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 | 4,015,961,320 | I_kwDOQb6kR87vXrTo | 32,906 | https://github.com/openclaw/openclaw/issues/32906 | https://api.github.com/repos/openclaw/openclaw/issues/32906 | Feature Request: Configuration Validation on Startup | ## Summary
Prevent Gateway crashes by validating configuration files before startup and providing clear error messages for invalid configuration keys.
## Problem
**Real-world incident:** An invalid configuration key `groupAllowFrom` caused Gateway to crash in a loop without any clear error message. The only way to d... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T10:42:54Z | 2026-04-19T04:24:31Z | 2026-04-16T04:25:32Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | SonicBotMan | 258,392,144 | U_kgDOD2bAUA | User | false |
openclaw/openclaw | 4,011,298,551 | I_kwDOQb6kR87vF473 | 31,720 | https://github.com/openclaw/openclaw/issues/31720 | https://api.github.com/repos/openclaw/openclaw/issues/31720 | [Bug]: transit station API like “https://api.akc.to/v1/” blocked by Cloudflare | ### Bug type
Crash (process/app exits or hangs)
### Summary
<img width="1092" height="266" alt="Image" src="https://github.com/user-attachments/assets/a8dc822e-e27c-45b4-a9a6-f1b1c7c67014" />
I found that using the transit station API like https://api.akc.to/v1/ is prone to being blocked by Cloudflare. The reason i... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-02T13:31:50Z | 2026-04-16T04:25:32Z | 2026-04-13T04:29:55Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | memoryoffish | 153,982,431 | U_kgDOCS2V3w | User | false |
openclaw/openclaw | 4,011,364,661 | I_kwDOQb6kR87vGJE1 | 31,733 | https://github.com/openclaw/openclaw/issues/31733 | https://api.github.com/repos/openclaw/openclaw/issues/31733 | Slack: users.list rate-limit loop on large workspaces when dmPolicy uses ID-based allowlist | ## Description
When `channels.slack.dmPolicy` is set to `allowlist` with user IDs (e.g., `allowFrom: ["U03A3QXEER3"]`), OpenClaw calls `users.list` to resolve the allowlist on startup and on incoming messages. On large Slack workspaces (thousands of users), this exhausts Slack's rate limit, causing an infinite retry l... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T13:44:11Z | 2026-04-16T04:25:31Z | 2026-04-13T04:29:59Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Skeptomenos | 22,348,986 | MDQ6VXNlcjIyMzQ4OTg2 | User | false |
openclaw/openclaw | 4,011,347,315 | I_kwDOQb6kR87vGE1z | 31,728 | https://github.com/openclaw/openclaw/issues/31728 | https://api.github.com/repos/openclaw/openclaw/issues/31728 | [Feature]: Add opt-out config for thread participation auto-reply (#29165) | ### Summary
Add a config option to disable the thread participation auto-reply behavior introduced in #29165, so operators can opt out of the bot auto-responding in threads it has previously participated in.
### Problem to solve
PR #29165 added always-on thread participation tracking: once the bot replies in a Slack... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T13:40:44Z | 2026-04-16T04:25:31Z | 2026-04-13T04:29:57Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,015,888,178 | I_kwDOQb6kR87vXZcy | 32,884 | https://github.com/openclaw/openclaw/issues/32884 | https://api.github.com/repos/openclaw/openclaw/issues/32884 | Feature: suppress auto-reply text output for embedded agent (message-tool-only reply mode) | ## Problem
When an embedded agent processes an inbound WhatsApp DM, **every text block** the model produces gets sent as a separate WhatsApp message via the `web-auto-reply` module. This includes internal narration between tool calls (e.g., "Now update the sheet...", "Let me check safe-send..."), tool error messages, ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T10:25:16Z | 2026-04-19T04:24:32Z | 2026-04-16T04:25:30Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | komalamee | 210,163,767 | U_kgDODIbYNw | User | false |
openclaw/openclaw | 4,011,386,282 | I_kwDOQb6kR87vGOWq | 31,734 | https://github.com/openclaw/openclaw/issues/31734 | https://api.github.com/repos/openclaw/openclaw/issues/31734 | Feature request: per-group triage model for non-mention messages | ## Problem
When a WhatsApp/Telegram/Discord group has `requireMention: false` (so the agent receives all messages), every single message triggers a full agent turn with the configured model. For groups where the agent primarily needs to:
1. **Log messages** for downstream processing (e.g., daily summary cron jobs rea... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T13:48:35Z | 2026-04-16T04:25:30Z | 2026-04-13T04:30:01Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | CopyJosh | 919,738 | MDQ6VXNlcjkxOTczOA== | User | false |
openclaw/openclaw | 4,011,457,228 | I_kwDOQb6kR87vGfrM | 31,751 | https://github.com/openclaw/openclaw/issues/31751 | https://api.github.com/repos/openclaw/openclaw/issues/31751 | Node run command fails with 'rawCommand does not match command' after v2026.3.1 update | ## Description
After updating both gateway and node to v2026.3.1, all `nodes run` commands fail with:
```
INVALID_REQUEST: rawCommand does not match command
```
This affects every command, including trivial ones like `echo hello`.
## Environment
- **Gateway:** v2026.3.1, running on Ubuntu (Atlas VM)
- **Node:** v2... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T14:02:35Z | 2026-04-16T04:25:29Z | 2026-04-13T04:30:03Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | clawbottim | 262,697,280 | U_kgDOD6hxQA | User | false |
openclaw/openclaw | 4,011,609,443 | I_kwDOQb6kR87vHE1j | 31,781 | https://github.com/openclaw/openclaw/issues/31781 | https://api.github.com/repos/openclaw/openclaw/issues/31781 | Enhanced Context Management for Long-Running Conversations | # Enhanced Context Management for Long-Running Conversations
## Problem
OpenClaw's current context management relies on reactive compaction that triggers when context windows fill up. This creates challenges for long-term projects:
1. **Loss of Detail**: Compaction performs lossy summarization that cannot be recover... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T14:35:34Z | 2026-04-16T04:25:28Z | 2026-04-13T04:30:12Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | qiuyuemartin-max | 263,818,784 | U_kgDOD7mOIA | User | false |
openclaw/openclaw | 4,011,504,718 | I_kwDOQb6kR87vGrRO | 31,760 | https://github.com/openclaw/openclaw/issues/31760 | https://api.github.com/repos/openclaw/openclaw/issues/31760 | Discord health-monitor stuck-restart loop after upgrading to 2026.3.1 | ## Description
After upgrading from 2026.2.26 to 2026.3.1, the Discord provider enters a stuck-restart loop where the health-monitor flags it as "stuck" every ~10 minutes, despite Discord successfully logging in and resolving channels each time.
## Environment
- **OpenClaw version:** 2026.3.1 (upgraded from 2026.2.26... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T14:12:31Z | 2026-04-16T04:25:28Z | 2026-04-13T04:30:07Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | orangecarrot-dev | 214,410,629 | U_kgDODMelhQ | User | false |
openclaw/openclaw | 4,011,629,006 | I_kwDOQb6kR87vHJnO | 31,785 | https://github.com/openclaw/openclaw/issues/31785 | https://api.github.com/repos/openclaw/openclaw/issues/31785 | [Bug] Feishu plugin fails to load due to /tmp/jiti permissions on Linux | ## Problem
The Feishu plugin fails to load on Linux systems due to permission issues with the `/tmp/jiti` directory, which is used for TypeScript compilation by the `jiti` module.
## Error Message
```
[plugins] feishu failed to load from /home/openclaw/.openclaw/extensions/feishu/index.ts:
Error: EACCES: permission... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T14:39:46Z | 2026-04-16T04:25:27Z | 2026-04-13T04:30:14Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | qiuyuemartin-max | 263,818,784 | U_kgDOD7mOIA | User | false |
openclaw/openclaw | 4,015,184,693 | I_kwDOQb6kR87vUts1 | 32,701 | https://github.com/openclaw/openclaw/issues/32701 | https://api.github.com/repos/openclaw/openclaw/issues/32701 | feat: Agent self-elected turn continuation (CONTINUE_WORK signal) | ## Problem: The Dwindle Pattern
When an agent completes a turn (processes a message, heartbeat, or sub-agent result), it becomes inert until the next external event. There is no mechanism for an agent to signal **"I have more work to do — give me another turn."**
This leads to the **dwindle pattern**: agents with act... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T07:31:20Z | 2026-04-25T04:19:10Z | 2026-04-22T04:30:20Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | karmafeast | 886,771 | MDQ6VXNlcjg4Njc3MQ== | User | false |
openclaw/openclaw | 4,011,631,286 | I_kwDOQb6kR87vHKK2 | 31,787 | https://github.com/openclaw/openclaw/issues/31787 | https://api.github.com/repos/openclaw/openclaw/issues/31787 | [Feature] Automatic session compaction and context length management | ## Problem
Long-running agent sessions accumulate too many messages, leading to:
- `HTTP 400: invalid_request_error: context is too long` errors
- Model API failures
- Manual intervention required to clear sessions
- No automatic cleanup mechanism
## Current Behavior
- Sessions grow indefinitely
- `compaction.mode:... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T14:40:12Z | 2026-04-16T04:25:26Z | 2026-04-13T04:30:16Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | qiuyuemartin-max | 263,818,784 | U_kgDOD7mOIA | User | false |
openclaw/openclaw | 4,011,717,370 | I_kwDOQb6kR87vHfL6 | 31,804 | https://github.com/openclaw/openclaw/issues/31804 | https://api.github.com/repos/openclaw/openclaw/issues/31804 | [Bug]: gateway connect failed: Error: pairing required | ### Bug type
Regression (worked before, now fails)
### Summary
"Chat works fine across WebUI, TUI, and even WeCom, but the cron tool triggers the error: 'gateway connect failed: Error: pairing required ◇ Error: gateway closed (1008): pairing required'."
