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 | 4,111,815,804 | I_kwDOQb6kR871FVR8 | 51,510 | https://github.com/openclaw/openclaw/issues/51510 | https://api.github.com/repos/openclaw/openclaw/issues/51510 | [Bug] Model selector UI incorrectly combines provider name with model ID | ## Bug Description
When trying to switch models in the OpenClaw Control UI, an error appears: "Invalid model: xxx is not a valid model".
The UI appears to be incorrectly combining provider names with model IDs. For example:
- Expected model: `custom-coding-dashscope-aliyuncs-com/qwen3-max-2026-01-23`
- UI tries to sa... | closed | completed | true | 3 | [] | [] | 2026-03-21T08:32:53Z | 2026-04-30T04:51:29Z | 2026-04-25T05:28:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | DearPeter | 1,520,265 | MDQ6VXNlcjE1MjAyNjU= | User | false |
openclaw/openclaw | 4,111,837,798 | I_kwDOQb6kR871Fapm | 51,516 | https://github.com/openclaw/openclaw/issues/51516 | https://api.github.com/repos/openclaw/openclaw/issues/51516 | [Bug]: openclaw devices/* and gateway RPC fail locally with pairing/operator scope mismatch on healthy loopback gateway (2026.3.13) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
### Environment
- OpenClaw version: `2026.3.13`
- Install method: global `npm`
- OS: Ubuntu 24.04.4 LTS
- Node: `24.14.0`
- Gateway mode: `local`
- Gateway bind: `loopback`
- Gateway URL: `ws://127.0.0.1:18789`
- Service manager: systemd u... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T08:47:39Z | 2026-04-30T04:51:28Z | 2026-04-25T05:28:37Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | anzulmos | 78,098,735 | MDQ6VXNlcjc4MDk4NzM1 | User | false |
openclaw/openclaw | 4,111,847,216 | I_kwDOQb6kR871Fc8w | 51,517 | https://github.com/openclaw/openclaw/issues/51517 | https://api.github.com/repos/openclaw/openclaw/issues/51517 | [Feature]: WhatsApp read-only / observe mode (dmPolicy: "observe") | ### Summary
Add a new `dmPolicy: "observe"` option that receives inbound messages without sending any agent responses back to the channel.
### Problem to solve
There is no way to passively observe incoming WhatsApp messages without the agent automatically responding. The current dmPolicy options are: pairing/allowli... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-03-21T08:53:08Z | 2026-04-30T04:51:27Z | 2026-04-25T05:28:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cmdrcool | 73,408,635 | MDQ6VXNlcjczNDA4NjM1 | User | false |
openclaw/openclaw | 4,111,849,426 | I_kwDOQb6kR871FdfS | 51,518 | https://github.com/openclaw/openclaw/issues/51518 | https://api.github.com/repos/openclaw/openclaw/issues/51518 | [Feature] Allow subagent.run in agent_end hook (async spawn without request context) | ## Problem
Plugin hooks like `agent_end` fire after the gateway request context has ended. At this point, `api.runtime.subagent.run` throws:
```
Error: Plugin runtime subagent methods are only available during a gateway request.
```
This makes it impossible for plugins to spawn background subagents in response to ag... | closed | completed | true | 2 | [] | [] | 2026-03-21T08:54:13Z | 2026-04-30T04:51:26Z | 2026-04-25T05:28:51Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kagura-agent | 268,167,063 | U_kgDOD_vnlw | User | false |
openclaw/openclaw | 4,111,873,121 | I_kwDOQb6kR871FjRh | 51,522 | https://github.com/openclaw/openclaw/issues/51522 | https://api.github.com/repos/openclaw/openclaw/issues/51522 | Exec approval dialog: no scroll + buttons unreachable when command is too long | ## Bug Report
### Summary
When an exec approval request contains a large payload (e.g. a script written via heredoc or a long command), the approval dialog window does not scroll. The entire command cannot be viewed, and more critically — **the accept/deny buttons are pushed off-screen and become unreachable**. There ... | closed | completed | true | 2 | [] | [] | 2026-03-21T09:05:39Z | 2026-04-30T04:51:25Z | 2026-04-25T05:29:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,111,957,719 | I_kwDOQb6kR871F37X | 51,532 | https://github.com/openclaw/openclaw/issues/51532 | https://api.github.com/repos/openclaw/openclaw/issues/51532 | [Feature]: Allow agent() RPC to honor caller-supplied sessionId when agentId is provided | ### Summary
When calling `agent({ sessionId, agentId })` without `sessionKey`, the caller's `sessionId` is silently discarded because `agentId` triggers automatic sessionKey derivation (`agent:<agentId>:main`), which then overwrites `resolvedSessionId`. This makes it impossible for plugins to manage independent sessio... | closed | completed | true | 2 | [] | [] | 2026-03-21T09:46:16Z | 2026-04-30T04:51:24Z | 2026-04-25T05:29:13Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | xiaohuaxi | 45,277,353 | MDQ6VXNlcjQ1Mjc3MzUz | User | false |
openclaw/openclaw | 4,111,961,066 | I_kwDOQb6kR871F4vq | 51,533 | https://github.com/openclaw/openclaw/issues/51533 | https://api.github.com/repos/openclaw/openclaw/issues/51533 | [Bug]: onboard wizard stores env var reference as apiKey; breaks channels that don't resolve it | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The onboard wizard stores "${MIMO_API_KEY}" (env var reference) in config instead of the resolved value. Webchat resolves it correctly at runtime but Feishu channel passes the literal string to the model API, causing 401 errors.
### Steps ... | closed | completed | true | 3 | [] | [] | 2026-03-21T09:48:42Z | 2026-04-30T04:51:23Z | 2026-04-25T05:29:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yuyzhen | 15,026,920 | MDQ6VXNlcjE1MDI2OTIw | User | false |
openclaw/openclaw | 4,111,980,705 | I_kwDOQb6kR871F9ih | 51,541 | https://github.com/openclaw/openclaw/issues/51541 | https://api.github.com/repos/openclaw/openclaw/issues/51541 | Gemini embedding provider: API key not loaded from .env (400 API_KEY_INVALID) | ## Environment
- **OpenClaw version:** 2026.3.13 (61d171a)
- **OS:** Ubuntu 24.04 LTS
- **Node:** v25.8.1
- **Gateway:** systemd user service
## Description
When configuring `agents.defaults.memorySearch.provider: "gemini"`, the Gemini embedding provider fails with:
```
gemini embeddings failed: 400 {"error":{"code":... | closed | not_planned | true | 4 | [] | [] | 2026-03-21T10:02:32Z | 2026-04-30T04:51:22Z | 2026-04-25T05:29:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lolaopenclaw | 259,362,584 | U_kgDOD3WPGA | User | false |
openclaw/openclaw | 4,111,997,641 | I_kwDOQb6kR871GBrJ | 51,545 | https://github.com/openclaw/openclaw/issues/51545 | https://api.github.com/repos/openclaw/openclaw/issues/51545 | Bug: Sub-agent model override not working in 2026.3.13 | ## Summary
Sub-agents ignore `agents.list[].model` configuration and always fall back to the requester's active model or `agents.defaults.subagents.model`. This appears to be a regression or incomplete fix for issues #6671, #8460.
## Environment
- OpenClaw Version: 2026.3.13 (61d171a)
- OS: macOS 26.3.2 (arm64)
- Node... | closed | completed | true | 3 | [] | [] | 2026-03-21T10:13:33Z | 2026-04-30T04:51:22Z | 2026-04-25T05:29:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | pjbarnett | 77,882,677 | MDQ6VXNlcjc3ODgyNjc3 | User | false |
openclaw/openclaw | 4,112,026,708 | I_kwDOQb6kR871GIxU | 51,551 | https://github.com/openclaw/openclaw/issues/51551 | https://api.github.com/repos/openclaw/openclaw/issues/51551 | Bug: Context Window shows unknown/0 despite extensive conversation | # Bug Report: Context Window Shows 0 Despite Extensive Conversation
## Problem Description
In a Feishu direct message session, `session_status` and `openclaw status` show context window usage as `unknown` or `0`, even after extensive conversation with large file processing.
## Environment
- OpenClaw version: 2026.3... | closed | completed | true | 4 | [] | [] | 2026-03-21T10:30:42Z | 2026-04-30T04:51:21Z | 2026-04-25T05:29:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wl1136 | 27,039,339 | MDQ6VXNlcjI3MDM5MzM5 | User | false |
openclaw/openclaw | 4,112,159,017 | I_kwDOQb6kR871GpEp | 51,564 | https://github.com/openclaw/openclaw/issues/51564 | https://api.github.com/repos/openclaw/openclaw/issues/51564 | LaunchAgent plist not updated on npm update — requires manual `openclaw doctor` after every upgrade | ## Bug Description
After updating OpenClaw via `npm update -g openclaw`, the LaunchAgent plist file (`~/Library/LaunchAgents/ai.openclaw.gateway.plist`) is not automatically updated to reflect the new entry point or version metadata. This causes the gateway service to fail on next restart/reboot.
## Steps to Reproduc... | closed | completed | true | 3 | [] | [] | 2026-03-21T11:21:45Z | 2026-04-30T04:51:20Z | 2026-04-25T05:29:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tarikpath | 227,801,721 | U_kgDODZP6eQ | User | false |
openclaw/openclaw | 4,112,243,603 | I_kwDOQb6kR871G9uT | 51,585 | https://github.com/openclaw/openclaw/issues/51585 | https://api.github.com/repos/openclaw/openclaw/issues/51585 | [求助] EvoMap 注册邀请码 | 我在配置 OpenClaw 的 capability-evolver 技能，需要注册 EvoMap 账号，但注册需要邀请码。
已经通过 API 获取到：
- node_id: node_e630762c5280_defc
- claim_url: https://evomap.ai/claim/88XT-YEQH
请问有人可以分享邀请码吗？或者有其他注册方式？
谢谢！ | closed | not_planned | true | 1 | [] | [] | 2026-03-21T12:01:01Z | 2026-04-30T04:51:19Z | 2026-04-25T05:29:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Qfeng-git | 84,396,374 | MDQ6VXNlcjg0Mzk2Mzc0 | User | false |
openclaw/openclaw | 4,112,339,733 | I_kwDOQb6kR871HVMV | 51,601 | https://github.com/openclaw/openclaw/issues/51601 | https://api.github.com/repos/openclaw/openclaw/issues/51601 | Discord thread sometimes leaks recovery / tool-use content after session reset (with lossless-claw enabled) | ## Summary
In a Discord thread, OpenClaw intermittently sent malformed/recovery-style assistant output into chat instead of a clean final reply.
