T04:24:16Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | taibaran | 6,942,166 | MDQ6VXNlcjY5NDIxNjY= | User | false |
openclaw/openclaw | 4,017,224,000 | I_kwDOQb6kR87vcflA | 33,155 | https://github.com/openclaw/openclaw/issues/33155 | https://api.github.com/repos/openclaw/openclaw/issues/33155 | Allow sessionTarget 'main' for non-default agents in cron jobs | ## Description
When setting up cron jobs for non-default agents, using `sessionTarget: "main"` returns an error:
> `sessionTarget "main" is only valid for the default agent. Use sessionTarget "isolated" with payload.kind "agentTurn" for non-default agents`
This limits multi-agent setups where each agent needs its ow... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T15:13:56Z | 2026-04-20T04:27:03Z | 2026-04-17T04:23:34Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | SammyBuildsIt | 265,308,422 | U_kgDOD9BJBg | User | false |
openclaw/openclaw | 4,015,783,821 | I_kwDOQb6kR87vW_-N | 32,871 | https://github.com/openclaw/openclaw/issues/32871 | https://api.github.com/repos/openclaw/openclaw/issues/32871 | [Bug]: routellm/gpt-5 fails with API format error (input_text vs output_text) | ## Description
When using `routellm/gpt-5` as the model for agents, OpenClaw fails with API format errors. The error has evolved across versions:
- **v2026.3.1:** `Message ordering conflict` — internal OpenClaw bug
- **v2026.3.2:** `HTTP 400: Invalid value: 'input_text'. Supported values are: 'output_text' and 'refus... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:03:43Z | 2026-04-17T04:23:34Z | 2026-04-14T04:24:20Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | ynik76 | 180,348,925 | U_kgDOCr_n_Q | User | false |
openclaw/openclaw | 4,015,897,633 | I_kwDOQb6kR87vXbwh | 32,887 | https://github.com/openclaw/openclaw/issues/32887 | https://api.github.com/repos/openclaw/openclaw/issues/32887 | fix(discord): chunked message sends silently drop on rate-limit mid-sequence | ## Bug
When a long agent response is split into multiple Discord message chunks (>2000 chars), if any chunk hits a 429 rate-limit or 5xx error mid-sequence, the error propagates up from `sendMessageDiscord` and **all remaining chunks are silently dropped**.
The webhook path in `sendDiscordChunkWithFallback` has a `tr... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:27:29Z | 2026-04-17T04:23:33Z | 2026-04-14T04:24:22Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | proto-genesys-x | 260,201,181 | U_kgDOD4Ja3Q | User | false |
openclaw/openclaw | 4,017,222,798 | I_kwDOQb6kR87vcfSO | 33,154 | https://github.com/openclaw/openclaw/issues/33154 | https://api.github.com/repos/openclaw/openclaw/issues/33154 | Telegram getUpdates conflict spam with multiple bot accounts (60K+ events) | ## Description
Running multiple Telegram bot accounts on a single OpenClaw instance generates massive `getUpdates conflict` errors — nearly **60,000 in one week**.
## Environment
- OpenClaw 2026.3.1
- 9 Telegram bot accounts (Fred, Penny, Atlas, Nova, Ace, etc.)
- Each bot bound to a different agent via bindings
- ma... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T15:13:41Z | 2026-04-20T04:27:03Z | 2026-04-17T04:23:32Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | SammyBuildsIt | 265,308,422 | U_kgDOD9BJBg | User | false |
openclaw/openclaw | 4,015,907,240 | I_kwDOQb6kR87vXeGo | 32,889 | https://github.com/openclaw/openclaw/issues/32889 | https://api.github.com/repos/openclaw/openclaw/issues/32889 | [Bug]: EISDIR error when AI tries to read directories during documentation generation | ### Bug type
Regression (worked before, now fails)
### Summary
## Description
When asking openclaw to generate documentation about itself (e.g., "generate... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T10:29:48Z | 2026-04-17T04:23:32Z | 2026-04-14T04:24:25Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | cyzwbincyzwbin | 76,512,254 | MDQ6VXNlcjc2NTEyMjU0 | User | false |
openclaw/openclaw | 4,015,914,835 | I_kwDOQb6kR87vXf9T | 32,892 | https://github.com/openclaw/openclaw/issues/32892 | https://api.github.com/repos/openclaw/openclaw/issues/32892 | [Bug]: models auth login --provider openai-codex fails with No provider plugins found on 2026.3.2 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Version / Environment
• OpenClaw: 2026.3.2
• OS: Linux
• Install path: system (/usr/bin/openclaw)
• Telegram channel plugin enabled (not relevant, but present)
What I expected
openclaw models auth login --provider openai-codex should star... | closed | not_planned | true | 14 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T10:31:41Z | 2026-04-17T04:23:31Z | 2026-04-14T04:24:28Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | GGM141 | 39,202,623 | MDQ6VXNlcjM5MjAyNjIz | User | false |
openclaw/openclaw | 4,015,913,776 | I_kwDOQb6kR87vXfsw | 32,891 | https://github.com/openclaw/openclaw/issues/32891 | https://api.github.com/repos/openclaw/openclaw/issues/32891 | Sub-agent completion coalescing: batch rapid completions into single parent turn | ## Problem
When multiple sub-agents complete in rapid succession (e.g., an "arrow shower" pattern with 5+ parallel dispatches), each completion triggers an independent parent session turn. The parent regenerates the same response for each completion event, producing duplicate messages.
When combined with context over... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:31:26Z | 2026-04-17T04:23:31Z | 2026-04-14T04:24:26Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | karmafeast | 886,771 | MDQ6VXNlcjg4Njc3MQ== | User | false |
openclaw/openclaw | 4,017,205,603 | I_kwDOQb6kR87vcbFj | 33,151 | https://github.com/openclaw/openclaw/issues/33151 | https://api.github.com/repos/openclaw/openclaw/issues/33151 | Gateway doesn't recover after network interruption during restart + hung requests block shutdown | ## Description
Two related issues observed when the gateway restarts during a period of network instability:
### Issue 1: Gateway doesn't self-recover when network returns
After a restart (SIGUSR1 / config.patch) during a network interruption, the gateway fails to recover even after connectivity is restored. Both pr... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T15:10:19Z | 2026-04-20T04:27:04Z | 2026-04-17T04:23:30Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | brockBAG | 264,715,768 | U_kgDOD8c9-A | User | false |
openclaw/openclaw | 4,015,942,210 | I_kwDOQb6kR87vXmpC | 32,900 | https://github.com/openclaw/openclaw/issues/32900 | https://api.github.com/repos/openclaw/openclaw/issues/32900 | [Bug]: ## 问题描述 TUI 运行后终端完全没有输出（没有错误信息，也没有正常显示内容），但消息其实能收到（从日志里能看到 chat.history 请求）。 | ### Bug type
Regression (worked before, now fails)
### Summary
## 问题描述
TUI 运行后终端完全没有输出（没有错误信息，也没有正常显示内容），但消息其实能收到（从日志里能看到 chat.history 请求）。
## 环境
- OpenClaw版本: v2026.3.2
- 操作系统: Windows
- 终端: PowerShell
### Steps to reproduce
## 复现步骤
1. 运行 `openclaw tui`
2. 终端没有任何输出显示
3. 发送消息，从日志看消息实际能收到并处理
### Expected b... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T10:38:22Z | 2026-04-17T04:23:30Z | 2026-04-14T04:24:30Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | cityflying | 55,661,656 | MDQ6VXNlcjU1NjYxNjU2 | User | false |
openclaw/openclaw | 4,015,949,838 | I_kwDOQb6kR87vXogO | 32,902 | https://github.com/openclaw/openclaw/issues/32902 | https://api.github.com/repos/openclaw/openclaw/issues/32902 | [voice-call] Inbound calls accepted but never answered - TwiML response missing | ## Description
Using OpenClaw 2026.3.2 with Twilio provider.
Inbound calls are being accepted by policy (see logs: "Inbound call accepted: policy is open") and call records are created, but the calls never connect - they just ring until the caller hangs up.
## Steps to reproduce
1. Configure voice-call plugin with:... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:40:18Z | 2026-04-17T04:23:29Z | 2026-04-14T04:24:32Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Jragyn | 7,897,461 | MDQ6VXNlcjc4OTc0NjE= | User | false |
openclaw/openclaw | 4,017,196,997 | I_kwDOQb6kR87vcY_F | 33,148 | https://github.com/openclaw/openclaw/issues/33148 | https://api.github.com/repos/openclaw/openclaw/issues/33148 | Isolated cron sessions can overwrite shared files (no per-session write restrictions) | ## Problem
Isolated cron sessions (sessionTarget: isolated) inherit the full AGENTS.md from the main agent, including rules like 'write diary after completing work'. Combined with the write tool (which overwrites entire files), this caused data loss:
- Isolated cron job ran geopolitical topic processing
- AGENTS.md ru... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T15:08:29Z | 2026-04-20T04:27:05Z | 2026-04-17T04:23:28Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | AD207warlord | 255,469,289 | U_kgDODzom6Q | User | false |
openclaw/openclaw | 4,015,952,363 | I_kwDOQb6kR87vXpHr | 32,905 | https://github.com/openclaw/openclaw/issues/32905 | https://api.github.com/repos/openclaw/openclaw/issues/32905 | feature: session-memory hook saves context but never recalls it in new sessions | ## Feature Request
The `session-memory` bundled hook (`src/hooks/bundled/session-memory/handler.ts`) saves session context to `workspace/memory/YYYY-MM-DD-slug.md` when the user runs `/new` or `/reset`. However, there is **no corresponding recall mechanism** — new sessions start with a blank slate, and the saved memor... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:40:53Z | 2026-04-17T04:23:28Z | 2026-04-14T04:24:36Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | velamints2 | 93,711,796 | U_kgDOBZXttA | User | false |
openclaw/openclaw | 4,015,950,544 | I_kwDOQb6kR87vXorQ | 32,903 | https://github.com/openclaw/openclaw/issues/32903 | https://api.github.com/repos/openclaw/openclaw/issues/32903 | bug: dispatch completes with replies=0 — user sees no response (silent drop) | ## Bug Description
When a user sends a message via Feishu (and likely other channels), the dispatch pipeline can complete with `queuedFinal=false` and `counts.final=0`, meaning the agent processed the message but produced **zero replies**. The user receives no response whatsoever and perceives the bot as "stuck" (卡死).... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:40:26Z | 2026-04-17T04:23:28Z | 2026-04-14T04:24:34Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | velamints2 | 93,711,796 | U_kgDOBZXttA | User | false |
openclaw/openclaw | 4,017,191,176 | I_kwDOQb6kR87vcXkI | 33,147 | https://github.com/openclaw/openclaw/issues/33147 | https://api.github.com/repos/openclaw/openclaw/issues/33147 | Health monitor 'stuck' detection should skip sessions with active responses | ## Description
When running multiple Telegram bot accounts (9 in our case), the health monitor detects idle bots as "stuck" and restarts them. However, this restart also kills active bots mid-response, causing replies to silently drop.