### Steps to reproduce
input :openclaw devices list
2026-03-... | closed | not_planned | true | 7 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-02T14:57:56Z | 2026-04-16T04:25:25Z | 2026-04-13T04:30:18Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | buptys1 | 53,006,043 | MDQ6VXNlcjUzMDA2MDQz | User | false |
openclaw/openclaw | 4,011,738,123 | I_kwDOQb6kR87vHkQL | 31,808 | https://github.com/openclaw/openclaw/issues/31808 | https://api.github.com/repos/openclaw/openclaw/issues/31808 | Discord DiscordMessageListener serializes all messages per account — cross-channel head-of-line blocking | ## Bug Report
**Version:** 2026.3.1
## Problem
`DiscordMessageListener` (introduced in commit `5056b643`, PR #29508) chains all incoming messages onto a single `messageQueue: Promise<void>` per account. This means a slow agent turn in **any** channel blocks message processing for **all** channels under the same Disc... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T15:02:26Z | 2026-04-16T04:25:24Z | 2026-04-13T04:30:20Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | herta-musubi | 263,009,710 | U_kgDOD601rg | User | false |
openclaw/openclaw | 4,011,788,522 | I_kwDOQb6kR87vHwjq | 31,823 | https://github.com/openclaw/openclaw/issues/31823 | https://api.github.com/repos/openclaw/openclaw/issues/31823 | BlueBubbles: text and balloon webhook events use different debounce keys, causing duplicate replies | ## Bug Description
When a user sends a message containing a URL, BlueBubbles Server fires two webhook events: one for the text message and one for the URL balloon (link preview). These two events represent the **same logical message**, but the debounce key builder assigns them different keys, so they are processed ind... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T15:13:29Z | 2026-04-16T04:25:23Z | 2026-04-13T04:30:24Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | zqchris | 4,436,110 | MDQ6VXNlcjQ0MzYxMTA= | User | false |
openclaw/openclaw | 4,011,753,198 | I_kwDOQb6kR87vHn7u | 31,813 | https://github.com/openclaw/openclaw/issues/31813 | https://api.github.com/repos/openclaw/openclaw/issues/31813 | BlueBubbles: unsupported tapback emoji throws instead of graceful fallback | ## Bug Description
When a user reacts to a message with an emoji that isn't in BlueBubbles' supported `REACTION_TYPES` set, the code throws an `Error("Unsupported BlueBubbles reaction: ...")` instead of gracefully degrading. This crashes the reaction handling flow for that message.
## Expected Behavior
Unsupported e... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T15:05:56Z | 2026-04-16T04:25:23Z | 2026-04-13T04:30:22Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | zqchris | 4,436,110 | MDQ6VXNlcjQ0MzYxMTA= | User | false |
openclaw/openclaw | 4,011,850,860 | I_kwDOQb6kR87vH_xs | 31,839 | https://github.com/openclaw/openclaw/issues/31839 | https://api.github.com/repos/openclaw/openclaw/issues/31839 | Feature request: Provider-independent gateway health / per-agent process isolation | ## Problem
When a single LLM provider (e.g., Anthropic) experiences an outage, the gateway process can become unstable — boot checks fail, the process crash-loops, and **all agents become unreachable**, even those configured to use a different, healthy provider (e.g., Google Gemini).
## Current Architecture
The gate... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T15:27:10Z | 2026-04-16T04:25:22Z | 2026-04-13T04:30:25Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kai-ocean-kurt | 264,042,516 | U_kgDOD7z4FA | User | false |
openclaw/openclaw | 4,011,894,411 | I_kwDOQb6kR87vIKaL | 31,846 | https://github.com/openclaw/openclaw/issues/31846 | https://api.github.com/repos/openclaw/openclaw/issues/31846 | Ollama auto-discovery causes connection storm when target is unreachable (no backoff/caching) | **Version**: 2026.3.1
**Severity**: High (causes downstream resource exhaustion)
## Description
`discoverOllamaModels()` in `model-selection` probes `http://127.0.0.1:11434/api/tags` on every provider rebuild, even when:
- Ollama is not installed
- No `OLLAMA_API_KEY` or explicit Ollama config exists
- Previous atte... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T15:36:38Z | 2026-04-16T04:25:21Z | 2026-04-13T04:30:27Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | zqchris | 4,436,110 | MDQ6VXNlcjQ0MzYxMTA= | User | false |
openclaw/openclaw | 4,012,098,083 | I_kwDOQb6kR87vI8Ij | 31,883 | https://github.com/openclaw/openclaw/issues/31883 | https://api.github.com/repos/openclaw/openclaw/issues/31883 | Feature: configurable bootstrapFiles list for custom workspace file injection | ## Problem
The bootstrap file list is hardcoded to a fixed set: `AGENTS.md`, `SOUL.md`, `TOOLS.md`, `IDENTITY.md`, `USER.md`, `HEARTBEAT.md`, `BOOTSTRAP.md`, `MEMORY.md`. There is no way to add custom workspace files to the bootstrap injection without resorting to hacks like symlinking to one of the existing filenames... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T16:22:06Z | 2026-04-16T04:25:20Z | 2026-04-13T04:30:31Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Adam-Researchh | 261,827,313 | U_kgDOD5sq8Q | User | false |
openclaw/openclaw | 4,011,957,445 | I_kwDOQb6kR87vIZzF | 31,855 | https://github.com/openclaw/openclaw/issues/31855 | https://api.github.com/repos/openclaw/openclaw/issues/31855 | Gateway crashes on undici Fetch.onAborted (TypeError: terminated) | ## Bug Report\n\n**OpenClaw:** 2026.3.1\n**OS:** macOS 15.6 (arm64)\n**Node:** 25.5.0\n\n### Summary\nGateway process crashes due to an **uncaught exception** from Node's built-in undici fetch when a request is aborted/terminated. This causes the gateway to stop entirely and requires a manual restart.\n\n### When\nObse... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T15:49:54Z | 2026-04-16T04:25:20Z | 2026-04-13T04:30:29Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | emilywinslow-source | 260,186,948 | U_kgDOD4IjRA | User | false |
openclaw/openclaw | 4,012,132,316 | I_kwDOQb6kR87vJEfc | 31,893 | https://github.com/openclaw/openclaw/issues/31893 | https://api.github.com/repos/openclaw/openclaw/issues/31893 | Feature: Inject current date/time into session startup sequence | ## Problem
When a new session starts (via `/new` or `/reset`), the agent's startup sequence reads memory files and greets the user, but there is no explicit step to check the current date and time. The runtime metadata in the system prompt includes a timezone but not a reliable, prominent timestamp at session init.
T... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T16:29:13Z | 2026-04-16T04:25:19Z | 2026-04-13T04:30:33Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | drkyleahuggins | 262,856,883 | U_kgDOD6rgsw | User | false |
openclaw/openclaw | 4,013,470,899 | I_kwDOQb6kR87vOLSz | 32,200 | https://github.com/openclaw/openclaw/issues/32200 | https://api.github.com/repos/openclaw/openclaw/issues/32200 | WhatsApp Voice Transcription Not Triggering (2026.3.1) | # Bug Report: WhatsApp Voice Message Transcription Not Triggering
## Summary
WhatsApp voice messages arrive but transcription pipeline never triggers. Audio files are received and stored, but `tools.media.audio` scope rules are not being evaluated correctly (or audio is not being routed to transcription at all).
## ... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-02T21:50:44Z | 2026-04-19T04:24:32Z | 2026-04-16T04:25:18Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | tomdailey55 | 22,225,525 | MDQ6VXNlcjIyMjI1NTI1 | User | false |
openclaw/openclaw | 4,012,251,127 | I_kwDOQb6kR87vJhf3 | 31,933 | https://github.com/openclaw/openclaw/issues/31933 | https://api.github.com/repos/openclaw/openclaw/issues/31933 | [Bug]: Announce delivery timeout hardcoded at 60s, causing timeout failures for complex cron payloads | ## Description
The `announce` delivery mode for cron jobs has a hardcoded 60-second timeout for generating the announcement message. Complex cron payloads (e.g., daily reports that require multiple tool calls and data aggregation) frequently exceed this timeout, causing the delivery to fail silently.
## Steps to Repro... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T16:53:51Z | 2026-04-16T04:25:18Z | 2026-04-13T04:30:37Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nova-openclaw-cgk | 262,160,754 | U_kgDOD6BBcg | User | false |
openclaw/openclaw | 4,012,250,102 | I_kwDOQb6kR87vJhP2 | 31,931 | https://github.com/openclaw/openclaw/issues/31931 | https://api.github.com/repos/openclaw/openclaw/issues/31931 | [Bug]: resolveSandboxDockerConfig() drops dangerouslyAllow* fields from sandbox config | ## Description
`resolveSandboxDockerConfig()` in the sandbox module does not pass through `dangerouslyAllow*` fields (like `dangerouslyAllowLocalSrc`, `dangerouslyAllowOutboundNetwork`, etc.) when resolving Docker sandbox configuration. This means any `dangerouslyAllow*` settings in `openclaw.json` are silently ignored... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T16:53:39Z | 2026-04-16T04:25:18Z | 2026-04-13T04:30:35Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nova-openclaw-cgk | 262,160,754 | U_kgDOD6BBcg | User | false |
openclaw/openclaw | 4,012,342,445 | I_kwDOQb6kR87vJ3yt | 31,952 | https://github.com/openclaw/openclaw/issues/31952 | https://api.github.com/repos/openclaw/openclaw/issues/31952 | Telegram delivery queue recovery uses markdown-length chunking instead of HTML-length (follow-up to #29342) | ## Summary
PR #29342 (shipped in `2026.3.1`) fixed Telegram oversize message splitting by routing it through the shared outbound pipeline with HTML-aware chunking. However, the **delivery queue recovery path** (`recoverPendingDeliveries`) still uses the Telegram plugin's `chunker` which calls `chunkMarkdownText` — spl... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T17:11:55Z | 2026-04-16T04:25:17Z | 2026-04-13T04:30:39Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | rtmuller | 66,280,176 | MDQ6VXNlcjY2MjgwMTc2 | User | false |
openclaw/openclaw | 4,013,343,856 | I_kwDOQb6kR87vNsRw | 32,179 | https://github.com/openclaw/openclaw/issues/32179 | https://api.github.com/repos/openclaw/openclaw/issues/32179 | [Bug] Azure Foundry Anthropic: SSE stream events concatenated without \n\n delimiter — raw JSON parse error surfaced to all user channels | ## Description
When using the `azure-foundry-anthropic` provider (Azure AI Foundry Anthropic endpoint at `*.services.ai.azure.com/anthropic`) with `api: "anthropic-messages"`, the SSE stream occasionally delivers chunks where **two SSE events are concatenated without a proper `\n\n` delimiter** between them. The `@ant... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T21:16:51Z | 2026-04-25T04:19:11Z | 2026-04-22T04:30:13Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | yvesgagnongesys | 242,308,872 | U_kgDODnFXCA | User | false |
openclaw/openclaw | 4,012,365,840 | I_kwDOQb6kR87vJ9gQ | 31,955 | https://github.com/openclaw/openclaw/issues/31955 | https://api.github.com/repos/openclaw/openclaw/issues/31955 | feat: Add tools.exec.shellInit config for transparent tool integration (RTK-AI use case) | **Note:** This issue was drafted with AI assistance.