Observed symptoms included:
- raw or semi-raw structured tool-use content appearing in Discord replies
- recovery-style prose such as `session reset mid-thought`
- mixed mes... | closed | completed | true | 3 | [] | [] | 2026-03-21T12:30:04Z | 2026-04-30T04:51:19Z | 2026-04-25T05:30:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | sidonsoft | 10,795,446 | MDQ6VXNlcjEwNzk1NDQ2 | User | false |
openclaw/openclaw | 4,327,161,547 | I_kwDOQb6kR88AAAABAes-yw | 71,407 | https://github.com/openclaw/openclaw/issues/71407 | https://api.github.com/repos/openclaw/openclaw/issues/71407 | DeepSeek thinking mode + tool calls = HTTP 400: reasoning_content not preserved | ---
# 🐛 DeepSeek thinking mode + tool calls = HTTP 400: `reasoning_content` not preserved
---
## 📋 Environment
| Item | Value |
|---|---|
| **OpenClaw** | `2026.4.15` ~ `2026.4.23` (confirmed unfixed) |
| **Model** | `deepseek-v4-pro` |
| **Config** | `reasoning: true` (thinking mode enabled) |
| **API** | OpenAI... | closed | completed | true | 1 | [] | [] | 2026-04-25T05:16:29Z | 2026-04-30T04:51:16Z | 2026-04-25T05:30:21Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cisheng119 | 113,488,900 | U_kgDOBsO0BA | User | false |
openclaw/openclaw | 4,112,558,300 | I_kwDOQb6kR871IKjc | 51,629 | https://github.com/openclaw/openclaw/issues/51629 | https://api.github.com/repos/openclaw/openclaw/issues/51629 | Gemini 2.5 Flash: model_not_found (empty 404) in agent/embedded runtime despite API key working directly | ## Summary
Herbie (a second OpenClaw instance running on port 18790, using `google/gemini-2.5-flash` as primary model) consistently fails with an empty 404 / `model_not_found` in the agent/embedded runtime. The same API key and model work perfectly when called directly against the Gemini API.
## Environment
- OpenCl... | closed | completed | true | 8 | [] | [] | 2026-03-21T13:37:01Z | 2026-04-30T04:51:15Z | 2026-04-25T05:30:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ramshenoy | 268,190,902 | U_kgDOD_xEtg | User | false |
openclaw/openclaw | 4,112,632,459 | I_kwDOQb6kR871IcqL | 51,632 | https://github.com/openclaw/openclaw/issues/51632 | https://api.github.com/repos/openclaw/openclaw/issues/51632 | openclaw cron run causes gateway closed 1000 error | See attached description | closed | completed | true | 7 | [] | [] | 2026-03-21T14:00:40Z | 2026-04-30T04:51:15Z | 2026-04-25T05:30:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | bxy3045134656 | 183,266,172 | U_kgDOCuxrfA | User | false |
openclaw/openclaw | 4,216,374,864 | I_kwDOQb6kR877UMZQ | 62,390 | https://github.com/openclaw/openclaw/issues/62390 | https://api.github.com/repos/openclaw/openclaw/issues/62390 | [Bug]: buildQualifiedChatModelValue ignores provider when model already contains slash | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## Description
ui/src/ui/chat-model-ref.ts
`buildQualifiedChatModelValue` function returns the model string directly when it contains a slash (`/`), ignoring the provided `provider` parameter.
## Current Behavior
```type... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-07T08:15:02Z | 2026-04-30T04:51:09Z | 2026-04-25T05:33:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | DeepUmbrella | 89,843,059 | MDQ6VXNlcjg5ODQzMDU5 | User | false |
openclaw/openclaw | 4,114,324,081 | I_kwDOQb6kR871O5px | 51,957 | https://github.com/openclaw/openclaw/issues/51957 | https://api.github.com/repos/openclaw/openclaw/issues/51957 | Control UI: Chat model picker uses bare catalog id, resolves to wrong provider (e.g. openai-codex/deepseek-chat) | ## Summary
The web **Control UI** chat **model** `<select>` uses each catalog entry's **bare `id`** as the `<option value>` (e.g. `deepseek-chat`) while the visible label can include the provider (e.g. `deepseek-chat · deepseek`).
When the user chooses a model, the UI calls `sessions.patch` with that bare id. The gat... | closed | completed | true | 4 | [] | [] | 2026-03-22T00:58:05Z | 2026-04-30T04:51:08Z | 2026-04-25T05:33:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | clawdianinan | 263,004,090 | U_kgDOD60fug | User | false |
openclaw/openclaw | 4,115,180,452 | I_kwDOQb6kR871SKuk | 52,127 | https://github.com/openclaw/openclaw/issues/52127 | https://api.github.com/repos/openclaw/openclaw/issues/52127 | [Bug] Control UI 选择 DeepSeek 模型时错误添加 moonshot/ 前缀 | ## 问题描述
在 Control UI 界面选择 DeepSeek 模型时，系统错误地将模型 ID 设置为 `moonshot/deepseek-chat`，导致请求失败。
## 错误信息
```
Failed to set model: GatewayRequestError: model not allowed: moonshot/deepseek-chat
```
## 预期行为
选择 DeepSeek 模型时应该使用正确的 provider 前缀：`deepseek/deepseek-chat`
## 实际行为
Control UI 错误地添加了 `moonshot/` 前缀，导致模型 ID 变成 `moon... | closed | completed | true | 5 | [] | [] | 2026-03-22T07:08:16Z | 2026-04-30T04:51:07Z | 2026-04-25T05:33:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | derekcscn | 269,405,863 | U_kgDOEA7Opw | User | false |
openclaw/openclaw | 4,103,987,025 | I_kwDOQb6kR870nd9R | 50,614 | https://github.com/openclaw/openclaw/issues/50614 | https://api.github.com/repos/openclaw/openclaw/issues/50614 | [Bug]: Dashboard command logs full gateway bearer URLs into read-scoped logs | ## Severity Assessment
### CVSS Assessment
| Metric | v3.1 | v4.0 |
|--------|------|------|
| **Score** | 9.0 / 10.0 | 9.4 / 10.0 |
| **Severity** | Critical | Critical |
| **Vector** | CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H | CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H |
| **Calculator** |... | closed | completed | true | 2 | [] | [] | 2026-03-19T20:09:39Z | 2026-04-30T04:51:07Z | 2026-04-25T05:36:12Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | coygeek | 65,363,919 | MDQ6VXNlcjY1MzYzOTE5 | User | false |
openclaw/openclaw | 4,041,230,262 | I_kwDOQb6kR87w4Ee2 | 39,844 | https://github.com/openclaw/openclaw/issues/39844 | https://api.github.com/repos/openclaw/openclaw/issues/39844 | [Bug] v2026.3.7 后端回归：多个 provider 的 tool calls 和 session 处理出现问题 | ## 摘要
v2026.3.7 版本存在多个后端 regression，影响多个 model provider 的 tool execution 和 session 处理。
## 受影响的 issues
### Tool Execution 问题
- **#39839**: Tool calls broken with kimi-k2.5 (anthropic-messages API) - tool calls 显示为纯文本而不是执行
- **#39788**: read and write tools not executing in v2026.3.7 - 只有 exec 工具正常工作
### Session 稳定性问... | closed | completed | true | 9 | [
"stale"
] | [] | 2026-03-08T12:45:35Z | 2026-04-30T04:50:55Z | 2026-04-25T05:39:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | liangruochong44-ui | 229,361,665 | U_kgDODavIAQ | User | false |
openclaw/openclaw | 4,041,356,204 | I_kwDOQb6kR87w4jOs | 39,914 | https://github.com/openclaw/openclaw/issues/39914 | https://api.github.com/repos/openclaw/openclaw/issues/39914 | iMessage: Configurable polling interval or push-based delivery to reduce response lag | **Channel:** iMessage
**Issue:**
Noticeable polling delay on incoming iMessage delivery. Messages are not received in real time — there is a lag causing delayed responses. On a primary business deployment where iMessage is the main channel, this makes the assistant feel unreliable compared to Telegram or Discord.
**C... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T14:00:45Z | 2026-04-30T04:50:54Z | 2026-04-25T05:39:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dgarf29 | 63,170,957 | MDQ6VXNlcjYzMTcwOTU3 | User | false |
openclaw/openclaw | 4,041,372,280 | I_kwDOQb6kR87w4nJ4 | 39,922 | https://github.com/openclaw/openclaw/issues/39922 | https://api.github.com/repos/openclaw/openclaw/issues/39922 | [Bug]: Agent does not resume in-progress tasks after restart without manual input (v2026.3.7) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Please people HELP xD :
If the agent is actively working on a task and the gateway restarts for any reason (crash, config change, watchdog, OS reboot), the agent comes back online silently and does nothing. The task is lost. The only way ... | closed | completed | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-08T14:10:36Z | 2026-04-30T04:50:54Z | 2026-04-25T05:39:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jokedul | 54,840,321 | MDQ6VXNlcjU0ODQwMzIx | User | false |
openclaw/openclaw | 4,041,497,727 | I_kwDOQb6kR87w5Fx_ | 39,989 | https://github.com/openclaw/openclaw/issues/39989 | https://api.github.com/repos/openclaw/openclaw/issues/39989 | Telegram inbound intermittently dies on 2026.3.2 while outbound still works | ## Summary
Telegram inbound handling is unstable on OpenClaw 2026.3.2: outbound works, but inbound messages intermittently stop being delivered to the agent. The provider often still reports as `Running` / `OK` in Control UI and `openclaw status --deep`.
This reproduces with:
- polling mode
- webhook mode (attempted v... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T15:21:11Z | 2026-04-30T04:50:53Z | 2026-04-25T05:39:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alickfine | 83,772,182 | MDQ6VXNlcjgzNzcyMTgy | User | false |
openclaw/openclaw | 4,041,523,152 | I_kwDOQb6kR87w5L_Q | 40,010 | https://github.com/openclaw/openclaw/issues/40010 | https://api.github.com/repos/openclaw/openclaw/issues/40010 | [Bug]: Feishu sends duplicate streaming cards for tool-using conversations | ## Bug Description
When using the Feishu channel, complex conversations (requiring tool calls/search) send two independent streaming cards instead of one.
## Root Cause
The agent returns **2 payloads** when executing tool calls:
1. **payload[0]**: Initial acknowledgment ("I'll help you... let me search...")