## Environment
- OpenClaw 2026.3.1
- macOS (Mac mini M4)
- 9 Telegram bot accounts... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T15:07:21Z | 2026-04-20T04:27:06Z | 2026-04-17T04:23:27Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | SammyBuildsIt | 265,308,422 | U_kgDOD9BJBg | User | false |
openclaw/openclaw | 4,015,974,071 | I_kwDOQb6kR87vXua3 | 32,912 | https://github.com/openclaw/openclaw/issues/32912 | https://api.github.com/repos/openclaw/openclaw/issues/32912 | [Bug]: Groundhog Day with GPT-5.x series. Total amnesia and failure to execute tasks. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using gpt-5.2 in multiple agentic harnesses openclaw, nanobot and other this is a very typical conversation and seems to be more of a training related issue. I have tried change prompts - review immediate context, review available tool... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T10:46:01Z | 2026-04-17T04:23:27Z | 2026-04-14T04:24:39Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | GratefulDave | 10,969,727 | MDQ6VXNlcjEwOTY5NzI3 | User | false |
openclaw/openclaw | 4,016,045,335 | I_kwDOQb6kR87vX_0X | 32,923 | https://github.com/openclaw/openclaw/issues/32923 | https://api.github.com/repos/openclaw/openclaw/issues/32923 | Feishu message reactions support | ## Feature Request: Add message reaction support for Feishu
### Description
Currently the Feishu plugin has im:message.reactions:write_only permission, but no tool to add reactions (emoji) to messages.
### Use Case
让机器人更像真人，可以在用户消息下加合适的emoji反应
### Example
- 用户说"好的" → 机器人加👍
- 用户说"哈哈" → 机器人加😂 | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T11:02:08Z | 2026-04-17T04:23:26Z | 2026-04-14T04:24:43Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kakakiya77 | 250,312,889 | U_kgDODut4uQ | User | false |
openclaw/openclaw | 4,015,978,347 | I_kwDOQb6kR87vXvdr | 32,913 | https://github.com/openclaw/openclaw/issues/32913 | https://api.github.com/repos/openclaw/openclaw/issues/32913 | openclaw status shows stale app version from wizard.lastRunVersion instead of running gateway version | ## Bug
`openclaw status` displays the `app` version field using `wizard.lastRunVersion` from `~/.openclaw/openclaw.json`, not the actual version of the running gateway process.
After updating from 2026.2.26 → 2026.3.2 and fully restarting the gateway (stop + start, new PID), `openclaw status` still shows `app 2026.2.... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T10:47:03Z | 2026-04-17T04:23:26Z | 2026-04-14T04:24:40Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | avelino | 31,996 | MDQ6VXNlcjMxOTk2 | User | false |
openclaw/openclaw | 4,017,179,284 | I_kwDOQb6kR87vcUqU | 33,143 | https://github.com/openclaw/openclaw/issues/33143 | https://api.github.com/repos/openclaw/openclaw/issues/33143 | Session-Start Bug: Empty/Invalid Tool Calls on Multi-Tool Invocations | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
# Title
Session-Start Bug: Empty/Invalid Tool Calls on Multi-Tool Invocations
## Description
### Symptom
During session start or multi-tool invocations, I generate empty/invalid tool calls:
- `Tool not found` → sending empty string `""... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T15:04:49Z | 2026-04-20T04:27:06Z | 2026-04-17T04:23:25Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | stefankubis | 62,284,720 | MDQ6VXNlcjYyMjg0NzIw | User | false |
openclaw/openclaw | 4,016,068,233 | I_kwDOQb6kR87vYFaJ | 32,927 | https://github.com/openclaw/openclaw/issues/32927 | https://api.github.com/repos/openclaw/openclaw/issues/32927 | Feature: globalSkillsPath in openclaw.json for multi-agent shared skills | # Feature Request: `globalSkillsPath` in openclaw.json
## Problem
Currently, skills in OpenClaw are resolved relative to the agent workspace. In multi-agent setups (e.g., 6+ agents), if you want to share a skill across all agents, you must either:
1. Copy the skill to each workspace (maintenance nightmare)
2. Use abs... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T11:07:31Z | 2026-04-17T04:23:25Z | 2026-04-14T04:24:45Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | okjpg | 239,251,923 | U_kgDODkKx0w | User | false |
openclaw/openclaw | 4,016,076,754 | I_kwDOQb6kR87vYHfS | 32,930 | https://github.com/openclaw/openclaw/issues/32930 | https://api.github.com/repos/openclaw/openclaw/issues/32930 | feat(feishu): bot cannot read Cardkit-based interactive card content via im.message.get | ## Problem
When a bot is @mentioned in a Feishu group and the quoted message is a Cardkit-based interactive card (i.e. a streaming card generated by an OpenClaw bot), the bot cannot read the card content — it always receives a minimal render snapshot.
This is distinct from #32023, which covers old-format cards from e... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:09:29Z | 2026-04-17T04:23:24Z | 2026-04-14T04:24:46Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Elarwei001 | 168,552,401 | U_kgDOCgvn0Q | User | false |
openclaw/openclaw | 4,017,153,934 | I_kwDOQb6kR87vcOeO | 33,135 | https://github.com/openclaw/openclaw/issues/33135 | https://api.github.com/repos/openclaw/openclaw/issues/33135 | Feature: expose dispatchAgentTurn in plugin runtime API | ## Summary
Plugins that register custom HTTP webhook endpoints (via `registerHttpRoute`) currently have no way to dispatch an isolated agent turn. This makes it impossible to build webhook-driven automation plugins that need to trigger agent processing — e.g., Gmail automation, CI/CD notification handlers, or any inbo... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T14:59:44Z | 2026-04-20T04:27:07Z | 2026-04-17T04:23:23Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | zhengpu-berkeley | 104,266,940 | U_kgDOBjb8vA | User | false |
openclaw/openclaw | 4,016,112,374 | I_kwDOQb6kR87vYQL2 | 32,936 | https://github.com/openclaw/openclaw/issues/32936 | https://api.github.com/repos/openclaw/openclaw/issues/32936 | [Bug]: Qwen3.5：LM Studio Jinja template error ("No user query found in messages") misclassified as context overflow | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Provider error (Jinja template exception) is misclassified as context overflow, causing misleading error messages and failed auto-compaction loop.
### Steps to reproduce
1. Configure LM Studio as provider with a Qwen3.x model (e.g. unslot... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T11:17:54Z | 2026-04-17T04:23:23Z | 2026-04-14T04:24:50Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Uplay111 | 76,837,767 | MDQ6VXNlcjc2ODM3NzY3 | User | false |
openclaw/openclaw | 4,016,095,985 | I_kwDOQb6kR87vYMLx | 32,933 | https://github.com/openclaw/openclaw/issues/32933 | https://api.github.com/repos/openclaw/openclaw/issues/32933 | [Bug]: gateway restart command loses result in exec tool (timeout issue) | ### Bug type
Regression (worked before, now fails)
### Summary
`openclaw gateway restart` takes 10-15 sec, exec tool times out and loses the result
### Steps to reproduce
1. Run `openclaw gateway restart` via the exec tool in an agent session
2. Wait for the result
3. Observe: tool returns "missing tool result in ... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T11:14:00Z | 2026-04-17T04:23:23Z | 2026-04-14T04:24:48Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | IIIyban | 137,302,231 | U_kgDOCC8Q1w | User | false |
openclaw/openclaw | 4,016,115,574 | I_kwDOQb6kR87vYQ92 | 32,937 | https://github.com/openclaw/openclaw/issues/32937 | https://api.github.com/repos/openclaw/openclaw/issues/32937 | [Bug]: WebUI 404 but telnet ok in version 2026.3.2 | ### Bug type
Regression (worked before, now fails)
### Summary
------------------------------
aaron@ubt:~$ telnet 127.0.0.1 18789
Trying 127.0.0.1...
Connected to 127.0.0.1.
Escape character is '^]'.
------------------------------
aaron@ubt:~$ curl -v http://127.0.0.1:18789
* Trying 127.0.0.1:18789...
* Connect... | closed | not_planned | true | 6 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T11:18:40Z | 2026-04-17T04:23:22Z | 2026-04-14T04:24:52Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | aaronchen2k | 3,222,874 | MDQ6VXNlcjMyMjI4NzQ= | User | false |
openclaw/openclaw | 4,017,144,767 | I_kwDOQb6kR87vcMO_ | 33,133 | https://github.com/openclaw/openclaw/issues/33133 | https://api.github.com/repos/openclaw/openclaw/issues/33133 | MiniMax (Anthropic Messages API) responses not delivered to TUI/WebChat client | ## Bug Description
MiniMax model responses via the `minimax-portal` provider (Anthropic Messages API) are **not delivered to the TUI/WebChat client**, while the same responses are successfully delivered to Telegram.