### Problem
The `exec` tool spawns non-interactive, non-login shell processes with no mechanism to:
- Initialize shell environment variables
- Prepend commands globally
- Load shell hooks or custom initialization scripts
This limits integration of external tools (l... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T17:16:23Z | 2026-04-16T04:25:16Z | 2026-04-13T04:30:41Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nSimonFR | 13,413,730 | MDQ6VXNlcjEzNDEzNzMw | User | false |
openclaw/openclaw | 4,012,371,731 | I_kwDOQb6kR87vJ-8T | 31,959 | https://github.com/openclaw/openclaw/issues/31959 | https://api.github.com/repos/openclaw/openclaw/issues/31959 | [Bug]: macOS Node Missing system.run.prepare Command | # Bug Report: macOS Node Missing `system.run.prepare` Command
## Environment
- **OpenClaw Version**: 2026.3.1
- **Gateway Platform**: Linux (Ubuntu)
- **Node Platform**: macOS 26.3.0 (Mac mini, Mac14,12)
- **Node Version**: 2026.3.1
## Description
The macOS companion app does not report the `system.run.prepare` com... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T17:17:29Z | 2026-04-16T04:25:15Z | 2026-04-13T04:30:42Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kongwy | 9,990,631 | MDQ6VXNlcjk5OTA2MzE= | User | false |
openclaw/openclaw | 4,013,644,551 | I_kwDOQb6kR87vO1sH | 32,245 | https://github.com/openclaw/openclaw/issues/32245 | https://api.github.com/repos/openclaw/openclaw/issues/32245 | [Bug]: Vertex embedded runs fail with "Cannot convert undefined or null to object" on v2026.3.1 (Telegram + cron) | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to OpenClaw `v2026.3.1`, Vertex-backed embedded runs fail with:
`Cannot convert undefined or null to object`
This affects both Telegram group replies and Vertex-based cron jobs.
Non-Vertex models (e.g. `openai-codex/...`, OpenRouter) c... | closed | not_planned | true | 20 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-02T22:32:40Z | 2026-04-16T04:25:14Z | 2026-04-13T04:30:50Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Frank683456 | 49,425,469 | MDQ6VXNlcjQ5NDI1NDY5 | User | false |
openclaw/openclaw | 4,013,640,901 | I_kwDOQb6kR87vO0zF | 32,244 | https://github.com/openclaw/openclaw/issues/32244 | https://api.github.com/repos/openclaw/openclaw/issues/32244 | Cron任务状态误报：显示error但实际执行成功 | ## 问题描述
Cron任务的状态记录（lastRunStatus）持续显示 "error"，但实际上任务执行成功，文件已写入、消息已送送送送。
## 复环步骤
1. 配置cron任务，执行写入文件操作（如保存新闻报告）
2. 任务执行后检查jobs.json中的state字段
## 实际表現
| 字段 | 值 |
|------|------|
| lastRunStatus | "error" |
| lastError | “✦ ✍ Write: to ~/文档/.../20260303.md (6827 chars) failed” |
| lastDelivered | true |
| 文件实际存在 | ✅ 写... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T22:31:50Z | 2026-04-16T04:25:14Z | 2026-04-13T04:30:48Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | xiaomei-01 | 262,327,818 | U_kgDOD6LOCg | User | false |
openclaw/openclaw | 4,013,651,004 | I_kwDOQb6kR87vO3Q8 | 32,248 | https://github.com/openclaw/openclaw/issues/32248 | https://api.github.com/repos/openclaw/openclaw/issues/32248 | Proposal: Optional OpenMork integration path (docs + adapter skeleton, opt-in) | ## Summary
Propose an optional **OpenMork provider/integration path** for OpenClaw, focused on uncensored local-first execution with strict isolation and explicit opt-in.
## Problem
Some users want a local-first, privacy-centric runtime profile with configurable model backends and minimal friction for autonomous tasks... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T22:34:11Z | 2026-04-16T04:25:13Z | 2026-04-13T04:30:52Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | openmork | 264,830,120 | U_kgDOD8j8qA | User | false |
openclaw/openclaw | 4,013,654,628 | I_kwDOQb6kR87vO4Jk | 32,250 | https://github.com/openclaw/openclaw/issues/32250 | https://api.github.com/repos/openclaw/openclaw/issues/32250 | Support HMAC signature verification as alternative to bearer token auth for hooks | ## Problem
External webhooks (GitHub, Mercury, Stripe, etc.) sign payloads with HMAC-SHA256 but cannot add bearer tokens to their requests. The current hooks system requires bearer token authentication, which means external webhook providers cannot directly invoke OpenClaw hooks.
## Current Workaround
Running a sepa... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-02T22:35:03Z | 2026-04-16T04:25:12Z | 2026-04-13T04:30:54Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | m-mohamed | 43,361,569 | MDQ6VXNlcjQzMzYxNTY5 | User | false |
openclaw/openclaw | 4,012,858,905 | I_kwDOQb6kR87vL14Z | 32,075 | https://github.com/openclaw/openclaw/issues/32075 | https://api.github.com/repos/openclaw/openclaw/issues/32075 | Mattermost: chatmode=onmessage does not disable mention requirement in channel messages | ## Bug Report
### Summary
Setting `chatmode: "onmessage"` in the Mattermost channel config does not cause the bot to respond to channel messages without an @mention, despite the docs stating it should.
### Environment
- OpenClaw version: 2026.3.1
- Mattermost plugin version: 2026.2.26
- Mattermost server: Team Edit... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T19:16:16Z | 2026-04-25T04:19:12Z | 2026-04-22T04:30:11Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | russellweiss | 44,983,860 | MDQ6VXNlcjQ0OTgzODYw | User | false |
openclaw/openclaw | 4,013,765,971 | I_kwDOQb6kR87vPTVT | 32,277 | https://github.com/openclaw/openclaw/issues/32277 | https://api.github.com/repos/openclaw/openclaw/issues/32277 | [Bug] Memory indexer doesn't detect embedding dimension mismatch between stored vectors and active provider | ## Problem
When switching memory search providers (e.g., from local `embeddinggemma-300M` fallback to Gemini `gemini-embedding-001`), the existing vector database contains embeddings with a different dimension (300 vs 3072). The indexer:
1. Doesn't detect the dimension mismatch
2. Doesn't warn the user
3. Returns zer... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T23:02:07Z | 2026-04-16T04:25:11Z | 2026-04-13T04:30:57Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Asclepeion | 169,803,573 | U_kgDOCh7_NQ | User | false |
openclaw/openclaw | 4,013,661,088 | I_kwDOQb6kR87vO5ug | 32,252 | https://github.com/openclaw/openclaw/issues/32252 | https://api.github.com/repos/openclaw/openclaw/issues/32252 | [Bug]: Nostr plugin — three bugs prevent DM delivery | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The Nostr plugin (`@openclaw/nostr` 2026.3.1) has three bugs that together prevent it from receiving NIP-04 DMs and sending replies: (1) double-nested subscription filter causes relays to reject the REQ, (2) `startAccount` resolves immediat... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-02T22:36:38Z | 2026-04-16T04:25:11Z | 2026-04-13T04:30:56Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | chrysolambda-ops | 264,110,048 | U_kgDOD73_4A | User | false |
openclaw/openclaw | 4,013,776,961 | I_kwDOQb6kR87vPWBB | 32,281 | https://github.com/openclaw/openclaw/issues/32281 | https://api.github.com/repos/openclaw/openclaw/issues/32281 | [Feature]: Sub-agent completion chaining (onComplete) | # [Feature]: Sub-agent completion chaining (`onComplete`)
## Summary
Add an optional `onComplete` field to `sessions_spawn` that automatically dispatches a follow-up sub-agent when the current one succeeds.
## Problem
Sequential sub-agent workflows (A finishes → B starts) currently require one of two workarounds:
... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T23:04:52Z | 2026-04-16T04:25:10Z | 2026-04-13T04:30:59Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | solosage1 | 158,075,162 | U_kgDOCWwJGg | User | false |
openclaw/openclaw | 4,013,905,503 | I_kwDOQb6kR87vP1Zf | 32,303 | https://github.com/openclaw/openclaw/issues/32303 | https://api.github.com/repos/openclaw/openclaw/issues/32303 | Dashboard: render suggested replies/commands as click-to-insert chips | Problem\n- When the assistant asks for a specific response phrase (e.g. "Reply with: YES keep purge" or "Choose one: SOURCE / DOC / HYBRID"), users currently have to copy/paste manually.\n- On some surfaces, silent turns (NO_REPLY) or system prompts make this extra confusing.\n\nRequest\n- In the OpenClaw Dashboard cha... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T23:42:39Z | 2026-04-16T04:25:09Z | 2026-04-13T04:31:02Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | lanciter | 159,670,295 | U_kgDOCYRgFw | User | false |
openclaw/openclaw | 4,013,903,949 | I_kwDOQb6kR87vP1BN | 32,301 | https://github.com/openclaw/openclaw/issues/32301 | https://api.github.com/repos/openclaw/openclaw/issues/32301 | Feature: Per-channel model overrides in Discord guild config | ## Summary
Currently `channels.discord.guilds.<guildId>` only supports `requireMention`. Would love per-channel model routing.
## Proposed config shape
```json
{
"channels": {
"discord": {
"guilds": {
"<guildId>": {
"requireMention": true,
"channelModels": {
"<channe... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-02T23:42:02Z | 2026-04-16T04:25:09Z | 2026-04-13T04:31:01Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | oclawtgk30-afk | 264,758,914 | U_kgDOD8fmgg | User | false |
openclaw/openclaw | 4,013,956,288 | I_kwDOQb6kR87vQBzA | 32,319 | https://github.com/openclaw/openclaw/issues/32319 | https://api.github.com/repos/openclaw/openclaw/issues/32319 | Canvas WebView: external links should open in system browser | ## Problem
Links in the Canvas WebView (A2UI) open inside the WebView itself. There's no way to navigate back to the canvas content after clicking a link, and the user gets stuck.