2. **pay... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T15:35:08Z | 2026-04-30T04:50:52Z | 2026-04-25T05:39:43Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mbC6zC | 1,839,207 | MDQ6VXNlcjE4MzkyMDc= | User | false |
openclaw/openclaw | 4,041,529,220 | I_kwDOQb6kR87w5NeE | 40,017 | https://github.com/openclaw/openclaw/issues/40017 | https://api.github.com/repos/openclaw/openclaw/issues/40017 | feat: designate a channel session as the primary/main session | ## Problem
The `agent:main:main` session is a special built-in session used for heartbeats and CLI interactions. There's no way to designate a specific channel session (e.g., a Slack DM) as the primary session instead.
Users who primarily interact via a specific channel (e.g., Slack DM) may want that session to be tre... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T15:38:40Z | 2026-04-30T04:50:51Z | 2026-04-25T05:39:49Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | orzelig | 2,146,989 | MDQ6VXNlcjIxNDY5ODk= | User | false |
openclaw/openclaw | 4,041,756,702 | I_kwDOQb6kR87w6FAe | 40,114 | https://github.com/openclaw/openclaw/issues/40114 | https://api.github.com/repos/openclaw/openclaw/issues/40114 | Feature Request: Dynamic browser profiles for multi-agent isolation | ## Problem
When multiple agents/sessions run concurrently (main session, cron jobs, sub-agents), they all share the single `openclaw` managed browser profile. This causes:
- **Tab hijacking** — one agent navigates a tab another agent is using
- **Cookie/session conflicts** — agent A's login state bleeds into agent B'... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T17:21:24Z | 2026-04-30T04:50:50Z | 2026-04-25T05:39:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | nbschultz97 | 126,931,519 | U_kgDOB5DSPw | User | false |
openclaw/openclaw | 4,041,859,734 | I_kwDOQb6kR87w6eKW | 40,145 | https://github.com/openclaw/openclaw/issues/40145 | https://api.github.com/repos/openclaw/openclaw/issues/40145 | [Bug]: Webchat sessions on 2026.3.7 expose zero tools (no web_search/exec even with tools.allow set) | ### Bug type
Regression (worked before, now fails)
### Summary
Summary: After migrating to a new machine (OpenClaw 2026.3.7, Node 22.22.1), the built‑in webchat/TUI channel no longer
hands the agent any tools—systemPromptReport.tools. entries is always empty and every request falls back to pure LLM
re... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-08T18:10:04Z | 2026-04-30T04:50:49Z | 2026-04-25T05:40:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chrism46220 | 262,900,639 | U_kgDOD6uLnw | User | false |
openclaw/openclaw | 4,041,919,416 | I_kwDOQb6kR87w6su4 | 40,162 | https://github.com/openclaw/openclaw/issues/40162 | https://api.github.com/repos/openclaw/openclaw/issues/40162 | [Bug]: WARNING: Plugin "feishu" contains dangerous code patterns | ### Bug type
Regression (worked before, now fails)
### Summary
Feishu can not installed well, configuration was terminated.
### Steps to reproduce
1. config Channels,
2. select Feishu,
3. Install Feishu plugin
then see this warning.
### Expected behavior
Feishu can be installed well, and work well.
### Actual... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-08T18:40:27Z | 2026-04-30T04:50:49Z | 2026-04-25T05:40:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yesmagic-claw | 266,546,865 | U_kgDOD-MusQ | User | false |
openclaw/openclaw | 4,041,976,066 | I_kwDOQb6kR87w66kC | 40,179 | https://github.com/openclaw/openclaw/issues/40179 | https://api.github.com/repos/openclaw/openclaw/issues/40179 | [Bug]: exec tool does not inherit gateway's environment variables (launchd) | ### Bug type
Regression (worked before, now fails)
### Summary
The exec tool does not properly inherit environment variables set in the launchd plist. Even after a full system reboot, exec commands receive stale/old environment variables instead of the current gateway process environment.
### Steps to reproduce
1... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-08T19:10:56Z | 2026-04-30T04:50:48Z | 2026-04-25T05:40:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tannensj-stack | 266,554,739 | U_kgDOD-NNcw | User | false |
openclaw/openclaw | 4,042,354,960 | I_kwDOQb6kR87w8XEQ | 40,326 | https://github.com/openclaw/openclaw/issues/40326 | https://api.github.com/repos/openclaw/openclaw/issues/40326 | Gateway crash - Doctor warnings being parsed as JSON | ## Description
After upgrading to v2026.3.7, the Gateway crashes when parsing Doctor warnings that contain ASCII box-drawing characters.
## Error
```
Failed to parse JSON message: │ channels.telegram...
SyntaxError: Unexpected token '│'
```
## Steps to reproduce
1. Run `openclaw gateway status`
2. Gateway receives... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-08T22:45:43Z | 2026-04-30T04:50:46Z | 2026-04-25T05:40:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lht0747 | 162,018,876 | U_kgDOCag2PA | User | false |
openclaw/openclaw | 4,204,474,919 | I_kwDOQb6kR876mzIn | 60,800 | https://github.com/openclaw/openclaw/issues/60800 | https://api.github.com/repos/openclaw/openclaw/issues/60800 | Feature: Allow LAN devices to auto-pair without manual approval | ## Problem
Currently, when a device (Android/iOS app) connects to the Gateway via LAN IP (e.g. 192.168.x.x), it requires manual pairing approval via CLI or UI every time the app is reinstalled or the device reconnects. This is inconvenient for personal LAN setups where trust is already established at the network level... | closed | completed | true | 0 | [] | [] | 2026-04-04T10:11:31Z | 2026-04-30T04:50:45Z | 2026-04-25T05:40:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | woaim65 | 98,373,717 | U_kgDOBd0QVQ | User | false |
openclaw/openclaw | 4,112,764,836 | I_kwDOQb6kR871I8-k | 51,661 | https://github.com/openclaw/openclaw/issues/51661 | https://api.github.com/repos/openclaw/openclaw/issues/51661 | [Bug]: Chrome MCP (`--browser-profile user`) does not auto-read `gateway.remote.token`; `open` fails with gateway closed (1000) | ### Bug type
Regression / browser-gateway auth + session stability issue
### Summary
In OpenClaw `2026.3.13`, Chrome MCP existing-session mode (`openclaw browser --browser-profile user`) appears partially broken:
1. `openclaw browser` does **not automatically read** `gateway.remote.token` from config
2. commands on... | closed | completed | true | 3 | [] | [] | 2026-03-21T14:53:42Z | 2026-04-30T04:50:45Z | 2026-04-25T05:40:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tengj | 7,495,019 | MDQ6VXNlcjc0OTUwMTk= | User | false |
openclaw/openclaw | 4,112,791,172 | I_kwDOQb6kR871JDaE | 51,669 | https://github.com/openclaw/openclaw/issues/51669 | https://api.github.com/repos/openclaw/openclaw/issues/51669 | [Bug]: Web chat crashes OpenClaw after sending an image, corrupting session file | ### Bug type
Crash (process/app exits or hangs)
### Summary
After sending an image via the web chat, the OpenClaw gateway crashes with error code=killed, status=4/ILL. Upon subsequent restart attempts, the gateway fails to start until the corrupted session file is manually removed from .openclaw/agents/main/sessions... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-21T15:03:45Z | 2026-04-30T04:50:44Z | 2026-04-25T05:41:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | eventslistener | 29,254,479 | MDQ6VXNlcjI5MjU0NDc5 | User | false |
openclaw/openclaw | 4,112,793,900 | I_kwDOQb6kR871JEEs | 51,670 | https://github.com/openclaw/openclaw/issues/51670 | https://api.github.com/repos/openclaw/openclaw/issues/51670 | Forum topic routing: bindings with peer.id containing topic suffix don't match inbound messages | ## Environment
- OpenClaw: 2026.3.13
- Channel: Telegram
- Group type: Supergroup with Forum/Topics enabled (`is_forum: true`)
- Bot: Admin with all permissions
## Setup
11 agents bound to specific forum topics in a Telegram supergroup:
```json
{
"agentId": "midas",
"match": {
"channel": "telegram",
"peer... | closed | completed | true | 2 | [] | [] | 2026-03-21T15:04:56Z | 2026-04-30T04:50:43Z | 2026-04-25T05:41:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | robertospighel-rgb | 258,860,172 | U_kgDOD23kjA | User | false |
openclaw/openclaw | 4,112,828,950 | I_kwDOQb6kR871JMoW | 51,677 | https://github.com/openclaw/openclaw/issues/51677 | https://api.github.com/repos/openclaw/openclaw/issues/51677 | sessions.json caches stale model after config change — session does not re-resolve on restart | ## Bug Description
When an agent's model is changed in `openclaw.json` config and the gateway is restarted (SIGUSR1), the `sessions.json` registry retains the **old model** for existing sessions. The agent continues trying to use the stale/broken model instead of re-resolving from config.
## Steps to Reproduce
1. Ag... | closed | completed | true | 4 | [] | [] | 2026-03-21T15:21:13Z | 2026-04-30T04:50:43Z | 2026-04-25T05:41:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | rlerikse | 19,270,118 | MDQ6VXNlcjE5MjcwMTE4 | User | false |
openclaw/openclaw | 4,112,830,713 | I_kwDOQb6kR871JND5 | 51,678 | https://github.com/openclaw/openclaw/issues/51678 | https://api.github.com/repos/openclaw/openclaw/issues/51678 | Feature: Option to suppress tool errors from chat | ## Feature Request: Option to suppress tool errors from chat
### Problem
When tools like `web_fetch`, `web_search`, `exec` fail, the error message is automatically sent to the user chat (WhatsApp, Telegram, etc.).
This creates noise for end users who don't need to see technical errors like:
- `Web fetch failed (403):... | closed | completed | true | 2 | [] | [] | 2026-03-21T15:22:08Z | 2026-04-30T04:50:42Z | 2026-04-25T05:41:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lua-executive | 263,506,981 | U_kgDOD7TMJQ | User | false |
openclaw/openclaw | 4,112,850,214 | I_kwDOQb6kR871JR0m | 51,685 | https://github.com/openclaw/openclaw/issues/51685 | https://api.github.com/repos/openclaw/openclaw/issues/51685 | [Bug]: Control UI freezes with high CPU when switching sessions via dropdown menu | ### Bug type
Crash (process/app exits or hangs)
### Summary
Clicking the session dropdown menu in Control UI and switching from "main" to "qqbot" causes the frontend to freeze with high CPU usage (fans spinning loudly). No console errors are logged.