## Environment
- **OpenClaw version:** 2026.3.2 (85377a2)
- **OS:** macOS 26.3 (arm64), Mac mini M4
- ... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T14:58:01Z | 2026-04-20T04:27:08Z | 2026-04-17T04:23:21Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | alexioklini | 1,140,889 | MDQ6VXNlcjExNDA4ODk= | User | false |
openclaw/openclaw | 4,016,144,184 | I_kwDOQb6kR87vYX84 | 32,940 | https://github.com/openclaw/openclaw/issues/32940 | https://api.github.com/repos/openclaw/openclaw/issues/32940 | feat(feishu): add legacy card format option to make bot responses quotable and forwardable | ## Background
Feishu has two distinct interactive card formats with different tradeoffs:
**Cardkit / streaming cards** (current default):
- ✅ Real-time streaming UX (typewriter effect while generating)
- ❌ Content stored as card_id reference — not readable via `im.message.get`
- ❌ Cannot be forwarded while streaming ... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T11:25:28Z | 2026-04-17T04:23:21Z | 2026-04-14T04:24:54Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Elarwei001 | 168,552,401 | U_kgDOCgvn0Q | User | false |
openclaw/openclaw | 4,016,200,143 | I_kwDOQb6kR87vYlnP | 32,948 | https://github.com/openclaw/openclaw/issues/32948 | https://api.github.com/repos/openclaw/openclaw/issues/32948 | Status line displays incorrect model (shows fallback instead of active primary) | ## Description
The status line displays `google/gemini-3-flash-preview` as the active model for the `main` agent, while Langfuse confirms that actual API calls during the same period used `openrouter/deepseek/deepseek-v3.2` and `openrouter/x-ai/grok-4.1-fast` (the configured primary + fallback).
## Version
OpenClaw ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:37:40Z | 2026-04-17T04:23:20Z | 2026-04-14T04:24:57Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | patbig-afk | 257,161,533 | U_kgDOD1P5PQ | User | false |
openclaw/openclaw | 4,016,184,797 | I_kwDOQb6kR87vYh3d | 32,947 | https://github.com/openclaw/openclaw/issues/32947 | https://api.github.com/repos/openclaw/openclaw/issues/32947 | [Bug]: web_fetch always fails (fetch failed) even for https://example.com, while curl works | ### Bug type
Regression (worked before, now fails)
### Summary
web_fetch always fails (fetch failed) even for https://example.com, while curl works
### Steps to reproduce
Repro steps
Configure web section (openclaw configure --section web) and enable web_fetch.
<img width="863" height="443" alt="Image" src="htt... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T11:33:58Z | 2026-04-17T04:23:20Z | 2026-04-14T04:24:56Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kompunetsa | 93,761,895 | U_kgDOBZaxZw | User | false |
openclaw/openclaw | 4,017,126,316 | I_kwDOQb6kR87vcHus | 33,126 | https://github.com/openclaw/openclaw/issues/33126 | https://api.github.com/repos/openclaw/openclaw/issues/33126 | [BUG] Cron nextRunAtMs calculation error - sets past time instead of next day | ## Description
Cron scheduler incorrectly calculates `nextRunAtMs`: when a task runs past its scheduled time for the day, `nextRunAtMs` is set to the same day's time (in the past) instead of the next day's scheduled time.
## Reproduction Steps
1. Create a daily cron job scheduled for 03:00 UTC (`0 3 * * *`)
2. Let the... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T14:54:16Z | 2026-04-20T04:27:09Z | 2026-04-17T04:23:19Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | 1249663781-design | 264,940,712 | U_kgDOD8qsqA | User | false |
openclaw/openclaw | 4,016,209,462 | I_kwDOQb6kR87vYn42 | 32,949 | https://github.com/openclaw/openclaw/issues/32949 | https://api.github.com/repos/openclaw/openclaw/issues/32949 | Telegram forum: message sent to wrong topic after Anthropic 529 retry | ## Bug Description
After Anthropic API returns 529 ("AI service temporarily overloaded") and OpenClaw retries, the reply message is sent to the wrong Telegram forum topic (or no topic at all). The message gets a valid `message_id` from Telegram (`sendMessage ok`), but it doesn't appear in the expected topic or in the ... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T11:40:05Z | 2026-04-17T04:23:19Z | 2026-04-14T04:24:59Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | AndrewArto | 247,877,121 | U_kgDODsZOAQ | User | false |
openclaw/openclaw | 4,016,210,778 | I_kwDOQb6kR87vYoNa | 32,951 | https://github.com/openclaw/openclaw/issues/32951 | https://api.github.com/repos/openclaw/openclaw/issues/32951 | MCP plugin init timeout hardcoded at 30s — no config to override | ## Description
The MCP `initialize` handshake timeout is hardcoded at 30,000ms. There is no configuration option to increase it. Plugins that need >30s to initialize (e.g., NeuralMemory loading sentence-transformers) always fail on first attempt, though they succeed on retry.
## Environment
- OpenClaw: 2026.3.2
- OS:... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:40:26Z | 2026-04-17T04:23:18Z | 2026-04-14T04:25:03Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | imtylervo | 13,073,663 | MDQ6VXNlcjEzMDczNjYz | User | false |
openclaw/openclaw | 4,016,210,156 | I_kwDOQb6kR87vYoDs | 32,950 | https://github.com/openclaw/openclaw/issues/32950 | https://api.github.com/repos/openclaw/openclaw/issues/32950 | Agent main violates SOUL.md: asks for data already in config + hallucinates requireMention default | ## Description
During Discord onboarding session (2026-03-03), the \`main\` agent (Steve) committed two distinct SOUL.md violations in the same conversation.
## Version
OpenClaw 2026.3.1 (2a8ac97)
## Violation 1 — "Jamais demander ce qui est déjà dans le système"
SOUL.md states explicitly:
> Avant toute question à... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:40:16Z | 2026-04-17T04:23:18Z | 2026-04-14T04:25:01Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | patbig-afk | 257,161,533 | U_kgDOD1P5PQ | User | false |
openclaw/openclaw | 4,017,120,092 | I_kwDOQb6kR87vcGNc | 33,124 | https://github.com/openclaw/openclaw/issues/33124 | https://api.github.com/repos/openclaw/openclaw/issues/33124 | [Bug]: 2026.3.2 doctor does not auto-seed gateway.controlUi.allowedOrigins for loopback deployments | ### Summary
When updating to OpenClaw 2026.3.2, the `openclaw doctor` command does not auto-add the required `gateway.controlUi.allowedOrigins` configuration for existing `loopback` deployments. This causes the Control UI to fail with "Origin localhost:18789 not allowed" even when accessing from the same machine.
###... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T14:52:57Z | 2026-04-20T04:27:09Z | 2026-04-17T04:23:17Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | guppibob-maker | 262,700,594 | U_kgDOD6h-Mg | User | false |
openclaw/openclaw | 4,016,231,394 | I_kwDOQb6kR87vYtPi | 32,952 | https://github.com/openclaw/openclaw/issues/32952 | https://api.github.com/repos/openclaw/openclaw/issues/32952 | [Bug]: openclaw 2026.3.2 require more than 2G for installation | ### Bug type
Regression (worked before, now fails)
### Summary
I'm running openclaw in Digital Ocean droplet with 2GB.
When trying to update to version 2026.3.2 (or installing from scratch), I got a memory error.
### Steps to reproduce
Run `openclaw update` or try to install from scratch
### Expected behavior
U... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T11:45:46Z | 2026-04-17T04:23:17Z | 2026-04-14T04:25:05Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | S1M0N38 | 22,257,750 | MDQ6VXNlcjIyMjU3NzUw | User | false |
openclaw/openclaw | 4,016,235,176 | I_kwDOQb6kR87vYuKo | 32,953 | https://github.com/openclaw/openclaw/issues/32953 | https://api.github.com/repos/openclaw/openclaw/issues/32953 | [Bug] PR #31526 breaks [[reply_to_current]] tag in Feishu DM (regression in 2026.3.2) | ## Bug Description
The `[[reply_to_current]]` tag functionality is broken in Feishu (飞书) direct messages after upgrading from version 2026.3.1 to 2026.3.2.
## Version Information
- **Working version**: 2026.3.1 ✅
- **Broken version**: 2026.3.2 ❌
- **Also tested**: 2026.2.14 ✅ (working)
## Symptoms
When using the `... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:46:43Z | 2026-04-17T04:23:16Z | 2026-04-14T04:25:07Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | AGI-XiaoBai-No1 | 259,090,809 | U_kgDOD3FpeQ | User | false |
openclaw/openclaw | 4,017,106,873 | I_kwDOQb6kR87vcC-5 | 33,122 | https://github.com/openclaw/openclaw/issues/33122 | https://api.github.com/repos/openclaw/openclaw/issues/33122 | WhatsApp DM allowlist silently drops inbound messages from certain international numbers after upgrade to 2026.3.x | ## Description
After upgrading from v2026.2.26 to v2026.3.2, inbound WhatsApp DM messages from a specific international number are silently dropped — no log entry, no error, no "blocked" or "rejected" trace. The same number worked correctly on v2026.2.26.
## Environment
- **From version:** 2026.2.26 (working)
- **To ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T14:50:14Z | 2026-04-20T04:27:10Z | 2026-04-17T04:23:15Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | rocky-balboa-ai | 257,493,578 | U_kgDOD1kKSg | User | false |
openclaw/openclaw | 4,016,236,043 | I_kwDOQb6kR87vYuYL | 32,954 | https://github.com/openclaw/openclaw/issues/32954 | https://api.github.com/repos/openclaw/openclaw/issues/32954 | fromMe propagation + sender identification in DM BodyForAgent | ## Problem
Re-filing of #32061 (closed by stale bot, cannot reopen).
When the agent monitors DM conversations, it cannot distinguish who sent each message because:
1. `msg.key.fromMe` is available at the adapter level but never propagated through the inbound pipeline (`inboundMessage` → `processMessage` → `buildInbo... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T11:46:56Z | 2026-04-17T04:23:15Z | 2026-04-14T04:25:08Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | mactsk | 5,073,360 | MDQ6VXNlcjUwNzMzNjA= | User | false |
openclaw/openclaw | 4,014,868,109 | I_kwDOQb6kR87vTgaN | 32,621 | https://github.com/openclaw/openclaw/issues/32621 | https://api.github.com/repos/openclaw/openclaw/issues/32621 | Feature request: hook point inside message tool (message_sending only catches implicit replies) | ## Problem
The `message_sending` hook currently only intercepts the agent's automatic outbound text — when the agent's reply gets routed to a channel implicitly through the delivery pipeline.