## Expected Behavior
Links with `target="_blank"` (or all non-canvas links) should open in the device's default browser (e.g. Chrome) ins... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T00:00:26Z | 2026-04-16T04:25:08Z | 2026-04-13T04:31:04Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | davelutztx | 48,866,029 | MDQ6VXNlcjQ4ODY2MDI5 | User | false |
openclaw/openclaw | 4,013,984,686 | I_kwDOQb6kR87vQIuu | 32,323 | https://github.com/openclaw/openclaw/issues/32323 | https://api.github.com/repos/openclaw/openclaw/issues/32323 | Cron任务在Gateway重启后未正确触发 | 问题描述：Gateway重启后，cron定时任务未在预定时间触发。
复现步骤：
1. 设置cron任务：每天8点执行
2. 重启Gateway
3. 观察第二天8点任务是否触发
日志证据：重启后日志显示 timer被错误地设置为60秒（clamped=true），而不是正确的调度时间。
环境：Windows_NT, OpenClaw最新
期望行为：Gateway重启后，cron任务应按原定计划时间触发。 | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T00:11:46Z | 2026-04-16T04:25:07Z | 2026-04-13T04:31:06Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | ZhangZhigang7627 | 188,031,052 | U_kgDOCzUgTA | User | false |
openclaw/openclaw | 4,010,908,014 | I_kwDOQb6kR87vEZlu | 31,649 | https://github.com/openclaw/openclaw/issues/31649 | https://api.github.com/repos/openclaw/openclaw/issues/31649 | Discord EventQueue still times out after network improved (MESSAGE_CREATE 30s timeout + Unknown interaction) | ## Summary
Even after switching to a faster proxy (network-layer errors dropped), Discord remains unstable due to EventQueue/listener timeouts.
## Environment
- OpenClaw: 2026.2.26
- Channel: Discord
- Host: macOS (arm64)
- Config highlights:
- `channels.discord.streaming = off`
- `commands.native = true`
- `com... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T12:02:40Z | 2026-04-25T04:19:13Z | 2026-04-22T04:30:07Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ZaneChen76 | 70,149,446 | MDQ6VXNlcjcwMTQ5NDQ2 | User | false |
openclaw/openclaw | 4,014,037,753 | I_kwDOQb6kR87vQVr5 | 32,341 | https://github.com/openclaw/openclaw/issues/32341 | https://api.github.com/repos/openclaw/openclaw/issues/32341 | [Bug] WhatsApp: sendPolicy 'deny' blocks message persistence/ingestion in Self-Chat Mode | **Description**
I am attempting to configure OpenClaw (v2026.3.1) to ingest WhatsApp group messages for logging/context *without* the agent replying (a "Spy Mode").
**Configuration**
- **Version:** 2026.3.1
- **Platform:** macOS
- **Mode:** WhatsApp Self-Chat Mode (bot uses same number as user).
- **Tools:** `tools.se... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T00:32:46Z | 2026-04-16T04:25:06Z | 2026-04-13T04:31:10Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | roman-sh | 29,175,447 | MDQ6VXNlcjI5MTc1NDQ3 | User | false |
openclaw/openclaw | 4,013,989,518 | I_kwDOQb6kR87vQJ6O | 32,326 | https://github.com/openclaw/openclaw/issues/32326 | https://api.github.com/repos/openclaw/openclaw/issues/32326 | Telegram: 'Failed to download media' when attaching files in group topics | ## Description
Attaching a `.md` file (document) in a **Telegram group topic** intermittently triggers:
> ⚠️ Failed to download media. Please try again.
The bot replies with this error instead of processing the file.
## Environment
- OpenClaw version: `2026.3.1`
- Channel: Telegram
- Chat type: Group with topics (f... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-03T00:13:38Z | 2026-04-16T04:25:06Z | 2026-04-13T04:31:08Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | a692570 | 182,830,946 | U_kgDOCuXHYg | User | false |
openclaw/openclaw | 4,014,049,097 | I_kwDOQb6kR87vQYdJ | 32,343 | https://github.com/openclaw/openclaw/issues/32343 | https://api.github.com/repos/openclaw/openclaw/issues/32343 | Feature: thread-info action for Discord (read thread metadata before posting) | Right now if you have a script that posts to forum threads, there's no way to check whether a thread is archived before you post to it. The only option is raw API calls. Posting to an archived Discord thread reopens it — which is exactly the wrong behavior if the user intentionally closed it.
The `channel-edit` action... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T00:37:17Z | 2026-04-16T04:25:05Z | 2026-04-13T04:31:11Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | danielbrodie | 11,337,994 | MDQ6VXNlcjExMzM3OTk0 | User | false |
openclaw/openclaw | 4,010,123,941 | I_kwDOQb6kR87vBaKl | 31,510 | https://github.com/openclaw/openclaw/issues/31510 | https://api.github.com/repos/openclaw/openclaw/issues/31510 | `openclaw agents list` should expose tool deny / sandbox status | ## Summary
`openclaw agents list` (and `--json`) does not expose whether an agent has `exec` blocked via `tools.deny` or sandbox mode enabled. This makes it impossible to audit agent capabilities from the CLI without manually parsing `openclaw.json`.
## Current behavior
```
$ openclaw agents list
- main (default) (H... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-02T09:07:42Z | 2026-04-25T04:19:14Z | 2026-04-22T04:30:02Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | tmchow | 517,103 | MDQ6VXNlcjUxNzEwMw== | User | false |
openclaw/openclaw | 4,014,059,750 | I_kwDOQb6kR87vQbDm | 32,348 | https://github.com/openclaw/openclaw/issues/32348 | https://api.github.com/repos/openclaw/openclaw/issues/32348 | [Linux/systemd] Gateway restart from within agent session causes SIGTERM crash loop | ## Summary
On Linux with systemd, triggering a gateway restart from within a running agent session causes the session to be killed via SIGTERM, often leading to a crash loop.
## Environment
- OpenClaw: 2026.3.1
- OS: Ubuntu 24.04 (Linux 6.17.0)
- Install: npm global (`sudo npm i -g openclaw`)
- Gateway: systemd user s... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T00:41:47Z | 2026-04-16T04:25:04Z | 2026-04-13T04:31:13Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | mohfoda1982-create | 252,742,484 | U_kgDODxCLVA | User | false |
openclaw/openclaw | 4,014,073,638 | I_kwDOQb6kR87vQecm | 32,362 | https://github.com/openclaw/openclaw/issues/32362 | https://api.github.com/repos/openclaw/openclaw/issues/32362 | Telegram document attachments (xlsx/Excel) not staged to sandbox workspace | ## Environment
- Openclaw version: 2026.3.1
- Channel: Telegram
- Sandbox mode: `all`, `workspaceAccess: "ro"`
- Docker image: `openclaw-sandbox:bookworm-slim`
## Problem
When a user sends an **Excel (.xlsx) file** via Telegram, the file is downloaded correctly to `~/.openclaw/media/inbound/`, but it is **never stage... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T00:47:27Z | 2026-04-16T04:25:03Z | 2026-04-13T04:31:15Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | boje1123-del | 253,819,830 | U_kgDODyD7tg | User | false |
openclaw/openclaw | 4,014,239,797 | I_kwDOQb6kR87vRHA1 | 32,403 | https://github.com/openclaw/openclaw/issues/32403 | https://api.github.com/repos/openclaw/openclaw/issues/32403 | Channel adapter leaks internal NO_REPLY/memory-flush artifacts as visible user messages | ## Problem
Internal maintenance/silent-response artifacts are being rendered to end users as visible chat messages (examples observed: `NO_REPLY`, `NO`, `None`, and transient typing bubbles with no content).
## Repro context
- During maintenance prompts (e.g., memory flush) the agent emits `NO_REPLY` as instructed.
- ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T01:54:20Z | 2026-04-16T04:25:02Z | 2026-04-13T04:31:19Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | michaelrobertsutton | 12,904,582 | MDQ6VXNlcjEyOTA0NTgy | User | false |
openclaw/openclaw | 4,014,235,652 | I_kwDOQb6kR87vRGAE | 32,400 | https://github.com/openclaw/openclaw/issues/32400 | https://api.github.com/repos/openclaw/openclaw/issues/32400 | [Bug]: Control UI hangs when WebSocket disconnects during long streaming responses | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
- **Problem**: When the WebSocket connection drops during a long streaming response in the Control UI (webchat), the UI hangs in "sending/thinking" state and does not recover automatically.
- **Why it matters**: Users cannot send new messag... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T01:52:40Z | 2026-04-16T04:25:02Z | 2026-04-13T04:31:17Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | lincoln310 | 7,071,490 | MDQ6VXNlcjcwNzE0OTA= | User | false |
openclaw/openclaw | 4,014,240,399 | I_kwDOQb6kR87vRHKP | 32,404 | https://github.com/openclaw/openclaw/issues/32404 | https://api.github.com/repos/openclaw/openclaw/issues/32404 | fs-safe hardcodes 0600 file mode, bypasses umask and default ACLs | ## Problem
The `fs-safe` module hardcodes `fs.open(path, flags, 384)` (octal `0600`, owner-only read/write) for all file writes. This bypasses both the system **umask** and any **default POSIX ACLs** set on parent directories.
**Affected code** (e.g. `dist/fs-safe-bED949rs.js:184-187`):
```js
handle = await fs$1.open... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T01:54:36Z | 2026-04-16T04:25:01Z | 2026-04-13T04:31:21Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | jeanmonet | 15,112,920 | MDQ6VXNlcjE1MTEyOTIw | User | false |
openclaw/openclaw | 4,014,263,187 | I_kwDOQb6kR87vRMuT | 32,409 | https://github.com/openclaw/openclaw/issues/32409 | https://api.github.com/repos/openclaw/openclaw/issues/32409 | Add config option for default reasoning mode | Currently, the `/reasoning` command is a session-level runtime switch that defaults to `off`. Users need to manually enable it each session.