### Steps to reproduce
Open Control UI at http://127.0.0.1:18789/... | closed | completed | true | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-21T15:31:33Z | 2026-04-30T04:50:41Z | 2026-04-25T05:41:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | liuxiao128 | 73,699,017 | MDQ6VXNlcjczNjk5MDE3 | User | false |
openclaw/openclaw | 4,112,871,141 | I_kwDOQb6kR871JW7l | 51,693 | https://github.com/openclaw/openclaw/issues/51693 | https://api.github.com/repos/openclaw/openclaw/issues/51693 | Request: configure ACP default agent / clarify sessions_spawn(acp) setup | When calling sessions_spawn with runtime="acp" from OpenClaw main agent, it fails with:
> ACP target agent is not configured. Pass agentId in sessions_spawn or set acp.defaultAgent in config.
On systems where users primarily run Codex, it would be helpful to:
- Document a recommended default (e.g. acp.defaultAgent="c... | closed | completed | true | 2 | [] | [] | 2026-03-21T15:41:07Z | 2026-04-30T04:50:40Z | 2026-04-25T05:41:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gohnjanotis | 410,024 | MDQ6VXNlcjQxMDAyNA== | User | false |
openclaw/openclaw | 4,112,890,302 | I_kwDOQb6kR871Jbm- | 51,698 | https://github.com/openclaw/openclaw/issues/51698 | https://api.github.com/repos/openclaw/openclaw/issues/51698 | gateway probe WebSocket handshake timeout but HTTP works fine |
## Bug Description
openclaw gateway probe shows 'Reachable: no' with WebSocket timeout, but:
- HTTP access works: curl http://127.0.0.1:18789/ returns 200
- Health endpoint works: /health returns {ok:true,status:live}
- Gateway service is running normally
## Environment
- OpenClaw: 2026.3.13
- Node: v25.8.1
## Logs
... | closed | completed | true | 2 | [] | [] | 2026-03-21T15:49:12Z | 2026-04-30T04:50:39Z | 2026-04-25T05:41:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ngxaix | 267,099,503 | U_kgDOD-udbw | User | false |
openclaw/openclaw | 4,112,965,841 | I_kwDOQb6kR871JuDR | 51,724 | https://github.com/openclaw/openclaw/issues/51724 | https://api.github.com/repos/openclaw/openclaw/issues/51724 | sessions_send 触发的子智能体任务完成后 auto-announce 未正常触发 | ## 问题描述
在 Telegram 渠道中，通过 `sessions_send` 向子智能体发送任务后，子智能体完成任务但未自动向主智能体汇报（auto-announce 未触发）。
## 复现步骤
1. 在 Telegram 渠道创建子智能体：`sessions_spawn`
2. 子智能体完成初始任务并正常汇报
3. 主智能体通过 `sessions_send` 向子智能体发送新任务
4. 子智能体完成任务（`sessions_history` 可见完整输出）
5. **问题**：子智能体未自动汇报结果给主智能体
## 预期行为
子智能体任务完成后，应自动向主智能体汇报结果（auto-announce）。
## 实... | closed | completed | true | 3 | [] | [] | 2026-03-21T16:20:58Z | 2026-04-30T04:50:38Z | 2026-04-25T05:41:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangweiweb3 | 174,498,242 | U_kgDOCmahwg | User | false |
openclaw/openclaw | 4,112,989,823 | I_kwDOQb6kR871Jz5_ | 51,736 | https://github.com/openclaw/openclaw/issues/51736 | https://api.github.com/repos/openclaw/openclaw/issues/51736 | [Mattermost] Markdown formatting lost in thread replies when message is chunked | ## Bug Description
When OpenClaw sends a reply to a Mattermost thread, Markdown formatting (including newlines) appears to be lost when the message is split into multiple chunks.
## Steps to Reproduce
1. Configure Mattermost channel in OpenClaw
2. Send a message in a thread that triggers a response
3. When the respo... | closed | completed | true | 3 | [] | [] | 2026-03-21T16:31:49Z | 2026-04-30T04:50:37Z | 2026-04-25T05:42:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lemoncat7 | 59,995,349 | MDQ6VXNlcjU5OTk1MzQ5 | User | false |
openclaw/openclaw | 4,113,000,791 | I_kwDOQb6kR871J2lX | 51,742 | https://github.com/openclaw/openclaw/issues/51742 | https://api.github.com/repos/openclaw/openclaw/issues/51742 | [Bug]: Openclaw not passing tools definition to nemotron | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Using nemotron 3 nono in LM Studio 18GB with openclaw, logic works, messaging works, but it can't seem to instantiate tools.
### Steps to reproduce
Macos
Latest openclaw
LM Studio
Nemotron nano 3 18GB mlx
-- it appears nemotron just do... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T16:36:54Z | 2026-04-30T04:50:36Z | 2026-04-25T05:42:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | styles01 | 11,393,400 | MDQ6VXNlcjExMzkzNDAw | User | false |
openclaw/openclaw | 4,113,001,310 | I_kwDOQb6kR871J2te | 51,743 | https://github.com/openclaw/openclaw/issues/51743 | https://api.github.com/repos/openclaw/openclaw/issues/51743 | [Bug]: Google provider: usageMetadata not mapped to usage.input/output — all Gemini calls record 0 tokens | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
<p class="font-claude-response-body break-words whitespace-normal leading-[1.7]">When calling <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">goog... | closed | completed | true | 4 | [
"bug",
"bug:behavior"
] | [
"vincentkoc"
] | 2026-03-21T16:37:09Z | 2026-04-30T04:50:36Z | 2026-04-25T05:42:15Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jcat59 | 34,663,690 | MDQ6VXNlcjM0NjYzNjkw | User | false |
openclaw/openclaw | 4,113,033,614 | I_kwDOQb6kR871J-mO | 51,755 | https://github.com/openclaw/openclaw/issues/51755 | https://api.github.com/repos/openclaw/openclaw/issues/51755 | Mattermost: Markdown formatting lost when replyToMode is set to "all" | ## Bug Description
When using the Mattermost integration with `replyToMode: "all"`, Markdown formatting (bold, italic, code blocks, lists, etc.) is lost and messages are sent as plain text.
## Steps to Reproduce
1. Configure Mattermost channel with `replyToMode: "all"`
2. Send a message containing Markdown formattin... | closed | completed | true | 3 | [] | [] | 2026-03-21T16:50:38Z | 2026-04-30T04:50:35Z | 2026-04-25T05:42:21Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lemoncat7 | 59,995,349 | MDQ6VXNlcjU5OTk1MzQ5 | User | false |
openclaw/openclaw | 4,113,079,970 | I_kwDOQb6kR871KJ6i | 51,771 | https://github.com/openclaw/openclaw/issues/51771 | https://api.github.com/repos/openclaw/openclaw/issues/51771 | Discord gateway WebSocket drops every ~10min - intents:content=limited treated as blocker | ## Summary
Discord gateway WebSocket connection drops every ~10 minutes and never maintains `gatewayConnected=true` state, despite all three Privileged Gateway Intents (Presence, Server Members, Message Content) being enabled in the Discord Developer Portal.
## Environment
- OpenClaw version: 2026.3.13 (61d171a)
- O... | closed | completed | true | 2 | [] | [] | 2026-03-21T17:12:09Z | 2026-04-30T04:50:34Z | 2026-04-25T05:42:28Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | maxfieldpm | 269,183,360 | U_kgDOEAtpgA | User | false |
openclaw/openclaw | 4,113,083,025 | I_kwDOQb6kR871KKqR | 51,773 | https://github.com/openclaw/openclaw/issues/51773 | https://api.github.com/repos/openclaw/openclaw/issues/51773 | [Bug]: Slack inbound attachments blocked by path security policy — image analysis fails | ### Bug type
Regression (worked before, now fails)
### Summary
When a user uploads an image via Slack (e.g., from mobile), OpenClaw downloads the attachment to a temp directory (`/var/folders/...`) that is not whitelisted for the image tool. Attempting to analyze the image fails with:
```
Local media path is not un... | closed | completed | true | 3 | [] | [] | 2026-03-21T17:13:32Z | 2026-04-30T04:50:32Z | 2026-04-25T05:42:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | integratedlifeorchestrator-ai | 267,233,091 | U_kgDOD-2nQw | User | false |
openclaw/openclaw | 4,113,106,451 | I_kwDOQb6kR871KQYT | 51,779 | https://github.com/openclaw/openclaw/issues/51779 | https://api.github.com/repos/openclaw/openclaw/issues/51779 | openclaw devices list/approve does not use cached paired operator device token, causing missing operator.read | ## Summary
`openclaw devices list` and `openclaw devices approve <requestId>` fail in a deployment where device pairing itself works correctly via:
- Telegram `/pair approve`
- Control UI browser pairing flow
The CLI can reach the local Gateway, but device-management commands fail with missing operator scopes.
## E... | closed | completed | true | 2 | [] | [] | 2026-03-21T17:23:44Z | 2026-04-30T04:50:31Z | 2026-04-25T05:42:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jolinliwip101-png | 269,974,573 | U_kgDOEBd8LQ | User | false |
openclaw/openclaw | 4,113,127,333 | I_kwDOQb6kR871KVel | 51,785 | https://github.com/openclaw/openclaw/issues/51785 | https://api.github.com/repos/openclaw/openclaw/issues/51785 | WhatsApp outbound send to new contact fails despite linked session | ## Bug report — WhatsApp outbound send to new contact fails despite linked session
**Context**
- OpenClaw version: `2026.3.13`
- OS: Windows 11
- Gateway local in loopback mode
- WhatsApp account: `default`
- inbound WhatsApp works with my own number
- goal: send an outbound WhatsApp message to a new contact (`+336959... | closed | completed | true | 2 | [] | [] | 2026-03-21T17:30:50Z | 2026-04-30T04:50:31Z | 2026-04-25T05:42:47Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kraffes | 6,930,081 | MDQ6VXNlcjY5MzAwODE= | User | false |
openclaw/openclaw | 4,113,155,918 | I_kwDOQb6kR871KcdO | 51,787 | https://github.com/openclaw/openclaw/issues/51787 | https://api.github.com/repos/openclaw/openclaw/issues/51787 | [Question] Is there interest in a managed hosting option for OpenClaw? | Hi, I've been working on a managed hosting service for OpenClaw ([openclawup.com](https://openclawup.com)) that handles all the server setup, AI API keys, and deployment automatically.