However, when the agent uses the `message(action=send)` tool explicitly (calling the Slack/channel API directly), it **bypasse... | closed | completed | true | 9 | [
"stale"
] | [
"vincentkoc"
] | 2026-03-03T05:58:39Z | 2026-04-28T04:45:39Z | 2026-04-25T23:30:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | duckshrug | 259,036,244 | U_kgDOD3CUVA | User | false |
openclaw/openclaw | 4,021,915,391 | I_kwDOQb6kR87vuY7_ | 34,452 | https://github.com/openclaw/openclaw/issues/34452 | https://api.github.com/repos/openclaw/openclaw/issues/34452 | maskApiKey leaks up to 16 characters of API keys in chat messages | ## Summary
`maskApiKey()` in `src/utils/mask-api-key.ts` exposes up to 16 characters of API keys (first 8 + last 8) when displaying `/models list` output. This masked value is sent as plaintext in chat channels (Telegram, Discord), creating a significant credential exposure risk.
## Affected Code
```typescript
// sr... | closed | completed | true | 6 | [
"stale"
] | [] | 2026-03-04T13:05:52Z | 2026-04-17T04:23:10Z | 2026-04-14T06:16:59Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | okuyam2y | 982,333 | MDQ6VXNlcjk4MjMzMw== | User | false |
openclaw/openclaw | 4,014,589,678 | I_kwDOQb6kR87vScbu | 32,513 | https://github.com/openclaw/openclaw/issues/32513 | https://api.github.com/repos/openclaw/openclaw/issues/32513 | Feature: Auto-retry for agent runs on timeout | ## Problem
When an agent run times out before generating a response, there's currently no automatic retry mechanism. The user has to manually re-run the request.
Error message users see:
```
Request timed out before a response was generated. Please try again, or increase agents.defaults.timeoutSeconds in your config.... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-03-03T04:10:07Z | 2026-04-25T04:19:10Z | 2026-04-22T04:30:16Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | yhdesai-ea | 260,208,258 | U_kgDOD4J2gg | User | false |
openclaw/openclaw | 4,214,502,059 | I_kwDOQb6kR877NDKr | 62,172 | https://github.com/openclaw/openclaw/issues/62172 | https://api.github.com/repos/openclaw/openclaw/issues/62172 | [Bug]: Discord native /status reports discord/undefined and no activity for active channel session | [Bug]: Discord native /status reports discord/undefined and no activity for active channel session
Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
Discord native `/status` in a guild channel reports `discord/undefined`, zero context, and `no activity` for the target chann... | closed | completed | true | 1 | [] | [] | 2026-04-06T23:04:55Z | 2026-04-17T04:23:04Z | 2026-04-14T08:20:30Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | hearts-crypter | 267,817,883 | U_kgDOD_aTmw | User | false |
openclaw/openclaw | 4,197,380,524 | I_kwDOQb6kR876LvGs | 59,943 | https://github.com/openclaw/openclaw/issues/59943 | https://api.github.com/repos/openclaw/openclaw/issues/59943 | image tool returns 'Unknown model' for configured ollama models with image input | ## Bug Description
The built-in `image` tool fails with `Unknown model` error when using ollama models that are configured with `image` input support.
## Environment
- **OpenClaw Version**: 2026.4.2
- **OS**: Linux (Ubuntu)
- **Node**: v25.9.0
## Steps to Reproduce
1. Configure `agents.defaults.imageModel.primary`... | closed | completed | true | 6 | [] | [] | 2026-04-02T23:02:51Z | 2026-04-17T04:23:03Z | 2026-04-14T08:23:54Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | clawd-bot-ai | 269,602,508 | U_kgDOEBHOzA | User | false |
openclaw/openclaw | 4,137,667,576 | I_kwDOQb6kR872n8v4 | 54,629 | https://github.com/openclaw/openclaw/issues/54629 | https://api.github.com/repos/openclaw/openclaw/issues/54629 | [Bug]: Status slash command not displaying data | ### Bug type
Regression (worked before, now fails)
### Summary
The /status slash command in Discord just displays "Done." Now instead of the information it used to display.
### Steps to reproduce
1. Update OpenClaw to latest
2. Set it up with Discord
3. Try running /status in a DM to your OpenClaw bot
### Expecte... | closed | completed | true | 0 | [
"bug",
"regression"
] | [
"vincentkoc"
] | 2026-03-25T18:42:01Z | 2026-04-17T04:23:02Z | 2026-04-14T08:55:04Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | BillChirico | 13,951,316 | MDQ6VXNlcjEzOTUxMzE2 | User | false |
openclaw/openclaw | 4,115,438,877 | I_kwDOQb6kR871TJ0d | 52,162 | https://github.com/openclaw/openclaw/issues/52162 | https://api.github.com/repos/openclaw/openclaw/issues/52162 | [Bug]: memory_search remote embeddings fail with ENOTFOUND when env proxy is configured — withRemoteHttpResponse skips TRUSTED_ENV_PROXY mode | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
`withRemoteHttpResponse()` in the memory remote embeddings path does not check `hasProxyEnvConfigured()` before calling `fetchWithSsrFGuard()`, causing it to default to strict mode which performs local DNS pre-resolution via `resolvePinnedH... | closed | completed | true | 5 | [] | [
"vincentkoc"
] | 2026-03-22T09:19:31Z | 2026-04-17T04:23:01Z | 2026-04-14T09:29:11Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | XIYBHK | 92,884,515 | U_kgDOBYlOIw | User | false |
openclaw/openclaw | 4,040,648,439 | I_kwDOQb6kR87w12b3 | 39,549 | https://github.com/openclaw/openclaw/issues/39549 | https://api.github.com/repos/openclaw/openclaw/issues/39549 | [Bug]: openclaw dashboard device signature bug report | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Control UI fails to connect to local gateway with persistent "device signature invalid" errors when token authentication is enabled, despite device auth flow completing correctly.
### Steps to reproduce
1. Configure gateway with token aut... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-08T06:25:41Z | 2026-04-17T04:22:59Z | 2026-04-14T09:54:11Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nanli2000cn | 259,904,494 | U_kgDOD33T7g | User | false |
openclaw/openclaw | 4,040,438,355 | I_kwDOQb6kR87w1DJT | 39,446 | https://github.com/openclaw/openclaw/issues/39446 | https://api.github.com/repos/openclaw/openclaw/issues/39446 | [Bug]: Gateway crashes with EBUSY errors when .openclaw is synced via cloud storage | ### Bug type
Crash (process/app exits or hangs)
### Summary
OpenClaw Gateway crashes with unhandled `EBUSY`, `EACCES`, or `EPERM` errors when the `.openclaw` configuration directory is synced across devices using cloud storage services (Baidusyncdist, OneDrive, Dropbox, Google Drive).
### Steps to reproduce
1. Con... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-08T04:21:20Z | 2026-04-17T04:22:58Z | 2026-04-14T09:54:38Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nanli2000cn | 259,904,494 | U_kgDOD33T7g | User | false |
openclaw/openclaw | 4,007,698,956 | I_kwDOQb6kR87u4KIM | 30,818 | https://github.com/openclaw/openclaw/issues/30818 | https://api.github.com/repos/openclaw/openclaw/issues/30818 | [Bug]: Web UI chat: Pasted images not sent to backend | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Description
When pasting images in the OpenClaw Web UI chat (Control UI at http://127.0.0.1:18789), the images are displayed as thumbnails in the input field but are not sent to the backend.
### Steps to reproduce
## Steps to Reprod... | closed | not_planned | true | 6 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-01T16:47:45Z | 2026-04-20T04:27:11Z | 2026-04-17T04:22:51Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | MilongWong | 13,269,445 | MDQ6VXNlcjEzMjY5NDQ1 | User | false |
openclaw/openclaw | 4,186,301,600 | I_kwDOQb6kR875heSg | 58,986 | https://github.com/openclaw/openclaw/issues/58986 | https://api.github.com/repos/openclaw/openclaw/issues/58986 | Slash commands return "✅ Done." instead of actual output in Discord | After upgrading to v2026.3.31, Discord slash commands like `/status`, `/help`, `/whoami` etc. return only "✅ Done." as an ephemeral response instead of the actual command output.
**Steps to reproduce:**
1. Use any slash command in Discord (e.g., `/status`)
2. Bot responds with "✅ Done." (ephemeral)
3. Expected: actual... | closed | completed | true | 9 | [] | [] | 2026-04-01T11:49:26Z | 2026-04-17T04:22:50Z | 2026-04-14T13:39:19Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Hansen1018 | 61,605,071 | MDQ6VXNlcjYxNjA1MDcx | User | false |
openclaw/openclaw | 4,218,735,128 | I_kwDOQb6kR877dMoY | 62,569 | https://github.com/openclaw/openclaw/issues/62569 | https://api.github.com/repos/openclaw/openclaw/issues/62569 | [Bug]: toolsAllow on cron agentTurn payload silently dropped — full tool catalog still shipped to model | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
`toolsAllow` set on a cron `agentTurn` payload has no effect at runtime — the model still receives the complete tool catalog (all built-in + all MCP plugin tools) on every run, because `runEmbeddedPiAgent` does... | closed | completed | true | 2 | [] | [] | 2026-04-07T15:24:48Z | 2026-04-17T04:22:47Z | 2026-04-14T16:06:17Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | hexsprite | 23,538 | MDQ6VXNlcjIzNTM4 | User | false |
openclaw/openclaw | 4,007,636,853 | I_kwDOQb6kR87u3691 | 30,784 | https://github.com/openclaw/openclaw/issues/30784 | https://api.github.com/repos/openclaw/openclaw/issues/30784 | hooks: unify internal + plugin hook registries to fix 5 systemic bugs | ## Summary
The codebase has two independent hook systems with overlapping responsibilities and duplicated module-level singleton state. This architectural split has been the root cause of 30+ bugs over the past 2 months.