It would be helpful to have a config option like:
```json5
{
agents: {
defaults: {
reasoning: "stream" // off | on | stream
}
}
}
```
This would allow users ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T02:02:50Z | 2026-04-16T04:25:00Z | 2026-04-13T04:31:23Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | xingzihai | 87,858,694 | MDQ6VXNlcjg3ODU4Njk0 | User | false |
openclaw/openclaw | 4,009,226,006 | I_kwDOQb6kR87u9-8W | 31,278 | https://github.com/openclaw/openclaw/issues/31278 | https://api.github.com/repos/openclaw/openclaw/issues/31278 | Feature Request: Per-Model Context Tokens Configuration | ## Summary
Add `contextTokens` field to `agents.defaults.models["provider/model"]` entries, allowing users to configure different context window limits for different models.
## Problem to solve
Currently, context window configuration is limited to:
1. **Global fallback**: `agents.defaults.model.contextTokens` (sing... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-02T04:33:49Z | 2026-04-25T04:19:14Z | 2026-04-22T04:30:01Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | xingzihai | 87,858,694 | MDQ6VXNlcjg3ODU4Njk0 | User | false |
openclaw/openclaw | 4,014,317,541 | I_kwDOQb6kR87vRZ_l | 32,430 | https://github.com/openclaw/openclaw/issues/32430 | https://api.github.com/repos/openclaw/openclaw/issues/32430 | [Bug]: Output Pipeline Race Condition - System Messages Preempt User Interactions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Description:
When user messages and system messages (memory flush, cron tasks, scheduled reminders) trigger concurrently, the output pipeline gets preempted, causing:
1. User receives incorrect response (e.g., NO_REPLY intended for syst... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T02:22:25Z | 2026-04-16T04:24:59Z | 2026-04-13T04:31:24Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | moyao-miao | 95,124,633 | U_kgDOBat8mQ | User | false |
openclaw/openclaw | 4,014,380,480 | I_kwDOQb6kR87vRpXA | 32,444 | https://github.com/openclaw/openclaw/issues/32444 | https://api.github.com/repos/openclaw/openclaw/issues/32444 | Auth profile round-robin broken on /new — always picks first profile | ## Summary
`/new` or `/reset` always selects the same auth profile (the one with the oldest `lastUsed`) instead of rotating between configured profiles.
## Environment
- OpenClaw 2026.3.1
- macOS (arm64)
- Two Anthropic token profiles configured: `anthropic:default` and `anthropic:secondary`
- No `auth.order` explic... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T02:48:57Z | 2026-04-16T04:24:58Z | 2026-04-13T04:31:28Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | lunanova-core | 258,436,214 | U_kgDOD2dsdg | User | false |
openclaw/openclaw | 4,014,318,985 | I_kwDOQb6kR87vRaWJ | 32,431 | https://github.com/openclaw/openclaw/issues/32431 | https://api.github.com/repos/openclaw/openclaw/issues/32431 | Feature Request: delivery.mode='main-only' for isolated cron jobs (announce to main session without channel delivery) | ## Problem
When running isolated cron jobs in a **group chat** context, the current delivery modes create an awkward UX:
- `delivery.mode = "announce"` delivers results to **both** the target channel (e.g., Telegram group) **and** posts a summary to the main session. This causes a "two people talking" problem — the g... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T02:23:05Z | 2026-04-16T04:24:58Z | 2026-04-13T04:31:26Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | renshaojie123-design | 260,354,309 | U_kgDOD4SxBQ | User | false |
openclaw/openclaw | 4,014,403,749 | I_kwDOQb6kR87vRvCl | 32,448 | https://github.com/openclaw/openclaw/issues/32448 | https://api.github.com/repos/openclaw/openclaw/issues/32448 | doctor --fix: service PATH includes non-existent directories (~/.npm-global/bin, ~/.nvm/current/bin) | # `doctor --fix` adds non-existent directories to service PATH
**Version:** 2026.3.1
**File:** `src/daemon/service-env.ts`
## Problem
`addCommonUserBinDirs()` and `resolveLinuxUserBinDirs()` unconditionally add hardcoded paths to the systemd service file's `PATH` without checking if they exist. This causes two issu... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T02:59:25Z | 2026-04-16T04:24:57Z | 2026-04-13T04:31:30Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | cliffchen | 7,713,792 | MDQ6VXNlcjc3MTM3OTI= | User | false |
openclaw/openclaw | 4,014,404,570 | I_kwDOQb6kR87vRvPa | 32,449 | https://github.com/openclaw/openclaw/issues/32449 | https://api.github.com/repos/openclaw/openclaw/issues/32449 | [Bug]: Chrome Extension Relay Authentication Failing with 401 - Valid HMAC Token Rejected | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
This is my first bug report prepared by Wren - my lovely OpenClaw
## Summary
The OpenClaw Chrome extension relay server (port 18791) rejects valid HMAC-SHA256 derived authentication tokens with 401 Unauthorized, preventing the browser ext... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T02:59:48Z | 2026-04-16T04:24:56Z | 2026-04-13T04:31:32Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | elizalin00 | 265,170,034 | U_kgDOD84scg | User | false |
openclaw/openclaw | 4,014,432,003 | I_kwDOQb6kR87vR18D | 32,458 | https://github.com/openclaw/openclaw/issues/32458 | https://api.github.com/repos/openclaw/openclaw/issues/32458 | Bug: Soft restart (SIGUSR1) does not kill previously-launched browser processes, causing port conflicts | ## Summary
On gateway soft restart (`SIGUSR1`), OpenClaw does not kill browser processes it previously launched. When the gateway restarts and spawns a new Chrome instance, the old one is still running on the same port. This causes a port conflict where the CDP WebSocket accepts connections but silently drops all comm... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T03:10:28Z | 2026-04-16T04:24:55Z | 2026-04-13T04:31:35Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | jilagan | 1,145,673 | MDQ6VXNlcjExNDU2NzM= | User | false |
openclaw/openclaw | 4,014,409,363 | I_kwDOQb6kR87vRwaT | 32,452 | https://github.com/openclaw/openclaw/issues/32452 | https://api.github.com/repos/openclaw/openclaw/issues/32452 | [Bug]: GGML Metal unload assertion crash on Apple Silicon when using local embeddings (rsets->data count != 0) | ### Bug type
Regression (worked before, now fails)
### Summary
On Apple Silicon (M-series), when using `"provider": "local"` for memory search embeddings (via node-llama-cpp), graceful shutdown (Ctrl+C / SIGINT) causes a GGML Metal assertion crash: GGML_ASSERT(... rsets->data count == 0 ...) // or similar Metal unl... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T03:01:45Z | 2026-04-16T04:24:55Z | 2026-04-13T04:31:34Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | AnilDaoud | 3,533,332 | MDQ6VXNlcjM1MzMzMzI= | User | false |
openclaw/openclaw | 4,014,447,122 | I_kwDOQb6kR87vR5oS | 32,466 | https://github.com/openclaw/openclaw/issues/32466 | https://api.github.com/repos/openclaw/openclaw/issues/32466 | Feature Request: Per-Agent Instructions File (INSTRUCTIONS.md) for persistent behavioral rules and sub-agent management | # Feature Request: Per-Agent Instructions File (AGENT_INSTRUCTIONS.md)
## Summary
Add support for a `.md` file (e.g., `AGENT_INSTRUCTIONS.md` or `INSTRUCTIONS.md`) that an agent reads on every message/turn. This file would contain persistent instructions that override or supplement the agent's default behavior.
## U... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T03:16:25Z | 2026-04-16T04:24:54Z | 2026-04-13T04:31:37Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | joemcurry | 1,192,832 | MDQ6VXNlcjExOTI4MzI= | User | false |
openclaw/openclaw | 4,008,791,077 | I_kwDOQb6kR87u8Uwl | 31,136 | https://github.com/openclaw/openclaw/issues/31136 | https://api.github.com/repos/openclaw/openclaw/issues/31136 | feat(feishu): Add streamingThrottleMs config parameter for streaming card update frequency | ## Problem
The streaming card update frequency `updateThrottleMs` was originally hardcoded at 100ms in the source code, making it impossible for users to configure via config. This leads to:
- High API call consumption
- Inability to adjust update frequency based on actual needs
## Solution
Add a `streamingThrottle... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-02T01:25:38Z | 2026-04-25T04:19:15Z | 2026-04-22T04:29:57Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | mosslee2042 | 264,580,147 | U_kgDOD8UsMw | User | false |
openclaw/openclaw | 4,014,464,856 | I_kwDOQb6kR87vR99Y | 32,472 | https://github.com/openclaw/openclaw/issues/32472 | https://api.github.com/repos/openclaw/openclaw/issues/32472 | [Feishu] 群消息 GroupSubject 应传递群名而非 chat_id | ## 问题描述
当前飞书群消息的 `GroupSubject` 传递的是 `chat_id`（如 `oc_d3c6404039344afa4d0e669460356aba`），而不是实际的群名称。
## 问题定位
`src/bot.ts` 第 1157 行：
```typescript
GroupSubject: isGroup ? ctx.chatId : undefined,
```
## 影响
- OpenClaw 会话列表显示的是 `oc_xxx` 这样的 ID，不够友好
- 用户难以区分不同的群聊会话
## 建议方案
在收到群消息时，调用 `getChatInfo` API 获取群信息，将 `GroupSub... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T03:24:36Z | 2026-04-16T04:24:53Z | 2026-04-13T04:31:39Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | sixgodIT | 5,539,093 | MDQ6VXNlcjU1MzkwOTM= | User | false |
openclaw/openclaw | 4,014,522,337 | I_kwDOQb6kR87vSL_h | 32,495 | https://github.com/openclaw/openclaw/issues/32495 | https://api.github.com/repos/openclaw/openclaw/issues/32495 | [Feature Request] Support switching between agents in Control UI | ## Problem
Currently, the OpenClaw Control UI (web dashboard) only connects to the default `main` agent. Users with multiple agents (e.g., code-agent, docs-agent, test-agent, claude) cannot easily switch between them in the same web interface.