Before I invest more time, I wanted to check with this community:
- Would a managed hosting option be useful for you or your team?
- W... | closed | not_planned | true | 2 | [] | [] | 2026-03-21T17:40:35Z | 2026-04-30T04:50:30Z | 2026-04-25T05:42:53Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wolfqing | 24,562,430 | MDQ6VXNlcjI0NTYyNDMw | User | false |
openclaw/openclaw | 4,113,161,045 | I_kwDOQb6kR871KdtV | 51,788 | https://github.com/openclaw/openclaw/issues/51788 | https://api.github.com/repos/openclaw/openclaw/issues/51788 | Bug: Control UI prefixes xAI model IDs with 'anthropic/' causing GatewayRequestError | ## Summary
When switching models via the Control UI (dropdown or `/model` command), xAI model IDs are sent to the gateway with an incorrect `anthropic/` prefix instead of `xai/`. This causes a `GatewayRequestError: model not allowed` for every xAI model.
## Environment
- **Version:** OpenClaw 2026.3.13 (61d171a)
- *... | closed | completed | true | 3 | [] | [] | 2026-03-21T17:42:14Z | 2026-04-30T04:50:29Z | 2026-04-25T05:43:00Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | neuvalbe | 241,164,360 | U_kgDODl_gSA | User | false |
openclaw/openclaw | 4,113,229,775 | I_kwDOQb6kR871KufP | 51,796 | https://github.com/openclaw/openclaw/issues/51796 | https://api.github.com/repos/openclaw/openclaw/issues/51796 | Inkonsistentes Verhalten bei exec Genehmigungs-Popups in OpenClaw Control UI | Hallo OpenClaw-Team,
ich möchte ein inkonsistentes Verhalten bei den `exec`-Genehmigungs-Popups in der OpenClaw Control UI melden, das wir bei umfangreichen Tests reproduzieren konnten.
**Problembeschreibung:**
Obwohl `tools.exec.ask` in der `openclaw.json` auf `"always"` gesetzt ist, erscheinen Genehmigungs-Popups (... | closed | completed | true | 2 | [] | [] | 2026-03-21T18:04:11Z | 2026-04-30T04:50:28Z | 2026-04-25T05:43:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Peilsender | 269,987,345 | U_kgDOEBeuEQ | User | false |
openclaw/openclaw | 4,113,233,925 | I_kwDOQb6kR871KvgF | 51,797 | https://github.com/openclaw/openclaw/issues/51797 | https://api.github.com/repos/openclaw/openclaw/issues/51797 | Bug: Exec命令在Windows下参数空格丢失导致命令失败 | ## 🐛 Bug描述
OpenClaw在执行Windows命令时，会出现命令参数空格丢失的问题，导致命令无法正确解析和执行。
### 复现步骤
1. 使用OpenClaw的exec工具执行Windows命令
2. 命令包含Windows参数格式（如
/<span class="hl-type">PID</span>
,
/<span class="hl-type">IM</span>
,
/<span class="hl-type">F</span>
）
3. 观察到参数与值之间缺少空格
### 示例
**期望命令:**一大段文本） | closed | not_planned | true | 3 | [] | [] | 2026-03-21T18:05:30Z | 2026-04-30T04:50:27Z | 2026-04-25T05:43:13Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | zy00010 | 265,507,976 | U_kgDOD9NUiA | User | false |
openclaw/openclaw | 4,113,234,250 | I_kwDOQb6kR871KvlK | 51,798 | https://github.com/openclaw/openclaw/issues/51798 | https://api.github.com/repos/openclaw/openclaw/issues/51798 | Bug: image tool returns 'Model does not support images' for ollama/qwen3-vl:2b despite VL model working via direct API | **OpenClaw Version:** 2026.3.13 (61d171a)
**Config:** agents.defaults.imageModel.primary = \"ollama/qwen3-vl:2b\"
**Steps to reproduce:**
1. Set agents.defaults.imageModel to ollama/qwen3-vl:2b
2. Call the image tool with any image, e.g.: image(prompt=\"Describe this image\", image=\"/path/to/photo.jpg\")
**Expected... | closed | completed | true | 3 | [] | [] | 2026-03-21T18:05:32Z | 2026-04-30T04:50:27Z | 2026-04-25T05:43:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangsen992 | 6,318,958 | MDQ6VXNlcjYzMTg5NTg= | User | false |
openclaw/openclaw | 4,113,247,061 | I_kwDOQb6kR871KytV | 51,800 | https://github.com/openclaw/openclaw/issues/51800 | https://api.github.com/repos/openclaw/openclaw/issues/51800 | Web UI: onAttachmentsChange callback is undefined, paste image not delivered to Gateway | ## Bug Description
Pasting an image in the Web UI control panel does not deliver the image to the Gateway/agent. The image is detected by the paste handler (console logs show the base64 dataUrl), but the `onAttachmentsChange` callback is undefined, causing the handler to return early without updating `chatAttachments`... | closed | not_planned | true | 3 | [] | [] | 2026-03-21T18:10:16Z | 2026-04-30T04:50:26Z | 2026-04-25T05:43:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lulu-claude | 262,723,017 | U_kgDOD6jVyQ | User | false |
openclaw/openclaw | 4,113,403,847 | I_kwDOQb6kR871LY_H | 51,818 | https://github.com/openclaw/openclaw/issues/51818 | https://api.github.com/repos/openclaw/openclaw/issues/51818 | Sub-agent completion results silently lost when parent session expires — channel-direct fallback proposal | ## Problem
When a parent agent spawns a sub-agent and the parent session ends/expires before the sub-agent completes, the completion announcement fails silently. The sub-agent's result is never delivered to the user.
This affects all channels (Telegram, Discord, WhatsApp, Slack).
### Reproduction steps
1. Agent rec... | closed | completed | true | 2 | [] | [] | 2026-03-21T18:55:19Z | 2026-04-30T04:50:24Z | 2026-04-25T05:43:51Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | hershey-g | 222,996,166 | U_kgDODUqmxg | User | false |
openclaw/openclaw | 4,113,411,367 | I_kwDOQb6kR871La0n | 51,819 | https://github.com/openclaw/openclaw/issues/51819 | https://api.github.com/repos/openclaw/openclaw/issues/51819 | Memory leak: seqByRun Map in agent-events.ts grows unbounded | ## Bug
The `seqByRun` Map in `src/infra/agent-events.ts` stores one entry per `runId` but never deletes them. `clearAgentRunContext()` cleans up `runContextById` but does not touch `seqByRun`.
Over time this Map grows by hundreds of entries per day and is never pruned, contributing to heap growth on long-running gate... | closed | completed | true | 3 | [] | [] | 2026-03-21T18:57:02Z | 2026-04-30T04:50:23Z | 2026-04-25T05:43:57Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | aneym | 43,046,102 | MDQ6VXNlcjQzMDQ2MTAy | User | false |
openclaw/openclaw | 4,113,461,776 | I_kwDOQb6kR871LnIQ | 51,830 | https://github.com/openclaw/openclaw/issues/51830 | https://api.github.com/repos/openclaw/openclaw/issues/51830 | [Bug]: message send can be misclassified as poll when empty poll fields are auto-included | ### Bug type
Behavior bug
### Summary
`message` send/file-send requests can be misclassified as poll actions when the caller/wrapper auto-includes empty `poll*` fields, causing normal sends to fail with a poll-validation error.
### Steps to reproduce
1. Run OpenClaw on a version/build that exposes the unified `messag... | closed | completed | true | 3 | [] | [] | 2026-03-21T19:08:29Z | 2026-04-30T04:50:21Z | 2026-04-25T05:44:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JUk1-GH | 236,887,312 | U_kgDODh6dEA | User | false |
openclaw/openclaw | 4,113,525,208 | I_kwDOQb6kR871L2nY | 51,837 | https://github.com/openclaw/openclaw/issues/51837 | https://api.github.com/repos/openclaw/openclaw/issues/51837 | Windows: exec spawns steal focus from active window | ## Description
On native Windows (not WSL), commands spawned via the `exec` tool open visible command windows that steal focus from whatever the user is actively working on. This is disruptive during normal use — typing gets interrupted, windows flash, etc.
## Current behavior
- `exec` tool spawns processes that cre... | closed | completed | true | 3 | [] | [] | 2026-03-21T19:22:35Z | 2026-04-30T04:50:21Z | 2026-04-25T05:44:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | junkyard-rules | 131,377,247 | U_kgDOB9SoXw | User | false |
openclaw/openclaw | 4,113,634,877 | I_kwDOQb6kR871MRY9 | 51,851 | https://github.com/openclaw/openclaw/issues/51851 | https://api.github.com/repos/openclaw/openclaw/issues/51851 | [Bug]: WebChat/VSCode UI hangs after /new — agent response not displayed until manual refresh | ## Bug Description
After the latest OpenClaw update (2026.3.13), running `/new` or `/reset` in the WebChat Control UI (and the VSCode extension) causes the UI to appear stuck. The agent processes the session reset and generates a startup greeting, but the response never appears in the chat. The user must manually clic... | closed | completed | true | 2 | [] | [] | 2026-03-21T19:49:29Z | 2026-04-30T04:50:20Z | 2026-04-25T05:44:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | segme | 8,915,558 | MDQ6VXNlcjg5MTU1NTg= | User | false |
openclaw/openclaw | 4,113,659,232 | I_kwDOQb6kR871MXVg | 51,854 | https://github.com/openclaw/openclaw/issues/51854 | https://api.github.com/repos/openclaw/openclaw/issues/51854 | agents.defaults.subagents.model config not applied to spawned sub-agents | ## Bug
`agents.defaults.subagents.model` is set in config but sub-agents still inherit the parent's model instead of using the configured override.
## Config
```json
{
"agents": {
"defaults": {
"model": {
"primary": "anthropic/claude-opus-4-6",
"fallbacks": []
},
"subagents": ... | closed | completed | true | 3 | [] | [] | 2026-03-21T19:58:12Z | 2026-04-30T04:50:19Z | 2026-04-25T05:44:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | vhicktour | 69,664,551 | MDQ6VXNlcjY5NjY0NTUx | User | false |
openclaw/openclaw | 4,113,760,234 | I_kwDOQb6kR871Mv_q | 51,862 | https://github.com/openclaw/openclaw/issues/51862 | https://api.github.com/repos/openclaw/openclaw/issues/51862 | [Feature]: Per-channel/group workspace file injection for context isolation | ### Summary
Allow per-channel/group workspace files to be specified in config so the agent loads only contextually relevant memory files for each Telegram group or Discord channel.