## Problem to solve
### Two hook systems, no shared infrastructure
1. **Internal hooks** (`src/... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-01T16:23:09Z | 2026-04-25T04:19:16Z | 2026-04-22T04:29:53Z | MEMBER | null | 20260425T060021Z | 2026-04-25T06:00:21Z | mcaxtr | 7,562,095 | MDQ6VXNlcjc1NjIwOTU= | User | false |
openclaw/openclaw | 3,898,181,655 | I_kwDOQb6kR87oWYgX | 9,044 | https://github.com/openclaw/openclaw/issues/9044 | https://api.github.com/repos/openclaw/openclaw/issues/9044 | [Feature Request] TTS post-processing hook for voice modulation | ## Feature Request
Allow users to customize TTS voice characteristics (pitch, speed, effects) through configurable post-processing.
### Use Cases
- **Deeper voices** for robot/assistant personas (e.g., TARS-like)
- **Higher/faster voices** for different character styles
- **Custom audio effects** (reverb, EQ, compres... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-02-04T19:57:16Z | 2026-04-17T04:22:45Z | 2026-04-14T16:40:49Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | robottwo | 1,307,660 | MDQ6VXNlcjEzMDc2NjA= | User | false |
openclaw/openclaw | 4,068,151,890 | I_kwDOQb6kR87yexJS | 44,496 | https://github.com/openclaw/openclaw/issues/44496 | https://api.github.com/repos/openclaw/openclaw/issues/44496 | Image model cannot recognize LM Studio deployed models (qwen3-vl) | ## Bug Description
Image tool cannot recognize LM Studio deployed vision models (e.g., qwen3-vl-30b-a3b-instruct-mlx).
## Steps to Reproduce
1. Configure agents.defaults.imageModel.primary to use LM Studio model
2. Try to use image analysis tool
3. Error: Unknown model anthropic/qwen3-vl-30b-a3b-instruct-mlx
## Roo... | closed | completed | true | 5 | [] | [] | 2026-03-13T01:05:26Z | 2026-04-27T04:48:18Z | 2026-04-24T23:26:13Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | trihook | 5,511,566 | MDQ6VXNlcjU1MTE1NjY= | User | false |
openclaw/openclaw | 4,097,641,742 | I_kwDOQb6kR870PQ0O | 50,017 | https://github.com/openclaw/openclaw/issues/50017 | https://api.github.com/repos/openclaw/openclaw/issues/50017 | [Bug]: Automatic model fallback does not recover cleanly from invalid primary model ID; user-facing HTTP 400 is surfaced instead of silent failover | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When OpenClaw is configured with a primary model plus valid fallback models, and the primary is intentionally or accidentally set to an invalid model ID, OpenClaw surfaces the provider error directly to the user:
HTTP 400: openrouter/inval... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-18T20:50:10Z | 2026-04-17T04:22:42Z | 2026-04-14T17:32:32Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | forrestgdean | 134,644,770 | U_kgDOCAaEIg | User | false |
openclaw/openclaw | 4,006,113,507 | I_kwDOQb6kR87uyHDj | 30,257 | https://github.com/openclaw/openclaw/issues/30257 | https://api.github.com/repos/openclaw/openclaw/issues/30257 | boot-md hook misses gateway:startup event (race condition) | ## Summary
The `boot-md` hook registers for the `gateway:startup` event, but the hook loader completes registration **after** the startup event has already been emitted. The hook never fires.
## Evidence from gateway logs
```
01:05:02.501Z [gateway] listening on ws://0.0.0.0:18789 (PID 454677) ← startup complete
0... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-03-01T03:25:50Z | 2026-04-25T04:19:17Z | 2026-04-22T04:29:49Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | sterling-prog | 261,519,752 | U_kgDOD5Z5iA | User | false |
openclaw/openclaw | 4,005,224,685 | I_kwDOQb6kR87uuuDt | 30,030 | https://github.com/openclaw/openclaw/issues/30030 | https://api.github.com/repos/openclaw/openclaw/issues/30030 | Anthropic API key fallback not triggered when subscription hits rate limit — both providers share cooldown bucket | ## Bug Report
**Version:** OpenClaw 2026.2.26 (bc50708)
## Summary
When the Anthropic subscription (OAuth) hits rate limits, the configured API key fallback (`anthropic-api/claude-opus-4-6`) is never triggered. Instead, the gateway immediately returns `All models failed (1)` — meaning only 1 model was attempted desp... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-28T19:25:11Z | 2026-04-20T04:27:12Z | 2026-04-17T04:22:39Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | finwitz | 14,839,657 | MDQ6VXNlcjE0ODM5NjU3 | User | false |
openclaw/openclaw | 4,207,170,446 | I_kwDOQb6kR876xFOO | 61,281 | https://github.com/openclaw/openclaw/issues/61281 | https://api.github.com/repos/openclaw/openclaw/issues/61281 | Z.AI network_error не триггерит fallback на другую модель | ## Проблема
При ошибке `network_error` от Z.AI fallback на следующую модель не срабатывает.
## Логи
```
[agent/embedded] embedded run agent end: runId=... isError=true model=glm-5 provider=zai error=Provider finish_reason: network_error
```
## Причина
Regex в `failover-matches.ts` ищет:
```js
/\bstop reason:\s*(?:... | closed | completed | true | 1 | [] | [] | 2026-04-05T10:08:54Z | 2026-04-17T04:22:35Z | 2026-04-14T20:51:33Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | oleg2medvedev | 15,245,206 | MDQ6VXNlcjE1MjQ1MjA2 | User | false |
openclaw/openclaw | 4,005,206,229 | I_kwDOQb6kR87uupjV | 30,025 | https://github.com/openclaw/openclaw/issues/30025 | https://api.github.com/repos/openclaw/openclaw/issues/30025 | [Bug]: [agent/embedded] embedded run agent end: runId=7cbbb196-a682-4aad-89fb-39ea01e72afb isError=true error=Connection error. | ### Summary
When I try to chat with openclaw i get this erorr [agent/embedded] embedded run agent end: runId=7cbbb196-a682-4aad-89fb-39ea01e72afb isError=true error=Connection error. And I also try after re install
### Steps to reproduce
idk
### Expected behavior
idk
### Actual behavior
idk
### OpenClaw versio... | closed | not_planned | true | 12 | [
"bug",
"stale"
] | [] | 2026-02-28T19:14:50Z | 2026-04-20T04:27:14Z | 2026-04-17T04:22:35Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | Istiyaq-Khan | 221,598,194 | U_kgDODTVR8g | User | false |
openclaw/openclaw | 3,955,295,607 | I_kwDOQb6kR87rwQV3 | 19,598 | https://github.com/openclaw/openclaw/issues/19598 | https://api.github.com/repos/openclaw/openclaw/issues/19598 | [Feature]: Add Slack Thread Support to OpenClaw Slack Integration | ### Summary
Request to implement thread-related functionality (create, reply, list) in the OpenClaw Slack channel plugin.
### Problem to solve
Currently all replies appear as separate top-level messages in DMs, making conversations disorganized.
### Proposed solution
Implement thread-create, thread-reply, thread-l... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-18T00:43:06Z | 2026-04-17T04:22:35Z | 2026-04-14T21:03:18Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kelfy | 1,072,806 | MDQ6VXNlcjEwNzI4MDY= | User | false |
openclaw/openclaw | 4,004,659,835 | I_kwDOQb6kR87uskJ7 | 29,772 | https://github.com/openclaw/openclaw/issues/29772 | https://api.github.com/repos/openclaw/openclaw/issues/29772 | [Bug]: Hardcoded system prompt conflicts with AGENTS.md | ### Summary
The instructions in the hardcoded system prompt do not match the instructions in `AGENTS.md`.
### Expected behavior
The prompts should not conflict with each other.
### Actual behavior
The system prompt asks for selective memory reading via tool calls, while `AGENTS.md` instructs the agent to read enti... | closed | not_planned | true | 4 | [
"bug",
"stale"
] | [] | 2026-02-28T13:25:51Z | 2026-04-20T04:27:15Z | 2026-04-17T04:22:33Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | jtatum | 340,001 | MDQ6VXNlcjM0MDAwMQ== | User | false |
openclaw/openclaw | 4,165,662,025 | I_kwDOQb6kR874SvVJ | 57,272 | https://github.com/openclaw/openclaw/issues/57272 | https://api.github.com/repos/openclaw/openclaw/issues/57272 | [Bug]: `openclaw status` fails on file-backed SecretRef configs because route/plugin loading re-reads unresolved raw config | # Bug: `openclaw status` fails on file-backed SecretRef configs because route/plugin loading re-reads unresolved raw config
## Summary
When a channel secret has been migrated out of `~/.openclaw/openclaw.json` into a file-backed SecretRef, `openclaw status` can fail before the command reaches its own SecretRef resolu... | closed | completed | true | 3 | [] | [] | 2026-03-29T19:58:23Z | 2026-04-17T04:22:32Z | 2026-04-14T22:59:31Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | laochendeai | 176,562,843 | U_kgDOCoYimw | User | false |
openclaw/openclaw | 4,170,246,037 | I_kwDOQb6kR874kOeV | 57,684 | https://github.com/openclaw/openclaw/issues/57684 | https://api.github.com/repos/openclaw/openclaw/issues/57684 | bug: firecrawl SecretRef env key stays unresolved after secrets reload | ## Summary
`plugins.entries.firecrawl.config.webSearch.apiKey` fails to resolve when configured as SecretRef (`env:default:FIRECRAWL_API_KEY`) even after `openclaw secrets reload`.
## Environment
- OpenClaw: 2026.3.28 (f9b1079)
- Gateway: local/systemd
- Firecrawl plugin enabled
## Repro
1. Configure Firecrawl with S... | closed | completed | true | 1 | [] | [] | 2026-03-30T12:14:11Z | 2026-04-17T04:22:31Z | 2026-04-14T22:59:32Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | ThirteenStrokes | 31,041,297 | MDQ6VXNlcjMxMDQxMjk3 | User | false |
openclaw/openclaw | 4,084,838,869 | I_kwDOQb6kR87zebHV | 48,495 | https://github.com/openclaw/openclaw/issues/48495 | https://api.github.com/repos/openclaw/openclaw/issues/48495 | Feature: per-agent session visibility + agentToAgent.allow should gate reads (list/history) | ## Problem
Currently, `tools.sessions.visibility` and `tools.agentToAgent` are **global-only** settings — they apply identically to all agents on the gateway. There is no way to give one agent broader cross-agent access while keeping others isolated.