## Current Behavior
- Control UI at `http://127.0.0.1:18789/` always conn... | closed | not_planned | true | 9 | [
"stale"
] | [] | 2026-03-03T03:48:25Z | 2026-04-16T04:24:52Z | 2026-04-13T04:31:45Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | survivor998 | 226,792,843 | U_kgDODYSViw | User | false |
openclaw/openclaw | 4,014,499,336 | I_kwDOQb6kR87vSGYI | 32,488 | https://github.com/openclaw/openclaw/issues/32488 | https://api.github.com/repos/openclaw/openclaw/issues/32488 | [Bug]: Docker image 2026.3.1 and 2026.3.1-beta.1 share identical digest but binary reports beta.1, triggering false "Update available" banner | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
**Environment**
- Install method: Docker (ghcr.io/openclaw/openclaw)
- Running version: 2026.3.1-beta.1 (self-reported by binary)
- Docker tag pulled: ghcr.io/openclaw/openclaw:2026.3.1
**Issue**
The Control UI displays "Update available: ... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T03:39:12Z | 2026-04-16T04:24:52Z | 2026-04-13T04:31:42Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | BMA-Aaron | 160,290,273 | U_kgDOCY3V4Q | User | false |
openclaw/openclaw | 4,006,486,032 | I_kwDOQb6kR87uziAQ | 30,390 | https://github.com/openclaw/openclaw/issues/30390 | https://api.github.com/repos/openclaw/openclaw/issues/30390 | [Feature]: Add one-click copy button for code blocks in WebChat | ## Summary
When sharing commands or code snippets in OpenClaw's WebChat interface, users must manually select text and copy it. This is cumbersome, especially on mobile devices or when copying long commands.
## Proposed solution
Add a copy button to code blocks in WebChat, similar to GitHub, Discord, and other platf... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-01T06:47:04Z | 2026-04-19T04:24:33Z | 2026-04-16T04:24:50Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | AceOld | 264,633,444 | U_kgDOD8X8ZA | User | false |
openclaw/openclaw | 4,014,519,848 | I_kwDOQb6kR87vSLYo | 32,493 | https://github.com/openclaw/openclaw/issues/32493 | https://api.github.com/repos/openclaw/openclaw/issues/32493 | [Bug]: CLI's Slack JSON manifest is framed in breaking characters | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When you go through the CLI to connect with a Slack channel, it will offer a JSON manifest for faster setup. But that manifest is framed by breaking characters. So for a user to copy the manifest, they would have to delete all those pipes.
... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T03:47:31Z | 2026-04-16T04:24:47Z | 2026-04-13T04:31:44Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | theDanielJLewis | 154,936 | MDQ6VXNlcjE1NDkzNg== | User | false |
openclaw/openclaw | 4,041,747,543 | I_kwDOQb6kR87w6CxX | 40,107 | https://github.com/openclaw/openclaw/issues/40107 | https://api.github.com/repos/openclaw/openclaw/issues/40107 | For God's sake provide LM Studio in onboarding! | ### Summary
Please add "LM Studio" as provider for local LLM in the onboarding. It's such a pita to set it up every time something breaks.
### Problem to solve
-LM Studio is not on the list as a provider for running local LLMs
-editing the JSON file with correct values is such a pain everytime something breaks
### ... | closed | completed | true | 0 | [
"enhancement"
] | [] | 2026-03-08T17:17:29Z | 2026-04-16T04:24:46Z | 2026-04-13T07:22:46Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | technerd-official | 143,012,532 | U_kgDOCIYytA | User | false |
openclaw/openclaw | 4,015,937,074 | I_kwDOQb6kR87vXlYy | 32,898 | https://github.com/openclaw/openclaw/issues/32898 | https://api.github.com/repos/openclaw/openclaw/issues/32898 | 功能建议：原生支持 LM Studio 本地模型配置 | ## 功能建议：原生支持 LM Studio 本地模型
### 背景
LM Studio 是流行的本地模型运行工具，提供 OpenAI 兼容 API。很多用户希望用本地模型降低成本和保护隐私。
### 已验证的工作方案
我在 OpenClaw 中成功配置了 LM Studio，配置如下：
```json
"lm-studio": {
"baseUrl": "http://192.168.11.193:8577/v1",
"api": "openai-completions",
"models": [
{
"id": "qwen/qwen3-coder-next",
"name": ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-03T10:37:07Z | 2026-04-16T04:24:45Z | 2026-04-13T07:22:47Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | SonicBotMan | 258,392,144 | U_kgDOD2bAUA | User | false |
openclaw/openclaw | 4,017,750,504 | I_kwDOQb6kR87vegHo | 33,291 | https://github.com/openclaw/openclaw/issues/33291 | https://api.github.com/repos/openclaw/openclaw/issues/33291 | [Feature] Feishu Bitable View Management API | ## Feature Request: Feishu Bitable View Management
### Description
Add support for managing Bitable views in the Feishu plugin.
### Background
- Feishu API supports view management: `POST /open-apis/bitable/v1/apps/{app_token}/tables/{table_id}/views`
- Ref: https://open.feishu.cn/document/server-docs/docs/bitable-v1... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-03T17:02:27Z | 2026-04-16T04:24:42Z | 2026-04-13T10:02:27Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | xiangyingchang | 10,992,221 | MDQ6VXNlcjEwOTkyMjIx | User | false |
openclaw/openclaw | 4,025,411,748 | I_kwDOQb6kR87v7uik | 35,300 | https://github.com/openclaw/openclaw/issues/35300 | https://api.github.com/repos/openclaw/openclaw/issues/35300 | Webchat Heartbeat Message Incorrectly Delivered to Feishu | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When OpenClaw's webchat heartbeat polling triggers an agent response, the system incorrectly sets the delivery context to the Feishu channel and attempts to deliver messages to `heartbeat` as a Feishu user ID, causing the Feishu API to retu... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-05T03:09:17Z | 2026-04-16T04:24:29Z | 2026-04-13T17:20:57Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Yaemikoreal | 141,380,188 | U_kgDOCG1KXA | User | false |
openclaw/openclaw | 4,003,177,656 | I_kwDOQb6kR87um6S4 | 29,186 | https://github.com/openclaw/openclaw/issues/29186 | https://api.github.com/repos/openclaw/openclaw/issues/29186 | sessions_spawn: gateway timeout too tight (10s) for embedded mode under load | ## Bug
The `sessions_spawn` tool dispatches sub-agent runs via `callGateway()` with a **hardcoded 10-second timeout** (`timeoutMs: 10_000` at `src/agents/tools/sessions-spawn-tool.ts:270`). In embedded gateway mode (e.g., Docker Compose), this is a loopback WebSocket call to the same process.
When the gateway is unde... | closed | not_planned | true | 8 | [
"stale"
] | [] | 2026-02-27T22:42:16Z | 2026-04-19T04:24:35Z | 2026-04-16T04:24:28Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | francisbrero | 2,491,181 | MDQ6VXNlcjI0OTExODE= | User | false |
openclaw/openclaw | 4,062,578,677 | I_kwDOQb6kR87yJgf1 | 43,764 | https://github.com/openclaw/openclaw/issues/43764 | https://api.github.com/repos/openclaw/openclaw/issues/43764 | BlueBubbles: reply threading silently degrades when Private API cache expires | > **Draft — validating fix before requesting upstream merge**
## Summary
BlueBubbles reply threading (iMessage `selectedMessageGuid`) silently degrades to plain sends after the server info cache expires (~10 minutes). The `message` tool returns `{"ok": true}` with no indication that threading was dropped.
## Root Ca... | closed | completed | true | 1 | [] | [] | 2026-03-12T06:45:40Z | 2026-04-16T04:24:27Z | 2026-04-13T18:03:49Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | omarshahine | 10,343,873 | MDQ6VXNlcjEwMzQzODcz | User | false |
openclaw/openclaw | 4,002,797,545 | I_kwDOQb6kR87uldfp | 29,124 | https://github.com/openclaw/openclaw/issues/29124 | https://api.github.com/repos/openclaw/openclaw/issues/29124 | No observability into message processing state (accepted → running → delivered/failed) | ## Problem
When the gateway processes an inbound message, there is no way to know what state it is in — whether it was accepted, is currently being processed by the AI, was successfully delivered, or failed silently.
This makes it difficult to diagnose:
- "I sent a message but never got a reply" — was it accepted? Di... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-02-27T20:45:24Z | 2026-04-19T04:24:36Z | 2026-04-16T04:24:27Z | CONTRIBUTOR | null | 20260419T052035Z | 2026-04-19T05:20:35Z | nohat | 838,027 | MDQ6VXNlcjgzODAyNw== | User | false |
openclaw/openclaw | 4,028,601,144 | I_kwDOQb6kR87wH5M4 | 36,260 | https://github.com/openclaw/openclaw/issues/36260 | https://api.github.com/repos/openclaw/openclaw/issues/36260 | Feature: Include Telegram forum topic name in inbound metadata | ## Problem
When receiving messages from Telegram forum topics, the inbound metadata currently includes:
- (the numeric thread ID)
- But NOT the topic name/title
This makes it impossible for agents to automatically identify which topic file to load, since they only receive the numeric ID.
## Current behavior
In a T... | closed | completed | true | 2 | [] | [] | 2026-03-05T14:30:48Z | 2026-04-16T04:24:26Z | 2026-04-13T18:08:16Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | dpraid | 71,937,886 | MDQ6VXNlcjcxOTM3ODg2 | User | false |
openclaw/openclaw | 4,002,661,668 | I_kwDOQb6kR87uk8Uk | 29,106 | https://github.com/openclaw/openclaw/issues/29106 | https://api.github.com/repos/openclaw/openclaw/issues/29106 | Gateway delivery-recovery replays already-delivered messages after restart | ## Summary
After restarting the gateway, the delivery-recovery system re-sends messages that were already successfully delivered in previous sessions. Recipients see duplicate messages that were originally sent days or weeks earlier.