### Problem to solve
Users are advised to create separate Telegram groups per topic to manage context window size — but this only works ... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-03-21T20:26:09Z | 2026-04-30T04:50:18Z | 2026-04-25T05:44:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LeoStehlik | 34,870,689 | MDQ6VXNlcjM0ODcwNjg5 | User | false |
openclaw/openclaw | 4,113,882,370 | I_kwDOQb6kR871NN0C | 51,879 | https://github.com/openclaw/openclaw/issues/51879 | https://api.github.com/repos/openclaw/openclaw/issues/51879 | [Bug]: CLI gateway handshake timeout on WSL2 — 2026.3.13 regression | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to 2026.3.13, all CLI-to-gateway commands (gateway health, devices list, etc.) fail with "gateway closed (1000 normal closure)" on WSL2. The gateway server log shows "handshake timeout" after 3s. The browser Control UI connects successful... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-21T21:17:33Z | 2026-04-30T04:50:17Z | 2026-04-25T05:44:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | markvshaney | 30,285,016 | MDQ6VXNlcjMwMjg1MDE2 | User | false |
openclaw/openclaw | 4,113,886,041 | I_kwDOQb6kR871NOtZ | 51,880 | https://github.com/openclaw/openclaw/issues/51880 | https://api.github.com/repos/openclaw/openclaw/issues/51880 | [Feature]: Agent-specific reply account for group messages | ## Feature Request: Agent-specific reply account for group messages
### Use Case
In a team environment with multiple Telegram bots (anna, nora, chloe, ethan), when anna assigns work to a specific agent in a group (e.g., "anna @Ethan please do X"), the responding agent should reply using its own bot account, not anna's... | closed | completed | true | 3 | [] | [] | 2026-03-21T21:19:14Z | 2026-04-30T04:50:16Z | 2026-04-25T05:44:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tslatop | 76,646,873 | MDQ6VXNlcjc2NjQ2ODcz | User | false |
openclaw/openclaw | 4,113,896,271 | I_kwDOQb6kR871NRNP | 51,882 | https://github.com/openclaw/openclaw/issues/51882 | https://api.github.com/repos/openclaw/openclaw/issues/51882 | Feature Request: Configurable connect-src or gateway proxy path for Mission Control local API access | ## Problem
Mission Control is served from the gateway origin (`127.0.0.1:18789`). We run a local API server on port 3001. All browser-side fetches to that API are blocked by the gateway's `Content-Security-Policy` response header:
```
connect-src 'self' ws: wss:
```
Port differs from the gateway origin, so the brows... | closed | completed | true | 2 | [] | [] | 2026-03-21T21:23:49Z | 2026-04-30T04:50:15Z | 2026-04-25T05:45:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | billpark88 | 234,947,759 | U_kgDODgEErw | User | false |
openclaw/openclaw | 4,113,916,133 | I_kwDOQb6kR871NWDl | 51,887 | https://github.com/openclaw/openclaw/issues/51887 | https://api.github.com/repos/openclaw/openclaw/issues/51887 | [Bug]: gateway-client / TUI loses operator.read and breaks openclaw status --all | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
On Windows (OpenClaw 2026.3.13), the local gateway-client/openclaw-tui path appears to drop operator.read and ends up effectively admin-only.
openclaw gateway status can show RPC probe ok, but openclaw status --all and openclaw security a... | closed | completed | true | 7 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T21:31:20Z | 2026-04-30T04:50:14Z | 2026-04-25T05:45:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | glassdidgeridoo | 211,490,435 | U_kgDODJsWgw | User | false |
openclaw/openclaw | 4,114,030,185 | I_kwDOQb6kR871Nx5p | 51,896 | https://github.com/openclaw/openclaw/issues/51896 | https://api.github.com/repos/openclaw/openclaw/issues/51896 | [Bug]: Model choice -> openAI codex configuration oauth link broken | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
w configure
│
◇ Existing config detected ──────────╮
│ │
│ model: openai-codex/gpt-5.3-codex │
│ gateway.mode: local │
│ │
├─────────────────────────... | closed | not_planned | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T22:22:44Z | 2026-04-30T04:50:13Z | 2026-04-25T05:45:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mileslow | 40,550,490 | MDQ6VXNlcjQwNTUwNDkw | User | false |
openclaw/openclaw | 4,114,070,985 | I_kwDOQb6kR871N73J | 51,904 | https://github.com/openclaw/openclaw/issues/51904 | https://api.github.com/repos/openclaw/openclaw/issues/51904 | memory_search tool ignores memorySearch.provider config, always uses OpenAI embeddings |
**Describe the bug**
memory_search tool ignores agents.defaults.memorySearch.provider config and always attempts to use OpenAI embeddings, even when configured for voyage (or other providers).
**To Reproduce**
1. Configure agents.defaults.memorySearch with provider: \"voyage\" and a VOYAGE_API_KEY env var
2. Attempt ... | closed | completed | true | 3 | [] | [] | 2026-03-21T22:41:27Z | 2026-04-30T04:50:13Z | 2026-04-25T05:45:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Jragyn | 7,897,461 | MDQ6VXNlcjc4OTc0NjE= | User | false |
openclaw/openclaw | 4,114,083,847 | I_kwDOQb6kR871N_AH | 51,908 | https://github.com/openclaw/openclaw/issues/51908 | https://api.github.com/repos/openclaw/openclaw/issues/51908 | Obfuscation detector pipe-to-shell pattern false-positives on || (logical OR) operator | ## Bug Description
The `pipe-to-shell` obfuscation detection pattern in `exec-obfuscation-detect.ts` produces false positives when commands use the `||` (logical OR) operator with shell interpreters like `bash`.
## Example
Command:
```
python3 bin/garmin_health.py 2>/dev/null || bash bin/garmin-health.sh 2>/dev/null... | closed | not_planned | true | 4 | [] | [] | 2026-03-21T22:47:49Z | 2026-04-30T04:50:12Z | 2026-04-25T05:45:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | TurboTheTurtle | 35,905,412 | MDQ6VXNlcjM1OTA1NDEy | User | false |
openclaw/openclaw | 4,114,096,143 | I_kwDOQb6kR871OCAP | 51,909 | https://github.com/openclaw/openclaw/issues/51909 | https://api.github.com/repos/openclaw/openclaw/issues/51909 | bug(acp): acpx gateway runtime fails with 'exited with code 1' — works fine via direct acpx invocation | ## Environment
- OpenClaw: 2026.3.13
- OS: macOS Darwin 25.3.0 (arm64)
- Node: v22.22.1
- acpx plugin version: 2026.3.13 (bundled)
- acpx binary version: 0.1.16
## What happens
Every ACP run session spawned via `sessions_spawn` with `runtime: "acp", agentId: "claude"` fails immediately with:
```
claude run failed: acp... | closed | completed | true | 3 | [] | [] | 2026-03-21T22:54:00Z | 2026-04-30T04:50:11Z | 2026-04-25T05:45:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mnyamolo | 52,471,852 | MDQ6VXNlcjUyNDcxODUy | User | false |
openclaw/openclaw | 4,114,146,036 | I_kwDOQb6kR871OOL0 | 51,921 | https://github.com/openclaw/openclaw/issues/51921 | https://api.github.com/repos/openclaw/openclaw/issues/51921 | Test: GitHub Skill Working | This is a test issue created by OpenClaw assistant to verify GitHub skill functionality. Account: casagerardon | closed | not_planned | true | 1 | [] | [] | 2026-03-21T23:14:36Z | 2026-04-30T04:50:10Z | 2026-04-25T05:45:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | casagerardon | 270,026,166 | U_kgDOEBhFtg | User | false |
openclaw/openclaw | 4,114,155,733 | I_kwDOQb6kR871OQjV | 51,928 | https://github.com/openclaw/openclaw/issues/51928 | https://api.github.com/repos/openclaw/openclaw/issues/51928 | Feature: Transparent exec logging — show commands before/as they run (log mode without confirmation) | ## Feature Request: Transparent exec logging — show commands before/as they run
### Summary
When an agent uses the `exec` tool to run commands on the host machine, there is no built-in way for users to see what is being executed in real time — especially in chat-based sessions (Telegram, Discord, etc.) where the term... | closed | completed | true | 4 | [] | [] | 2026-03-21T23:19:44Z | 2026-04-30T04:50:10Z | 2026-04-25T05:45:49Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ernestodeoliveira | 26,804,139 | MDQ6VXNlcjI2ODA0MTM5 | User | false |
openclaw/openclaw | 4,114,161,750 | I_kwDOQb6kR871OSBW | 51,929 | https://github.com/openclaw/openclaw/issues/51929 | https://api.github.com/repos/openclaw/openclaw/issues/51929 | Test: GitHub Skill Working | Test issue | closed | not_planned | true | 1 | [] | [] | 2026-03-21T23:22:57Z | 2026-04-30T04:50:09Z | 2026-04-25T05:45:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | casagerardon | 270,026,166 | U_kgDOEBhFtg | User | false |
openclaw/openclaw | 4,114,182,152 | I_kwDOQb6kR871OXAI | 51,932 | https://github.com/openclaw/openclaw/issues/51932 | https://api.github.com/repos/openclaw/openclaw/issues/51932 | [Bug]: exec approval fails in control-ui over plain HTTP due to missing approval client recognition | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Summary:
In a plain HTTP webchat/control-ui setup, exec approval requests fail immediately with a no-approval-route outcome because the connected control-ui client is not recognized as a valid approval client.
### Steps to reproduce
1. Ru... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-21T23:33:18Z | 2026-04-30T04:50:08Z | 2026-04-25T05:46:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | playaopindo-dev | 257,603,176 | U_kgDOD1q2aA | User | false |
openclaw/openclaw | 4,114,249,975 | I_kwDOQb6kR871Onj3 | 51,941 | https://github.com/openclaw/openclaw/issues/51941 | https://api.github.com/repos/openclaw/openclaw/issues/51941 | [Bug]: cron timezone handling inconsistent between scheduled jobs and manual trigger | ## Bug Description
### Summary
Cron jobs scheduled with timezone-aware cron expressions behave differently when triggered manually via `cron job run` vs when triggered automatically by the scheduler. The scheduler correctly applies timezone conversion, but manual triggers use UTC, causing unexpected execution times.