Additionally, `tools.agentToAgent.allow` only gates `sessions_send`... | closed | completed | true | 1 | [] | [] | 2026-03-16T21:19:14Z | 2026-04-17T04:22:29Z | 2026-04-15T00:06:07Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | toruvieI | 258,645,309 | U_kgDOD2qdPQ | User | false |
openclaw/openclaw | 4,003,723,931 | I_kwDOQb6kR87uo_qb | 29,367 | https://github.com/openclaw/openclaw/issues/29367 | https://api.github.com/repos/openclaw/openclaw/issues/29367 | BUG: Default subagents inherit parent workspace instead of agents.defaults.workspace | ### Summary
BUG — Subagent Workspace Inheritance
───
When a parent agent has an explicit workspace set in agents.list[], its default child sub-agents (spawned without agentId) inherit the parent's workspace instead of using agents.defaults.workspace. This makes it impossible to isolate default sub-agents from the pa... | closed | not_planned | true | 7 | [
"bug",
"stale"
] | [] | 2026-02-28T03:01:56Z | 2026-04-25T04:19:17Z | 2026-04-22T04:29:44Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,164,129,145 | I_kwDOQb6kR874M5F5 | 57,033 | https://github.com/openclaw/openclaw/issues/57033 | https://api.github.com/repos/openclaw/openclaw/issues/57033 | [Bug]: Native Google provider sends OpenAI-specific `store` field, causing 400 from Gemini API | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using the native Google provider (`google/gemini-2.5-flash`) with OpenClaw's OpenAI-compatible completions path, requests to Google's `generativelanguage.googleapis.com/v1beta/openai/chat/completions` endp... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-29T12:43:34Z | 2026-04-17T04:22:28Z | 2026-04-15T00:43:34Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | jcat59 | 34,663,690 | MDQ6VXNlcjM0NjYzNjkw | User | false |
openclaw/openclaw | 4,228,997,091 | I_kwDOQb6kR878EV_j | 63,500 | https://github.com/openclaw/openclaw/issues/63500 | https://api.github.com/repos/openclaw/openclaw/issues/63500 | [Bug]: HTTP 400 when calling gpt-5.2-codex / gpt-5.3-chat via OpenClaw (Azure OpenAI Responses adapter / reasoning item issue) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When calling gpt-5.2-codex and/or gpt-5.3-chat through OpenClaw, requests fail with HTTP 400. This has been happening consistently and blocks normal usage of these models via OpenClaw (including via the Control... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-09T02:42:06Z | 2026-04-17T04:22:27Z | 2026-04-15T01:04:20Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | zjunsen | 40,018,419 | MDQ6VXNlcjQwMDE4NDE5 | User | false |
openclaw/openclaw | 4,003,716,972 | I_kwDOQb6kR87uo99s | 29,363 | https://github.com/openclaw/openclaw/issues/29363 | https://api.github.com/repos/openclaw/openclaw/issues/29363 | BUG: Subagent sessions inject ALL workspace files instead of AGENTS.md + TOOLS.md only | ### Summary
BUG — Bootstrap File Injection
───
The documentation states in multiple places that sub-agent sessions should only inject AGENTS.md and TOOLS.md into the system prompt context. However, in practice, ALL workspace bootstrap files are injected — including SOUL.md, IDENTITY.md, USER.md, HEARTBEAT.md, and BO... | closed | not_planned | true | 5 | [
"bug",
"stale"
] | [] | 2026-02-28T02:58:22Z | 2026-04-19T04:22:40Z | 2026-04-17T04:22:27Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,051,035,961 | I_kwDOQb6kR87xdec5 | 42,074 | https://github.com/openclaw/openclaw/issues/42074 | https://api.github.com/repos/openclaw/openclaw/issues/42074 | [Bug]: Dashboard inaccessible from macOS | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The dashboard was accessible during initial onboarding on an earlier version (v2.9). After upgrading through 3.9 the dashboard URL became unreachable from the Mac chrome browser. The Telegram bot continues to work perfectly throughout — thi... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T11:07:43Z | 2026-04-17T04:22:24Z | 2026-04-15T02:31:45Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | hckal | 219,231,102 | U_kgDODREzfg | User | false |
openclaw/openclaw | 4,038,037,881 | I_kwDOQb6kR87wr5F5 | 38,717 | https://github.com/openclaw/openclaw/issues/38717 | https://api.github.com/repos/openclaw/openclaw/issues/38717 | Control UI config.set coerces string arrays to numbers, breaking allowFrom | ## Summary
The Control UI's `config.set` WebSocket command re-serializes the entire config JSON and silently coerces string values to numbers when they look numeric. This broke `channels.bluebubbles.allowFrom` by converting phone number strings like `"+12069106512"` to bare integers `12069106512`, causing all inbound ... | closed | completed | true | 0 | [] | [] | 2026-03-07T09:30:14Z | 2026-04-17T04:22:21Z | 2026-04-15T03:12:53Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | omarshahine | 10,343,873 | MDQ6VXNlcjEwMzQzODcz | User | false |
openclaw/openclaw | 3,998,041,848 | I_kwDOQb6kR87uTUb4 | 27,963 | https://github.com/openclaw/openclaw/issues/27963 | https://api.github.com/repos/openclaw/openclaw/issues/27963 | Feature Request: Pre-Send Queue Flush — check for new inbound messages before sending response | ## Problem
When the agent is processing a response (especially longer ones with multiple tool calls), the user may send additional messages that add context, corrections, or follow-up info. Currently, these messages are handled **after** the response is sent:
- **`collect`** mode: queues them for the next turn
- **`s... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-26T22:20:03Z | 2026-04-19T04:22:41Z | 2026-04-17T04:22:21Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | biohazardxxx | 1,370,362 | MDQ6VXNlcjEzNzAzNjI= | User | false |
openclaw/openclaw | 4,027,929,343 | I_kwDOQb6kR87wFVL_ | 36,090 | https://github.com/openclaw/openclaw/issues/36090 | https://api.github.com/repos/openclaw/openclaw/issues/36090 | [Bug]: WhatsApp group reactions silently fail when agent omits participant field | ## Summary
When an agent uses the `message` tool with `action=react` on a WhatsApp **group** message, the reaction silently fails if the `participant` field is omitted. The tool returns `{ ok: true }` but the reaction never appears. No error in logs or tool output. DM reactions work fine.
## Steps to Reproduce
1. Co... | closed | completed | true | 9 | [] | [] | 2026-03-05T12:29:19Z | 2026-04-17T04:22:20Z | 2026-04-15T03:12:54Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | omarshahine | 10,343,873 | MDQ6VXNlcjEwMzQzODcz | User | false |
openclaw/openclaw | 3,997,193,890 | I_kwDOQb6kR87uQFai | 27,843 | https://github.com/openclaw/openclaw/issues/27843 | https://api.github.com/repos/openclaw/openclaw/issues/27843 | Allowlisted commands still trigger approval prompts for complex arguments (2026.2.25) | ## Summary
In 2026.2.25, commands that are explicitly in the exec-approvals allowlist still trigger approval prompts when their arguments contain shell-like characters (backticks, $, backslashes, etc.).
## Steps to reproduce
1. Set `tools.exec.security=allowlist` with `ask=on-miss`
2. Add `/opt/homebrew/bin/gh` to t... | closed | not_planned | true | 13 | [
"stale"
] | [] | 2026-02-26T18:48:54Z | 2026-04-19T04:22:41Z | 2026-04-17T04:22:18Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | xrom2863 | 7,476,098 | MDQ6VXNlcjc0NzYwOTg= | User | false |
openclaw/openclaw | 4,014,586,959 | I_kwDOQb6kR87vSbxP | 32,511 | https://github.com/openclaw/openclaw/issues/32511 | https://api.github.com/repos/openclaw/openclaw/issues/32511 | Feature Request: Discord inline buttons support | ## Feature Request
**Is your feature request related to a problem?**
Currently, inline buttons (interactive message components) are only supported on Telegram. When using the `buttons` parameter with `message send` on Discord, the buttons are silently ignored and only the text message is delivered.
**Describe the sol... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T04:09:12Z | 2026-04-17T04:22:16Z | 2026-04-15T04:20:23Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | otcombo | 8,352,042 | MDQ6VXNlcjgzNTIwNDI= | User | false |
openclaw/openclaw | 4,014,586,972 | I_kwDOQb6kR87vSbxc | 32,512 | https://github.com/openclaw/openclaw/issues/32512 | https://api.github.com/repos/openclaw/openclaw/issues/32512 | feat: add blockInterToolText config to suppress text between tool calls | ## Problem
When models produce text between tool calls during a multi-step response, each text block gets delivered as a separate message to chat channels (Slack, Telegram, Discord, etc.). This "narration leak" clutters conversations with internal reasoning that users don't need to see.
Related issues: #25592, #20005... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T04:09:12Z | 2026-04-17T04:22:15Z | 2026-04-15T04:20:25Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Asclepeion | 169,803,573 | U_kgDOCh7_NQ | User | false |
openclaw/openclaw | 4,014,636,333 | I_kwDOQb6kR87vSn0t | 32,532 | https://github.com/openclaw/openclaw/issues/32532 | https://api.github.com/repos/openclaw/openclaw/issues/32532 | Chrome Extension Relay startup deadlock - browser start requires connected tab before relay starts | ### Bug type
Regression (worked before, now fails)
### Summary
Chrome Extension Browser Relay has a startup deadlock: `openclaw browser start` fails because no tab is connected, but the relay service (port 18792) is never started to allow the extension to connect.