## Steps to Reproduce
1. Have agents running normally with delivery history
2. Rest... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-02-27T20:04:41Z | 2026-04-19T04:24:37Z | 2026-04-16T04:24:25Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | petersykim | 35,867,020 | MDQ6VXNlcjM1ODY3MDIw | User | false |
openclaw/openclaw | 4,112,176,137 | I_kwDOQb6kR871GtQJ | 51,571 | https://github.com/openclaw/openclaw/issues/51571 | https://api.github.com/repos/openclaw/openclaw/issues/51571 | OpenRouter JSON 404 payloads do not trigger model fallback | ## Problem
OpenRouter sometimes returns HTTP 404 with a JSON body like:
```json
{"error":{"message":"Healer Alpha was a stealth model revealed on March 18th as an early testing version of MiMo-V2-Omni. Find it here: https://openrouter.ai/xiaomi/mimo-v2-omni","code":404},"user_id":"user_33GTyP8uDSYYbaeBO48AGHXyuMC"}
`... | closed | completed | true | 1 | [] | [] | 2026-03-21T11:27:15Z | 2026-04-16T04:24:23Z | 2026-04-13T18:53:58Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | mrdear | 13,889,883 | MDQ6VXNlcjEzODg5ODgz | User | false |
openclaw/openclaw | 4,002,557,512 | I_kwDOQb6kR87uki5I | 29,087 | https://github.com/openclaw/openclaw/issues/29087 | https://api.github.com/repos/openclaw/openclaw/issues/29087 | Feature: Idle-aware auto-compaction (compact during agent downtime) | ## Summary
When a session is idle (no user messages for a configurable duration) and context usage exceeds a threshold, automatically trigger compaction. This reclaims context space during downtime so the next conversation starts with maximum available context.
## Motivation
In long-running agent sessions (e.g., per... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-02-27T19:35:09Z | 2026-04-19T04:24:38Z | 2026-04-16T04:24:23Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | ArtDimensions | 212,239,527 | U_kgDODKaEpw | User | false |
openclaw/openclaw | 4,001,486,196 | I_kwDOQb6kR87ugdV0 | 28,845 | https://github.com/openclaw/openclaw/issues/28845 | https://api.github.com/repos/openclaw/openclaw/issues/28845 | RFC: Agent-controlled compaction via tool call | # RFC: Agent-controlled compaction via tool call
## Problem
Current compaction is a system-controlled black box that fires automatically when context pressure builds. This causes several compounding issues:
### 1. Cache invalidation cascade
A compaction event triggers up to 3 separate LLM calls, each invalidating pr... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-27T15:10:05Z | 2026-04-19T04:24:38Z | 2026-04-16T04:24:21Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | kami-saia | 204,396,543 | U_kgDODC7X_w | User | false |
openclaw/openclaw | 4,164,413,489 | I_kwDOQb6kR874N-gx | 57,072 | https://github.com/openclaw/openclaw/issues/57072 | https://api.github.com/repos/openclaw/openclaw/issues/57072 | [Bug]: Chat UI State Desync: Rendered messages do not match active session after route navigation | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After navigating away from Chat and returning (via sidebar/top nav), the active session and rendered conversation can diverge.
### Steps to reproduce
1. Open Chat and select a non-default session (for example old).
2. Send... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-29T13:56:19Z | 2026-04-16T04:24:19Z | 2026-04-13T20:24:59Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | bbddbb1 | 75,060,417 | MDQ6VXNlcjc1MDYwNDE3 | User | false |
openclaw/openclaw | 4,001,446,256 | I_kwDOQb6kR87ugTlw | 28,834 | https://github.com/openclaw/openclaw/issues/28834 | https://api.github.com/repos/openclaw/openclaw/issues/28834 | Slack message_changed events re-trigger AI response causing duplicate messages | ## Bug
When a user edits any Slack message in a DM, OpenClaw fires a `message_changed` event which re-triggers the AI session. The agent responds again to the previous user message, producing an identical duplicate reply.
## Steps to reproduce
1. Send a message in a Slack DM with the bot
2. Receive a response
3. Edit... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-27T15:01:46Z | 2026-04-19T04:24:39Z | 2026-04-16T04:24:19Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | jfulmines-star | 262,581,161 | U_kgDOD6arqQ | User | false |
openclaw/openclaw | 4,001,073,229 | I_kwDOQb6kR87ue4hN | 28,742 | https://github.com/openclaw/openclaw/issues/28742 | https://api.github.com/repos/openclaw/openclaw/issues/28742 | fix(slack): message tool does not auto-thread in DM conversations | ## Bug Description
When an agent uses the `message` tool from a DM thread session, the reply goes top-level instead of staying in the thread. Regular auto-replies (without the message tool) work correctly.
## Steps to Reproduce
1. Send a DM to the bot in Slack
2. Bot replies in a thread (correct, via auto-reply)
3. ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-02-27T13:35:45Z | 2026-04-19T04:24:40Z | 2026-04-16T04:24:17Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | Victorrrior | 208,812,903 | U_kgDODHI7Zw | User | false |
openclaw/openclaw | 4,000,882,244 | I_kwDOQb6kR87ueJ5E | 28,693 | https://github.com/openclaw/openclaw/issues/28693 | https://api.github.com/repos/openclaw/openclaw/issues/28693 | bug: signoff event never fires for NO_REPLY — parseReplyDirectives strips token before emission | ## Bug Description
The watchdog sign-off detection mechanism (`stream: "signoff"` event) never fires when an agent responds with `NO_REPLY` (or `HEARTBEAT_OK`) after using a tool like the message tool. This causes the reply-chain-enforcer to incorrectly ARM and trigger false recovery runs.
## Root Cause
Two strippin... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-27T12:48:20Z | 2026-04-19T04:24:40Z | 2026-04-16T04:24:15Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | kami-saia | 204,396,543 | U_kgDODC7X_w | User | false |
openclaw/openclaw | 4,015,967,138 | I_kwDOQb6kR87vXsui | 32,910 | https://github.com/openclaw/openclaw/issues/32910 | https://api.github.com/repos/openclaw/openclaw/issues/32910 | [Feature]: Inject system event on model fallback | # Feature Request: Inject system event on model fallback
## Summary
When the Gateway falls back from the primary model to a fallback (e.g., due to rate limiting or API errors), the agent has no awareness that a model switch occurred. The agent continues responding as if nothing changed, while the user may notice diff... | closed | completed | true | 0 | [
"enhancement"
] | [] | 2026-03-03T10:44:14Z | 2026-04-16T04:24:13Z | 2026-04-13T22:54:17Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | aaajiao | 49,885 | MDQ6VXNlcjQ5ODg1 | User | false |
openclaw/openclaw | 4,000,463,266 | I_kwDOQb6kR87ucjmi | 28,617 | https://github.com/openclaw/openclaw/issues/28617 | https://api.github.com/repos/openclaw/openclaw/issues/28617 | Feature: noOutputTimeoutMs for API provider responses (stuck streaming connections) | ## Problem
When using `openai-codex/gpt-5.2-codex` as a subagent model, ~10% of runs get stuck: the tool call completes and returns a result, but the subsequent API call to the provider never returns. No error, no timeout — the connection just hangs indefinitely.
## Current Behavior
- `timeoutSeconds` (agent-level) is... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-02-27T11:04:14Z | 2026-04-19T04:24:41Z | 2026-04-16T04:24:13Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | kinnoautomations-dot | 250,111,624 | U_kgDODuhmiA | User | false |
openclaw/openclaw | 3,999,711,543 | I_kwDOQb6kR87uZsE3 | 28,472 | https://github.com/openclaw/openclaw/issues/28472 | https://api.github.com/repos/openclaw/openclaw/issues/28472 | Canvas returns 401 Unauthorized when gateway runs behind Tailscale Serve | ## Summary
`isLocalDirectRequest()` fails for Canvas/A2UI endpoints when the gateway runs behind Tailscale Serve, even with `gateway.allowRealIpFallback: true` and `gateway.trustedProxies` correctly configured. This causes `{"error":{"message":"Unauthorized","type":"unauthorized"}}` for all Canvas and A2UI requests fr... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-02-27T07:56:35Z | 2026-04-19T04:24:42Z | 2026-04-16T04:24:11Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | SNRose07 | 133,899,507 | U_kgDOB_sk8w | User | false |
openclaw/openclaw | 3,999,531,176 | I_kwDOQb6kR87uZACo | 28,423 | https://github.com/openclaw/openclaw/issues/28423 | https://api.github.com/repos/openclaw/openclaw/issues/28423 | [Bug]: macOS update leaves LaunchAgent entrypoint stale (entry.js -> index.js) until manual doctor | ### Summary
After nightly update on macOS, openclaw doctor repeatedly reports:
Gateway service entrypoint does not match the current install (/opt/homebrew/lib/node_modules/openclaw/dist/entry.js -> /opt/homebrew/lib/node_modules/openclaw/dist/index.js)
This happens almost every morning and requires manual fix (doct... | closed | not_planned | true | 6 | [
"bug",
"stale"
] | [] | 2026-02-27T07:08:25Z | 2026-04-19T04:24:43Z | 2026-04-16T04:24:09Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | alexandergripas2019-prog | 261,962,708 | U_kgDOD5071A | User | false |
openclaw/openclaw | 3,999,130,680 | I_kwDOQb6kR87uXeQ4 | 28,327 | https://github.com/openclaw/openclaw/issues/28327 | https://api.github.com/repos/openclaw/openclaw/issues/28327 | Feature Request: Add Cursor OAuth provider for model access | ## Request
Add **Cursor** as a model provider using OAuth, similar to how Codex (openai-codex) works.
## Rationale
Currently OpenClaw supports:
- MiniMax Portal (OAuth)
- Codex via openai-codex (OAuth)
Many users have Cursor subscriptions and would love to use them as their AI model provider instead of paying for s... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-02-27T05:10:31Z | 2026-04-19T04:24:43Z | 2026-04-16T04:24:07Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | cbcampos | 63,550,718 | MDQ6VXNlcjYzNTUwNzE4 | User | false |
openclaw/openclaw | 4,098,920,894 | I_kwDOQb6kR870UJG- | 50,161 | https://github.com/openclaw/openclaw/issues/50161 | https://api.github.com/repos/openclaw/openclaw/issues/50161 | doctor fails on exec SecretRefs despite running gateway with resolved snapshot | ## Summary
`openclaw doctor` fails with `unresolved SecretRef` errors for channel credentials managed by an `exec` provider (credexec), even when the gateway is running and has successfully resolved all secrets. This blocks `openclaw update`, which gates on `doctor`.