... | closed | not_planned | true | 3 | [] | [] | 2026-03-22T00:03:36Z | 2026-04-30T04:50:07Z | 2026-04-25T05:46:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,114,347,714 | I_kwDOQb6kR871O_bC | 51,975 | https://github.com/openclaw/openclaw/issues/51975 | https://api.github.com/repos/openclaw/openclaw/issues/51975 | Error finding Sentry configuration | Sentry CLI authentication failed with error message 'Auth token is required for this request.' | closed | not_planned | true | 1 | [] | [] | 2026-03-22T01:16:15Z | 2026-04-30T04:50:06Z | 2026-04-25T05:46:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gitbotv1-pixel | 262,690,772 | U_kgDOD6hX1A | User | false |
openclaw/openclaw | 4,085,646,677 | I_kwDOQb6kR87zhgVV | 48,614 | https://github.com/openclaw/openclaw/issues/48614 | https://api.github.com/repos/openclaw/openclaw/issues/48614 | [Bug]: Signal voice notes — MediaPath not populated, audio transcription pipeline never triggers | ### Bug type
Regression (worked before, now fails)
### Summary
Signal voice notes arrive as [media attached: /path.aac (audio/x-m4a)] text in the message body but ctx.MediaPath / ctx.MediaType are not set, so tools.media.audio transcription never runs.
### Steps to reproduce
Steps to reproduce:
1. Configure to... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-17T01:18:02Z | 2026-04-30T04:50:05Z | 2026-04-25T05:46:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | vector761 | 268,498,748 | U_kgDOEAD3PA | User | false |
openclaw/openclaw | 4,114,348,158 | I_kwDOQb6kR871O_h- | 51,976 | https://github.com/openclaw/openclaw/issues/51976 | https://api.github.com/repos/openclaw/openclaw/issues/51976 | Sentry CLI authentication failed | Failed to authenticate with Sentry CLI. No .sentryclirc file found and authentication requires a token. | closed | not_planned | true | 1 | [] | [] | 2026-03-22T01:16:32Z | 2026-04-30T04:50:04Z | 2026-04-25T05:46:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gitbotv1-pixel | 262,690,772 | U_kgDOD6hX1A | User | false |
openclaw/openclaw | 4,114,350,675 | I_kwDOQb6kR871PAJT | 51,978 | https://github.com/openclaw/openclaw/issues/51978 | https://api.github.com/repos/openclaw/openclaw/issues/51978 | Feature: Support Ollama with proper fallback model configuration | ## Issue Description
Currently OpenClaw supports Ollama as a model provider but there are configuration issues that make it difficult to use:
### Problem 1: Ollama API Path Error
When configuring Ollama as a provider, the base URL gets doubled:
```
Ollama API error 404: {"error":"path "/api/api/chat" not found"}
```
... | closed | completed | true | 2 | [] | [] | 2026-03-22T01:18:27Z | 2026-04-30T04:50:03Z | 2026-04-25T05:46:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | larrybuckalew | 969,197 | MDQ6VXNlcjk2OTE5Nw== | User | false |
openclaw/openclaw | 4,114,400,769 | I_kwDOQb6kR871PMYB | 51,987 | https://github.com/openclaw/openclaw/issues/51987 | https://api.github.com/repos/openclaw/openclaw/issues/51987 | Local gateway websocket handshake times out on 127.0.0.1:18789 and closes before connect | ### Bug type
Regression (worked before, now fails)
### Summary
With a local loopback gateway at ws://127.0.0.1:18789, browser/gateway connection attempts can fail because the websocket closes before connect and the gateway logs `handshake timeout` from 127.0.0.1.
### Steps to reproduce
1. Configure OpenClaw 2026.3... | closed | completed | true | 7 | [
"bug",
"regression"
] | [] | 2026-03-22T01:59:11Z | 2026-04-30T04:50:02Z | 2026-04-25T05:46:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MeridianWorld | 58,415,149 | MDQ6VXNlcjU4NDE1MTQ5 | User | false |
openclaw/openclaw | 4,114,412,808 | I_kwDOQb6kR871PPUI | 51,991 | https://github.com/openclaw/openclaw/issues/51991 | https://api.github.com/repos/openclaw/openclaw/issues/51991 | fix(sessions): updateLastRoute bumps updatedAt, preventing idle session reset | ## Why this matters
With the current stable release, **idle session reset is completely broken for Discord (and likely all channels)**. This means every message — even a simple "hi" after hours of inactivity — carries the full conversation context (30k+ tokens in our case). Users have no way to benefit from `session.r... | closed | completed | true | 4 | [] | [] | 2026-03-22T02:08:21Z | 2026-04-30T04:50:01Z | 2026-04-25T05:46:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alderaan | 4,752,845 | MDQ6VXNlcjQ3NTI4NDU= | User | false |
openclaw/openclaw | 4,114,420,369 | I_kwDOQb6kR871PRKR | 51,998 | https://github.com/openclaw/openclaw/issues/51998 | https://api.github.com/repos/openclaw/openclaw/issues/51998 | [Feature]:Feature Request — Telegram Group Channel Context Binding | ### Summary
Subject: Feature Request — Telegram Group Channel Context Binding
Hi OpenClaw team,
I'm running a multi-agent setup where two instances (Claude Code + OpenClaw
gateway) share a Telegram group chat. I've run into a routing issue and wanted to flag it as
a feature request.
Current behavior:
... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-22T02:14:01Z | 2026-04-30T04:50:01Z | 2026-04-25T05:46:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | docktaclawedbot-arch | 270,043,447 | U_kgDOEBiJNw | User | false |
openclaw/openclaw | 4,114,422,313 | I_kwDOQb6kR871PRop | 52,001 | https://github.com/openclaw/openclaw/issues/52001 | https://api.github.com/repos/openclaw/openclaw/issues/52001 | Web UI model selector assigns wrong provider prefix for models shared across multiple providers | **Description**%0A%0AWhen switching models in the OpenClaw Web UI (control-ui), if a model ID exists in multiple providers under different names, the UI incorrectly assigns the wrong provider prefix when sending the sessions.patch request.%0A%0A**Example**%0A%0AThe model astron-code-latest belongs to provider astroncod... | closed | completed | true | 2 | [] | [] | 2026-03-22T02:15:30Z | 2026-04-30T04:50:00Z | 2026-04-25T05:47:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tcyxk | 129,282,805 | U_kgDOB7Sy9Q | User | false |
openclaw/openclaw | 4,214,504,563 | I_kwDOQb6kR877NDxz | 62,173 | https://github.com/openclaw/openclaw/issues/62173 | https://api.github.com/repos/openclaw/openclaw/issues/62173 | Audio transcription fails on 2026.4.5: SSRF guard corrupts multipart FormData for Whisper API | Voice note transcription via OpenAI Whisper API fails on 2026.4.5. The `fetchWithSsrFGuard` function uses undici's fetch with a pinned DNS dispatcher, which corrupts multipart FormData boundaries. OpenAI rejects the request with "you must provide a model parameter" even though it's present in the form data.
This is th... | closed | completed | true | 2 | [] | [] | 2026-04-06T23:05:46Z | 2026-04-30T04:49:59Z | 2026-04-25T05:47:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ketani23 | 11,549,089 | MDQ6VXNlcjExNTQ5MDg5 | User | false |
openclaw/openclaw | 4,114,422,447 | I_kwDOQb6kR871PRqv | 52,002 | https://github.com/openclaw/openclaw/issues/52002 | https://api.github.com/repos/openclaw/openclaw/issues/52002 | Web UI model selector assigns wrong provider prefix for models shared across multiple providers | **Description**%0A%0AWhen switching models in the OpenClaw Web UI (control-ui), if a model ID exists in multiple providers under different names, the UI incorrectly assigns the wrong provider prefix when sending the sessions.patch request.%0A%0A**Example**%0A%0AThe model astron-code-latest belongs to provider astroncod... | closed | completed | true | 3 | [] | [] | 2026-03-22T02:15:36Z | 2026-04-30T04:49:58Z | 2026-04-25T05:47:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tcyxk | 129,282,805 | U_kgDOB7Sy9Q | User | false |
openclaw/openclaw | 4,114,442,452 | I_kwDOQb6kR871PWjU | 52,014 | https://github.com/openclaw/openclaw/issues/52014 | https://api.github.com/repos/openclaw/openclaw/issues/52014 | Gateway crash-loops on config reload when node host pairing fails | ## Description
After installing the `lossless-claw` plugin, the config file was updated (`meta.lastTouchedAt`, `plugins.installs`). The gateway detected the config change and triggered a SIGTERM self-restart.
On restart, the **node host** subsystem failed to reconnect with error:
```
gateway connect failed: GatewayCl... | closed | not_planned | true | 4 | [] | [] | 2026-03-22T02:26:02Z | 2026-04-30T04:49:57Z | 2026-04-25T05:47:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | antimeat | 79,799,747 | MDQ6VXNlcjc5Nzk5NzQ3 | User | false |
openclaw/openclaw | 4,114,443,001 | I_kwDOQb6kR871PWr5 | 52,015 | https://github.com/openclaw/openclaw/issues/52015 | https://api.github.com/repos/openclaw/openclaw/issues/52015 | Heartbeat: overlapping runs overwrite session and cause nested session keys | ## Bug
When `isolatedSession: true` is configured for heartbeat, two issues occur:
### 1. No overlap protection — concurrent heartbeats overwrite each other
`runHeartbeatOnce` calls `resolveCronSession` with `forceNew: true` every time it fires. If a previous heartbeat is still running (e.g., LLM is mid-execution), ... | closed | completed | true | 9 | [] | [] | 2026-03-22T02:26:14Z | 2026-04-30T04:49:57Z | 2026-04-25T05:47:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wiwaxia-ai | 258,413,720 | U_kgDOD2cUmA | User | false |
openclaw/openclaw | 4,114,454,251 | I_kwDOQb6kR871PZbr | 52,019 | https://github.com/openclaw/openclaw/issues/52019 | https://api.github.com/repos/openclaw/openclaw/issues/52019 | Add Fish Audio as TTS provider | **Feature Request**
Please add Fish Audio (fish.audio) as a TTS provider option alongside ElevenLabs, OpenAI, and Edge.