### Steps to reproduce
1. Start OpenClaw gateway: ... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T04:26:50Z | 2026-04-17T04:22:14Z | 2026-04-15T04:20:29Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | yushuaibing9084-oss | 262,818,327 | U_kgDOD6pKFw | User | false |
openclaw/openclaw | 3,988,875,894 | I_kwDOQb6kR87twWp2 | 26,422 | https://github.com/openclaw/openclaw/issues/26422 | https://api.github.com/repos/openclaw/openclaw/issues/26422 | message_sending plugin hook never fires — dead code in all outbound paths | ## Summary
The `message_sending` plugin hook registers successfully via `api.on("message_sending", ...)`, appears in `openclaw hooks list`, and the hook runner has it available (`hasHooks("message_sending")` returns `true`). However, **it never fires** because none of the actual outbound send paths call `deliverOutbou... | closed | completed | true | 12 | [] | [
"vincentkoc"
] | 2026-02-25T11:25:05Z | 2026-04-28T04:45:44Z | 2026-04-25T23:24:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | AndrewAndrewsen | 34,192,400 | MDQ6VXNlcjM0MTkyNDAw | User | false |
openclaw/openclaw | 4,014,638,467 | I_kwDOQb6kR87vSoWD | 32,533 | https://github.com/openclaw/openclaw/issues/32533 | https://api.github.com/repos/openclaw/openclaw/issues/32533 | Fallback does not escalate to different provider on overload errors | ## Bug Report
### Description
When the primary provider (Anthropic) returns 'overloaded' errors, the failover logic retries alternate **profiles** of the same provider instead of escalating to configured **fallback providers** (Google Gemini, OpenAI).
### Configuration
```json
{
"primary": "anthropic/claude-opus-4-... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T04:27:37Z | 2026-04-17T04:22:12Z | 2026-04-15T04:20:31Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | lopponep | 151,858,740 | U_kgDOCQ0uNA | User | false |
openclaw/openclaw | 3,986,633,393 | I_kwDOQb6kR87tnzKx | 25,951 | https://github.com/openclaw/openclaw/issues/25951 | https://api.github.com/repos/openclaw/openclaw/issues/25951 | Sandbox env sanitizer blocks skill primaryEnv vars (e.g. NOTION_API_KEY) | ## Bug Description
Built-in skills that declare `primaryEnv` with names matching the sandbox env sanitizer's blocklist cannot work in sandboxed agents.
**Example:** The bundled `notion` skill declares `primaryEnv: "NOTION_API_KEY"`. When configured via `skills.entries.notion.apiKey`, the host-side `env-overrides.ts` ... | closed | not_planned | true | 11 | [
"stale"
] | [] | 2026-02-25T01:23:38Z | 2026-04-25T04:19:18Z | 2026-04-22T04:29:36Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | hanamizuki | 741,302 | MDQ6VXNlcjc0MTMwMg== | User | false |
openclaw/openclaw | 4,014,639,584 | I_kwDOQb6kR87vSong | 32,535 | https://github.com/openclaw/openclaw/issues/32535 | https://api.github.com/repos/openclaw/openclaw/issues/32535 | Inter-tool-call text blocks leak as separate messages (Telegram streaming preview) | # Issue: Inter-tool-call text blocks leak as separate messages to user (Telegram)
**Repo:** openclaw/openclaw
## Problem
When an LLM response involves multiple tool-use rounds, each intermediate text block between tool calls is delivered as a separate Telegram message. Only the final text should be visible.
### Exa... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T04:28:02Z | 2026-04-17T04:22:11Z | 2026-04-15T04:20:33Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | fridayyi | 258,521,054 | U_kgDOD2i33g | User | false |
openclaw/openclaw | 4,016,289,048 | I_kwDOQb6kR87vY7UY | 32,961 | https://github.com/openclaw/openclaw/issues/32961 | https://api.github.com/repos/openclaw/openclaw/issues/32961 | Graceful gateway shutdown: drain active sessions before restart | ## Problem
When the gateway restarts (`openclaw gateway restart`), all in-flight agent sessions are killed immediately. Any agent mid-task — writing a Linear comment, running a multi-step cron job, composing a Discord message — gets hard-killed with no chance to finish or checkpoint. This causes:
- **Consecutive erro... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T11:59:51Z | 2026-04-17T04:22:10Z | 2026-04-15T04:20:48Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | immeasurablematt | 135,199,143 | U_kgDOCA75pw | User | false |
openclaw/openclaw | 4,016,246,814 | I_kwDOQb6kR87vYxAe | 32,955 | https://github.com/openclaw/openclaw/issues/32955 | https://api.github.com/repos/openclaw/openclaw/issues/32955 | Feature Request: Batch context management (compact/reset) for all agents in a Telegram topic or group | ## Summary
When multiple agents are active in the same Telegram forum topic, there is currently no way to compact or reset all of their sessions at once. Users must manually run `/compact` or `/reset` on each bot individually.
## Use Case
In a multi-agent workflow (e.g., 3 agents collaborating in one Telegram topic)... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T11:49:40Z | 2026-04-17T04:22:10Z | 2026-04-15T04:20:46Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | ai-freer | 194,205,508 | U_kgDOC5NXRA | User | false |
openclaw/openclaw | 3,983,329,458 | I_kwDOQb6kR87tbMiy | 25,371 | https://github.com/openclaw/openclaw/issues/25371 | https://api.github.com/repos/openclaw/openclaw/issues/25371 | [Bug]: OpenRouter 402 billing error misclassified as 'Context overflow', auto-compaction retries drain remaining credits | ### Summary
OpenRouter 402 billing errors ("This request requires more credits, or fewer max_tokens") are misclassified as "Context overflow: prompt too large for the model", triggering futile auto-compaction retries that drain remaining credits faster.
### Steps to reproduce
1. Configure OpenClaw with an OpenRouter... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-24T12:06:44Z | 2026-04-19T04:22:42Z | 2026-04-17T04:22:10Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | marcospgp | 3,464,445 | MDQ6VXNlcjM0NjQ0NDU= | User | false |
openclaw/openclaw | 4,016,296,183 | I_kwDOQb6kR87vY9D3 | 32,964 | https://github.com/openclaw/openclaw/issues/32964 | https://api.github.com/repos/openclaw/openclaw/issues/32964 | Node WebSocket clients receive global agent/chat event broadcasts (conversation content leakage) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
I’m implementing an independent Node client for OpenClaw Gateway, and I noticed that a role: "node" WS connection receives a large amount of unrelated conversation traffic (agent and chat events), including assistant text deltas/finals from... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T12:01:36Z | 2026-04-17T04:22:09Z | 2026-04-15T04:20:50Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | 188080501 | 6,997,074 | MDQ6VXNlcjY5OTcwNzQ= | User | false |
openclaw/openclaw | 4,016,297,093 | I_kwDOQb6kR87vY9SF | 32,966 | https://github.com/openclaw/openclaw/issues/32966 | https://api.github.com/repos/openclaw/openclaw/issues/32966 | feat: pluggable memory_search backend — support external stores (pgvector, etc.) | ## What I'm asking for
A stable interface for skills to register as a `memory_search` backend, so external stores like PostgreSQL + pgvector can replace the default markdown + SQLite setup.
## Background
The [Elite Longterm Memory](https://clawhub.com/skills/elite-longterm-memory) skill uses a `memory-lancedb` plugi... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T12:01:50Z | 2026-04-17T04:22:08Z | 2026-04-15T04:20:53Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | jbushman | 14,206,799 | MDQ6VXNlcjE0MjA2Nzk5 | User | false |
openclaw/openclaw | 4,016,296,514 | I_kwDOQb6kR87vY9JC | 32,965 | https://github.com/openclaw/openclaw/issues/32965 | https://api.github.com/repos/openclaw/openclaw/issues/32965 | memory index --force should reindex session transcripts | ## Description
When using session memory search (`experimental.sessionMemory: true`), `openclaw memory index --force` only reindexes memory files (`MEMORY.md` + `memory/*.md`) but does not force-reindex session transcript files.
## Current Behavior
- `openclaw memory status` shows `sessions: 1/119 files`
- `opencla... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T12:01:40Z | 2026-04-17T04:22:08Z | 2026-04-15T04:20:51Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | L8ton-crypto | 232,345,273 | U_kgDODdlOuQ | User | false |
openclaw/openclaw | 4,016,312,520 | I_kwDOQb6kR87vZBDI | 32,967 | https://github.com/openclaw/openclaw/issues/32967 | https://api.github.com/repos/openclaw/openclaw/issues/32967 | @openclaw/acpx plugin install fails with 404 — scoped package not published to npm | ## Bug
`openclaw update` (and `openclaw plugins install acpx`) fails to install the acpx plugin because `@openclaw/acpx` does not exist on the npm registry:
```
Downloading @openclaw/acpx…
Failed to install acpx: npm pack failed:
npm error code E404
npm error 404 Not Found - GET https://registry.npmjs.org/@openclaw%2... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T12:05:49Z | 2026-04-17T04:22:07Z | 2026-04-15T04:20:55Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | antek-eth | 75,264,212 | MDQ6VXNlcjc1MjY0MjEy | User | false |
openclaw/openclaw | 4,016,360,367 | I_kwDOQb6kR87vZMuv | 32,976 | https://github.com/openclaw/openclaw/issues/32976 | https://api.github.com/repos/openclaw/openclaw/issues/32976 | [Bug]: Webchat: Final reply replaced by tool output when tools are called before responding | ### Bug type
Regression (worked before, now fails)
### Summary
## Description
When an agent calls tools and then sends a final reply in the same turn,
the webchat UI replaces the final reply with the tool output. The user
sees raw tool results instead of the agent's composed response.
## Environment
- Channel: we... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T12:17:00Z | 2026-04-17T04:22:06Z | 2026-04-15T04:20:57Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | greatscottd | 55,542,752 | MDQ6VXNlcjU1NTQyNzUy | User | false |
openclaw/openclaw | 4,016,407,772 | I_kwDOQb6kR87vZYTc | 32,986 | https://github.com/openclaw/openclaw/issues/32986 | https://api.github.com/repos/openclaw/openclaw/issues/32986 | Bug: `openclaw logs --follow` triggers Feishu /open-apis/bot/v3/info probe every second via post-connect health refresh | ## Summary
Running `openclaw logs --follow` (default `--interval 1000`) causes the gateway to call Feishu's `/open-apis/bot/v3/info` API **once per second**, resulting in 30,000+ unnecessary API calls over two days and potential rate-limiting by Feishu.