## Reproduction
**Environment:** openclaw 2026.3.... | closed | completed | true | 3 | [] | [] | 2026-03-19T02:51:19Z | 2026-04-16T04:24:05Z | 2026-04-14T02:34:32Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | lolisaigao1234 | 92,764,750 | U_kgDOBYd6Tg | User | false |
openclaw/openclaw | 4,091,422,741 | I_kwDOQb6kR87z3igV | 49,264 | https://github.com/openclaw/openclaw/issues/49264 | https://api.github.com/repos/openclaw/openclaw/issues/49264 | doctor --non-interactive crashes with unresolved SecretRef when using exec: secret provider | ## Bug Report
**Version:** OpenClaw (latest as of 2026-03-17)
### Description
`openclaw doctor --non-interactive` crashes with an error when the config uses `exec:` SecretRef providers (e.g. `exec:skrytka:...` backed by macOS Keychain).
### Error
```
Error: channels.telegram.botToken: unresolved SecretRef "exec:sk... | closed | completed | true | 2 | [] | [] | 2026-03-17T22:00:36Z | 2026-04-16T04:24:05Z | 2026-04-14T02:33:23Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | toruvieI | 258,645,309 | U_kgDOD2qdPQ | User | false |
openclaw/openclaw | 3,999,130,351 | I_kwDOQb6kR87uXeLv | 28,326 | https://github.com/openclaw/openclaw/issues/28326 | https://api.github.com/repos/openclaw/openclaw/issues/28326 | Feature Request: Sandboxed agent profiles for communication channels | We need a better sandbox feature for communication channels that is not subject to prompt injection or jailbreaking.
**Use case:** Building a customer-facing bot that handles real-time conversations with end users via OpenClaw's messaging channels. The bot needs to be strictly scoped to one knowledge domain and must N... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-02-27T05:10:23Z | 2026-04-19T04:24:44Z | 2026-04-16T04:24:05Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,102,873,724 | I_kwDOQb6kR870jOJ8 | 50,537 | https://github.com/openclaw/openclaw/issues/50537 | https://api.github.com/repos/openclaw/openclaw/issues/50537 | doctor --fix crashes on env-source SecretRef for Discord token | ## Summary
`openclaw doctor --fix` crashes with an unresolved SecretRef error when `channels.discord.accounts.default.token` is configured as an env-source SecretRef. The gateway resolves the token correctly at runtime (Discord is fully functional), but doctor fails before completing checks.
This is the same class of... | closed | completed | true | 2 | [] | [] | 2026-03-19T16:36:32Z | 2026-04-16T04:24:04Z | 2026-04-14T02:36:41Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | protomota | 1,140,777 | MDQ6VXNlcjExNDA3Nzc= | User | false |
openclaw/openclaw | 3,998,538,069 | I_kwDOQb6kR87uVNlV | 28,106 | https://github.com/openclaw/openclaw/issues/28106 | https://api.github.com/repos/openclaw/openclaw/issues/28106 | RFC: Agent-to-Agent Task Delegation Protocol (Agent Economy) | ## Summary
A protocol enabling OpenClaw agents to discover, negotiate with, and delegate tasks to agents running on other OpenClaw instances. This creates a decentralized agent economy where specialized agents can offer services to other agents across the network.
## Motivation
OpenClaw has solved the "one human, on... | closed | not_planned | true | 9 | [
"stale"
] | [] | 2026-02-27T01:20:06Z | 2026-04-28T04:30:30Z | 2026-04-25T05:37:58Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dsantoreis | 220,753,637 | U_kgDODShu5Q | User | false |
openclaw/openclaw | 3,998,139,635 | I_kwDOQb6kR87uTsTz | 27,984 | https://github.com/openclaw/openclaw/issues/27984 | https://api.github.com/repos/openclaw/openclaw/issues/27984 | [Bug]: Telegram: files ~5-20MB cause silent deadlock (no error, no timeout, locks entire chat) | ### Summary
When a user sends a file between approximately 5MB and 20MB via Telegram DM, the gateway hangs silently — no error, no log entry, no timeout. The sequentialize(chatId) middleware locks the entire chat permanently. All subsequent messages (including text) are blocked. Only a gateway restart clears it. This ... | closed | not_planned | true | 8 | [
"bug",
"stale"
] | [] | 2026-02-26T22:48:02Z | 2026-04-19T04:24:45Z | 2026-04-16T04:23:59Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | rkcollins | 104,794,829 | U_kgDOBj8KzQ | User | false |
openclaw/openclaw | 4,204,198,345 | I_kwDOQb6kR876lvnJ | 60,758 | https://github.com/openclaw/openclaw/issues/60758 | https://api.github.com/repos/openclaw/openclaw/issues/60758 | feat(discord): resolve exec approvals via emoji reactions (👍/👎) | ## Summary
Exec approval cards in Discord require button clicks or Web UI interaction to resolve. Emoji reactions (👍 to approve, 👎 to deny) on the approval message should work as an alternative resolution method.
This is a continuation of #24639, which was closed for inactivity. That issue targeted BlueBubbles; thi... | closed | not_planned | true | 5 | [] | [] | 2026-04-04T08:22:36Z | 2026-04-16T04:23:57Z | 2026-04-14T04:13:19Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | cattyclaw-bot | 263,644,359 | U_kgDOD7bkxw | User | false |
openclaw/openclaw | 4,184,259,791 | I_kwDOQb6kR875ZrzP | 58,760 | https://github.com/openclaw/openclaw/issues/58760 | https://api.github.com/repos/openclaw/openclaw/issues/58760 | [Bug]: Discord fails to start on 2026.3.31 - TypeError: this.client.registerListener is not a function | ### Bug type
Regression (worked before on earlier versions)
### Summary
After upgrading to OpenClaw 2026.3.31, the Discord plugin fails to initialize with an unhandled promise rejection:
```
TypeError: this.client.registerListener is not a function
at SafeGatewayPlugin.registerClient (file:///.../openclaw/dist/pr... | closed | completed | true | 2 | [] | [
"thewilloftheshadow"
] | 2026-04-01T05:34:59Z | 2026-04-16T04:23:56Z | 2026-04-14T04:13:29Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Leegenux | 34,061,155 | MDQ6VXNlcjM0MDYxMTU1 | User | false |
openclaw/openclaw | 4,043,567,000 | I_kwDOQb6kR87xA--Y | 40,669 | https://github.com/openclaw/openclaw/issues/40669 | https://api.github.com/repos/openclaw/openclaw/issues/40669 | Discord voice: missing @snazzah/davey dependency causes silent 4017 (DAVE required) rejection | # Discord voice: missing `@snazzah/davey` dependency causes 4017 (DAVE required) rejection
## Bug
Discord voice connections fail silently with a 15-second timeout when joining voice channels that require DAVE E2E encryption (now the default on most Discord voice channels).
## Root Cause
`@snazzah/davey` is not incl... | closed | completed | true | 1 | [] | [] | 2026-03-09T06:10:18Z | 2026-04-16T04:23:55Z | 2026-04-14T04:17:05Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | jaffer1979 | 4,798,221 | MDQ6VXNlcjQ3OTgyMjE= | User | false |
openclaw/openclaw | 4,032,877,659 | I_kwDOQb6kR87wYNRb | 37,449 | https://github.com/openclaw/openclaw/issues/37449 | https://api.github.com/repos/openclaw/openclaw/issues/37449 | npm audit: 8 high severity vulnerabilities in 2026.3.1 deps | ## Issue\n\n`npm audit` reports 8 high severity vulnerabilities in `openclaw@2026.3.1`.\n\n## Root Cause\n\nVulnerable transitive dependencies:\n- `@hono/node-server <1.19.10` — authorization bypass (GHSA-wc8c-qw6v-h7f6)\n- `tar <=7.5.9` — multiple path traversal / arbitrary file write vulns\n- `@discordjs/opus`, `@dis... | closed | not_planned | true | 0 | [] | [] | 2026-03-06T08:12:35Z | 2026-04-16T04:23:55Z | 2026-04-14T04:17:00Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | mkdelta221 | 200,819,132 | U_kgDOC_hBvA | User | false |
openclaw/openclaw | 3,945,749,687 | I_kwDOQb6kR87rL1y3 | 17,731 | https://github.com/openclaw/openclaw/issues/17731 | https://api.github.com/repos/openclaw/openclaw/issues/17731 | [Bug]: Subagent announce doesn't preserve Slack assistant app thread context | ### Summary
When spawning a subagent via sessions_spawn from within a Slack assistant app thread (topic session), the announce routes to the base DM session instead of the originating thread. The announce response either gets lost or delivers as a flat DM, never reaching the thread where the user is actively chatting.... | closed | not_planned | true | 11 | [
"bug",
"stale"
] | [] | 2026-02-16T04:01:06Z | 2026-04-16T04:23:54Z | 2026-04-14T04:20:05Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Taskle | 1,445,636 | MDQ6VXNlcjE0NDU2MzY= | User | false |
openclaw/openclaw | 3,991,831,007 | I_kwDOQb6kR87t7oHf | 26,871 | https://github.com/openclaw/openclaw/issues/26871 | https://api.github.com/repos/openclaw/openclaw/issues/26871 | [Build] Why does building download rolldown instead of adding to depdencies? | https://github.com/openclaw/openclaw/blob/a898acbd55c1ef267f1f4cf858d03ee76129d0f1/scripts/bundle-a2ui.sh#L89-L93
Why is is that this can download rolldown instead of specifying in dependencies? Is it to avoid possibly building rolldown on weaker computers?
This is currently causing issues in newer versions when buil... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-02-25T21:07:52Z | 2026-04-19T04:24:46Z | 2026-04-16T04:23:53Z | NONE | null | 20260419T052035Z | 2026-04-19T05:20:35Z | chrisportela | 505,649 | MDQ6VXNlcjUwNTY0OQ== | User | false |
openclaw/openclaw | 3,952,608,386 | I_kwDOQb6kR87rmASC | 19,148 | https://github.com/openclaw/openclaw/issues/19148 | https://api.github.com/repos/openclaw/openclaw/issues/19148 | Feature request: Customizable compaction prompts and post-compaction instructions | ## Summary
Allow users to customize compaction behavior with three configuration options:
1. **Custom summarization prompt** — instructions for the model performing the summary
2. **Pre-compaction system message with optional agent turn** — notify the agent before compaction and optionally allow a final turn (e.g., t... | closed | not_planned | true | 12 | [
"stale"
] | [] | 2026-02-17T12:46:49Z | 2026-04-16T04:23:52Z | 2026-04-14T04:20:10Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | AshColton | 261,042,354 | U_kgDOD48wsg | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.