**Why:**
- Fish Audio has unique voice models not available on other platforms
- Specifically need the T-800 ENG voice (Austrian/Terminator style)
- Already have API key and voice ID ready to use
*... | closed | not_planned | true | 3 | [] | [] | 2026-03-22T02:33:43Z | 2026-04-30T04:49:56Z | 2026-04-25T05:47:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Rheingold777 | 104,510,750 | U_kgDOBjq1Hg | User | false |
openclaw/openclaw | 4,114,464,098 | I_kwDOQb6kR871Pb1i | 52,023 | https://github.com/openclaw/openclaw/issues/52023 | https://api.github.com/repos/openclaw/openclaw/issues/52023 | message:sent hook events don't fire for tool-dispatched messages | ## Summary
`message:sent` hook events never fire when messages are sent via the `message` tool (e.g., `message action=send target=-5245845023`). Only `message:received` and `message:preprocessed` events fire.
## Steps to Reproduce
1. Create a custom hook listening for `message:sent` events
2. Send a message via the ... | closed | completed | true | 4 | [] | [] | 2026-03-22T02:40:44Z | 2026-04-30T04:49:55Z | 2026-04-25T05:47:33Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wrentheai | 257,617,545 | U_kgDOD1ruiQ | User | false |
openclaw/openclaw | 4,114,471,899 | I_kwDOQb6kR871Pdvb | 52,024 | https://github.com/openclaw/openclaw/issues/52024 | https://api.github.com/repos/openclaw/openclaw/issues/52024 | Context trimmer splits tool_use/tool_result pairs, causing API rejection | ## Bug Description
When a session transcript grows large enough to require context trimming (to fit within the model's context window), the trimmer can remove an assistant message containing a `tool_use` block while keeping the subsequent user message containing the corresponding `tool_result` block. This produces an ... | closed | completed | true | 7 | [] | [] | 2026-03-22T02:46:24Z | 2026-04-30T04:49:54Z | 2026-04-25T05:47:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | nrubioru | 256,029,441 | U_kgDOD0KzAQ | User | false |
openclaw/openclaw | 4,114,501,485 | I_kwDOQb6kR871Pk9t | 52,034 | https://github.com/openclaw/openclaw/issues/52034 | https://api.github.com/repos/openclaw/openclaw/issues/52034 | [Bug]: The result of "clawhub search self-improving-agent" missing exact match | ### Bug type
Regression (worked before, now fails)
### Summary
npx clawhub search self-improving-agent
or
https://clawhub.ai/skills?q=self-improving-agent
### Steps to reproduce
npx clawhub search self-improving-agent
or
https://clawhub.ai/skills?q=self-improving-agent
### Expected behavior
return self-improving... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-22T03:06:23Z | 2026-04-30T04:49:54Z | 2026-04-25T05:47:46Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | zerolyd2026 | 269,719,430 | U_kgDOEBOXhg | User | false |
openclaw/openclaw | 4,114,540,984 | I_kwDOQb6kR871Pum4 | 52,040 | https://github.com/openclaw/openclaw/issues/52040 | https://api.github.com/repos/openclaw/openclaw/issues/52040 | Control UI: Model selector dropdown sends bare model ID without provider prefix | ## Bug Description
When switching models via the Control UI chat model dropdown, the selected model's `value` is set to just the model ID (e.g., `Qwen3.5-4B-MLX-4bit`) without the provider prefix. The backend's `resolveModelRefFromString` then uses the **current session's provider** as the default prefix, which produc... | closed | completed | true | 4 | [] | [] | 2026-03-22T03:23:07Z | 2026-04-30T04:49:52Z | 2026-04-25T05:48:00Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Evan-Luo-Engnieer | 12,590,510 | MDQ6VXNlcjEyNTkwNTEw | User | false |
openclaw/openclaw | 4,114,553,432 | I_kwDOQb6kR871PxpY | 52,044 | https://github.com/openclaw/openclaw/issues/52044 | https://api.github.com/repos/openclaw/openclaw/issues/52044 | Bug: gateway restart spawns duplicate processes on Windows (3 windows) | # Bug: `gateway restart` spawns duplicate gateway processes on Windows (3 windows)
## Summary
`openclaw gateway restart` on Windows produces **3 cmd windows** instead of 1:
1. The original gateway window (being shut down)
2. A new window from the CLI's `schtasks /Run`
3. A third window from the gateway's **self-resta... | closed | completed | true | 7 | [] | [] | 2026-03-22T03:27:50Z | 2026-04-30T04:49:50Z | 2026-04-25T05:48:06Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | andyk-ms | 91,510,251 | U_kgDOBXRV6w | User | false |
openclaw/openclaw | 4,114,573,387 | I_kwDOQb6kR871P2hL | 52,049 | https://github.com/openclaw/openclaw/issues/52049 | https://api.github.com/repos/openclaw/openclaw/issues/52049 | Bug: gateway stop doesn't terminate node.exe process on Windows | # Bug: `gateway stop` doesn't terminate the node.exe process on Windows
## Summary
`openclaw gateway stop` reports "Stopped Scheduled Task: OpenClaw Gateway" but the actual `node.exe` gateway process continues running. The gateway remains accessible on port 18789 after `stop`.
## Root Cause
The "OpenClaw Gateway" s... | closed | completed | true | 4 | [] | [] | 2026-03-22T03:34:43Z | 2026-04-30T04:49:49Z | 2026-04-25T05:48:13Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | andyk-ms | 91,510,251 | U_kgDOBXRV6w | User | false |
openclaw/openclaw | 4,114,637,479 | I_kwDOQb6kR871QGKn | 52,061 | https://github.com/openclaw/openclaw/issues/52061 | https://api.github.com/repos/openclaw/openclaw/issues/52061 | [Bug]: Control UI model picker prefixes openai/ when selecting Anthropic Claude → model not allowed | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Version
OpenClaw: 2026.3.13 (commit 61d171a)
OS: macOS (Apple Silicon) on Mac mini
Browser: Safari
Channel: Control UI (webchat)
What I’m trying to do
Switch the session model in the Control UI from OpenAI GPT to Anthropic Claude (e.g., ant... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T03:59:39Z | 2026-04-30T04:49:48Z | 2026-04-25T05:48:19Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jlg-oh | 266,536,066 | U_kgDOD-MEgg | User | false |
openclaw/openclaw | 4,114,684,316 | I_kwDOQb6kR871QRmc | 52,069 | https://github.com/openclaw/openclaw/issues/52069 | https://api.github.com/repos/openclaw/openclaw/issues/52069 | [Bug] Token usage 显示为 0，dashscope-coding-plan API 返回的字段名不匹配 | ## 问题描述
使用 `dashscope-coding-plan` 提供商时，session status 显示 Context: 0/128k (0%)，所有 usage 字段都为 0。
## 根本原因
API 实际返回了正确的 usage 信息，但 OpenClaw 代码中的字段名不匹配：
**OpenClaw 代码期望的字段名**（auth-profiles-DDVivXkv.js）：
```javascript
input: response.usage?.input_tokens ?? 0,
output: response.usage?.output_tokens ?? 0,
totalTokens: resp... | closed | completed | true | 3 | [] | [] | 2026-03-22T04:20:03Z | 2026-04-30T04:49:47Z | 2026-04-25T05:48:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ZeroXenith | 132,541,447 | U_kgDOB-ZsBw | User | false |
openclaw/openclaw | 4,114,688,349 | I_kwDOQb6kR871QSld | 52,071 | https://github.com/openclaw/openclaw/issues/52071 | https://api.github.com/repos/openclaw/openclaw/issues/52071 | CLI status/audit probe can fail with missing scope: operator.read while main Control UI auth works | ## Summary
`openclaw gateway status` / `openclaw security audit --deep` can report `gateway.probe_failed` with `missing scope: operator.read` even when the main Control UI session is authenticated and working.
This appears to be a probe-path credential resolution bug in the CLI/daemon status code, not a real failure o... | closed | completed | true | 4 | [] | [] | 2026-03-22T04:22:09Z | 2026-04-30T04:49:46Z | 2026-04-25T05:48:33Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | schabot3 | 270,057,071 | U_kgDOEBi-bw | User | false |
openclaw/openclaw | 4,114,753,826 | I_kwDOQb6kR871Qiki | 52,081 | https://github.com/openclaw/openclaw/issues/52081 | https://api.github.com/repos/openclaw/openclaw/issues/52081 | [Bug]: [plugins] feishu failed to load from /app/extensions/feishu/index.ts: TypeError: (0 , _runtimeApi.buildChannelConfigSchema) is not a function | ### Bug type
Regression (worked before, now fails)
### Summary
◇ Select a channel
│ Feishu/Lark (飞书)
12:38:09 [plugins] feishu failed to load from /app/extensions/feishu/index.ts: TypeError: (0 , _runtimeApi.buildChannelConfigSchema) is not a function
### Steps to reproduce
◇ Select a channel
│ Feishu/Lark (飞书... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-03-22T04:43:00Z | 2026-04-30T04:49:46Z | 2026-04-25T05:48:39Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | hhxdestiny | 44,420,757 | MDQ6VXNlcjQ0NDIwNzU3 | User | false |
openclaw/openclaw | 4,114,793,695 | I_kwDOQb6kR871QsTf | 52,085 | https://github.com/openclaw/openclaw/issues/52085 | https://api.github.com/repos/openclaw/openclaw/issues/52085 | TUI / gateway-client connects but RPC calls fail with `missing scope: operator.read` despite paired device/token showing full operator scopes | ## Summary
On a fresh local macOS install of OpenClaw, the TUI / gateway-client can connect to the local gateway, but scope-gated RPC calls fail with:
- `missing scope: operator.read`
This happens even after:
- fresh device auth reset
- device token rotation with full operator scopes
- removing + re-pairing the loca... | closed | completed | true | 4 | [] | [] | 2026-03-22T04:53:59Z | 2026-04-30T04:49:45Z | 2026-04-25T05:48:46Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | danshoop | 6,692,679 | MDQ6VXNlcjY2OTI2Nzk= | User | false |
openclaw/openclaw | 4,114,794,231 | I_kwDOQb6kR871Qsb3 | 52,086 | https://github.com/openclaw/openclaw/issues/52086 | https://api.github.com/repos/openclaw/openclaw/issues/52086 | [openclaw-weixin] Support receiving forwarded WeChat article (link) messages | ## Feature Request
### Problem
When a user forwards a WeChat Official Account (公众号) article to the bot, the message is silently dropped — the agent receives nothing. This makes it impossible to discuss or process shared articles.
### Expected Behavior
When a link/article message is forwarded, the agent should receive... | closed | not_planned | true | 3 | [] | [] | 2026-03-22T04:54:15Z | 2026-04-30T04:49:44Z | 2026-04-25T05:48:52Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Xiaochengzi2048 | 15,998,677 | MDQ6VXNlcjE1OTk4Njc3 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.