## Root Cause
The call chain is:
```
openclaw logs --follow (i... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T12:28:21Z | 2026-04-17T04:22:05Z | 2026-04-15T04:21:01Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | wuzhihui1123 | 4,246,626 | MDQ6VXNlcjQyNDY2MjY= | User | false |
openclaw/openclaw | 4,016,368,250 | I_kwDOQb6kR87vZOp6 | 32,977 | https://github.com/openclaw/openclaw/issues/32977 | https://api.github.com/repos/openclaw/openclaw/issues/32977 | test | test | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T12:19:00Z | 2026-04-17T04:22:05Z | 2026-04-15T04:20:59Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | wardcrew-bots[bot] | 265,147,710 | BOT_kgDOD83VPg | Bot | false |
openclaw/openclaw | 4,016,439,515 | I_kwDOQb6kR87vZgDb | 32,990 | https://github.com/openclaw/openclaw/issues/32990 | https://api.github.com/repos/openclaw/openclaw/issues/32990 | BlueBubbles (webhook channel) incorrectly triggers stale-socket health monitor restart during quiet periods | ## Summary
The channel health monitor's stale-socket detection restarts the BlueBubbles provider after 30 minutes of no inbound events. For webhook-based channels like BlueBubbles, this is incorrect behavior — a quiet period simply means no messages were sent, not that the channel is broken.
## Problem
`evaluateChan... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T12:35:31Z | 2026-04-17T04:22:04Z | 2026-04-15T04:21:03Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | amzzzzzzz | 154,392,693 | U_kgDOCTPYdQ | User | false |
openclaw/openclaw | 3,971,344,175 | I_kwDOQb6kR87stecv | 22,438 | https://github.com/openclaw/openclaw/issues/22438 | https://api.github.com/repos/openclaw/openclaw/issues/22438 | feat: Tiered bootstrap file loading for progressive context control | ## Motivation
Bootstrap files consume LLM tokens on every session. For users with large workspaces, loading all files into every session — including sub-agents and cron jobs — wastes context window budget on files the agent never references. This proposal introduces tiered loading so users can control how much context... | open | null | false | 16 | [] | [] | 2026-02-21T05:16:33Z | 2026-05-05T03:10:15Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | 882soft | 14,934,673 | MDQ6VXNlcjE0OTM0Njcz | User | false |
openclaw/openclaw | 4,016,459,736 | I_kwDOQb6kR87vZk_Y | 32,995 | https://github.com/openclaw/openclaw/issues/32995 | https://api.github.com/repos/openclaw/openclaw/issues/32995 | Gateway should truncate raw error payloads before injecting into session context | ## Problem
When the LLM provider returns an error (e.g. Anthropic `overloaded_error`, Cloudflare challenge pages), the raw error payload is injected into the session transcript as-is. This includes:
- Full JSON error objects (`{"type":"error","error":{"type":"overloaded_error",...}}`)
- Cloudflare HTML challenge page... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T12:39:29Z | 2026-04-17T04:22:03Z | 2026-04-15T04:21:05Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | One4AllServices | 263,356,188 | U_kgDOD7J_HA | User | false |
openclaw/openclaw | 3,960,412,480 | I_kwDOQb6kR87sDxlA | 20,478 | https://github.com/openclaw/openclaw/issues/20478 | https://api.github.com/repos/openclaw/openclaw/issues/20478 | Duplicate plugin id warning when channels config coexists with bundled extensions | ## Description
When `channels.bluebubbles` and `channels.feishu` are enabled in `openclaw.json`, the gateway logs a `duplicate plugin id detected` warning on every config reload/heartbeat cycle:
```
[plugins] duplicate plugin id detected; later plugin may be overridden
(/path/to/node_modules/openclaw/extensions/blueb... | closed | completed | true | 11 | [] | [] | 2026-02-19T00:36:16Z | 2026-04-22T04:30:41Z | 2026-04-19T16:16:39Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | liuwenyong1985 | 48,443,240 | MDQ6VXNlcjQ4NDQzMjQw | User | false |
openclaw/openclaw | 3,895,722,653 | I_kwDOQb6kR87oNAKd | 8,731 | https://github.com/openclaw/openclaw/issues/8731 | https://api.github.com/repos/openclaw/openclaw/issues/8731 | [Bug]: OpenClaw bulks messages only after task is done | ## Summary
When chatting with OpenClaw and asking to do a task, instead of replying to me directly and acknowledging the task immediately, it goes "Typing..." for few minutes until it's done with the task and sends in bulk everything that he has been saying all the time... in reverse.
## Steps to reproduce
1. Chat w... | closed | not_planned | true | 14 | [
"bug",
"stale"
] | [] | 2026-02-04T10:08:38Z | 2026-04-19T04:22:44Z | 2026-04-17T04:21:41Z | CONTRIBUTOR | null | 20260419T052035Z | 2026-04-19T05:20:35Z | jabezborja | 64,759,159 | MDQ6VXNlcjY0NzU5MTU5 | User | false |
openclaw/openclaw | 4,143,016,186 | I_kwDOQb6kR8728Wj6 | 55,117 | https://github.com/openclaw/openclaw/issues/55117 | https://api.github.com/repos/openclaw/openclaw/issues/55117 | read 工具读取图片失败 - 模型无法看到图片内容 | ## 问题描述
使用 read 工具读取 JPEG 图片时，模型无法获取图片内容。
## 复现步骤
1. 通过微信发送图片到 OpenClaw
2. 使用 read 工具读取图片文件
3. 模型返回 "image data removed - already processed by model"，无法描述图片内容
## 期望行为
模型应该能够识别图片内容并描述
## 实际行为
只返回 "image data removed - already processed by model"，模型无法识别
## 环境
- OpenClaw 版本：2026.3.8
- 操作系统：macOS 25.3.0 (arm64)
- ... | closed | completed | true | 0 | [] | [] | 2026-03-26T11:33:26Z | 2026-04-19T04:22:45Z | 2026-04-17T04:09:14Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | Fangtianwd | 42,469,250 | MDQ6VXNlcjQyNDY5MjUw | User | false |
openclaw/openclaw | 4,236,258,393 | I_kwDOQb6kR878gCxZ | 64,046 | https://github.com/openclaw/openclaw/issues/64046 | https://api.github.com/repos/openclaw/openclaw/issues/64046 | [Feature]: 希望支持敏感数据脱敏 | ### Summary
1. 当前openclaw.json等配置文件里apikey、token、secretkey等敏感信息都是明文存储
2. openclaw日志、gateway日志中敏感信息未脱敏展示，当前仅支持控制台打印脱敏
3. 自带UI里敏感信息明文展示
<img width="1871" height="537" alt="Image" src="https://github.com/user-attachments/assets/9e26e08a-f06e-494b-9fa9-11e791e02fad" />
综上，希望能够支持敏感数据脱敏
### Problem to solve
能够支持敏感数据脱敏
... | open | null | false | 8 | [
"enhancement"
] | [
"hxy91819"
] | 2026-04-10T03:07:28Z | 2026-05-03T13:21:18Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | NiuDongyao | 39,486,383 | MDQ6VXNlcjM5NDg2Mzgz | User | false |
openclaw/openclaw | 4,204,824,344 | I_kwDOQb6kR876oIcY | 60,858 | https://github.com/openclaw/openclaw/issues/60858 | https://api.github.com/repos/openclaw/openclaw/issues/60858 | Bug: Three `hasRealConversationContent` guards silently block compaction — `session.messages` arrives empty | ## Environment
- OpenClaw 2026.3.13 (61d171a), Homebrew, macOS 15.6 arm64, Node 25.6.0
- Plugin: @martian-engineering/lossless-claw v0.5.3 (contextThreshold: 0.75)
- Agent model: anthropic/claude-opus-4-6 (200k ctx)
- Channel: Telegram
## Problem
Compaction never triggers despite context reaching 287% capacity (2.9M/1... | open | null | false | 4 | [] | [] | 2026-04-04T12:34:26Z | 2026-05-03T06:49:25Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | whe17ch | 62,474,939 | MDQ6VXNlcjYyNDc0OTM5 | User | false |
openclaw/openclaw | 4,188,994,501 | I_kwDOQb6kR875rvvF | 59,198 | https://github.com/openclaw/openclaw/issues/59198 | https://api.github.com/repos/openclaw/openclaw/issues/59198 | [Bug] Telegram getUpdates 409 conflict when multiple bot accounts poll from same gateway | ## Description
When running a single gateway with multiple Telegram bot accounts (8 in our case), the gateway produces persistent `409 Conflict` errors on `getUpdates` — even though every account has a **unique bot token**.
## Environment
- OpenClaw version: `2026.3.31 (213a704)`
- Node.js: `22.22.1`
- OS: Ubuntu Li... | closed | completed | false | 2 | [] | [] | 2026-04-01T18:25:48Z | 2026-04-25T09:30:39Z | 2026-04-25T08:57:04Z | NONE | null | 20260425T120014Z | 2026-04-25T12:00:14Z | mrossit | 207,357,822 | U_kgDODFwHfg | User | false |
openclaw/openclaw | 4,093,913,076 | I_kwDOQb6kR870BCf0 | 49,705 | https://github.com/openclaw/openclaw/issues/49705 | https://api.github.com/repos/openclaw/openclaw/issues/49705 | Chrome 146+ user profile (existing-session) fails with 'Could not find DevToolsActivePort' error | # GitHub Issue: Chrome 146+ user profile (existing-session) cannot connect - DevToolsActivePort not found
## 🐛 Bug Description
**Title:** Chrome 146+ user profile (existing-session) fails with "Could not find DevToolsActivePort" error
**Severity:** High - Blocks real Chrome browser automation
**OpenClaw Version:**... | closed | completed | true | 4 | [] | [] | 2026-03-18T09:34:25Z | 2026-04-27T04:32:13Z | 2026-04-25T04:27:19Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | zhangshu9206 | 12,078,497 | MDQ6VXNlcjEyMDc4NDk3 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.