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 | 4,198,945,704 | I_kwDOQb6kR876RtOo | 60,131 | https://github.com/openclaw/openclaw/issues/60131 | https://api.github.com/repos/openclaw/openclaw/issues/60131 | [Feature]: Make TTS directive parsing honor explicit provider selection | ### Summary
Honor `provider=` in `[[tts:...]]` directives so only the selected speech provider consumes overlapping directive keys, and unsupported keys produce warnings instead of being silently dropped.
### Problem to solve
Today, TTS directive parsing is provider-order driven, not provider-aware. In `src/tts/dire... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-04-03T07:30:36Z | 2026-04-28T04:39:45Z | 2026-04-26T03:15:06Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sliekens | 1,583,241 | MDQ6VXNlcjE1ODMyNDE= | User | false |
openclaw/openclaw | 3,933,358,217 | I_kwDOQb6kR87qckiJ | 14,861 | https://github.com/openclaw/openclaw/issues/14861 | https://api.github.com/repos/openclaw/openclaw/issues/14861 | Gateway boot: no observability for subsystem startup failures (gmail-watcher, hooks) | ## Problem
When the OpenClaw gateway starts (or restarts), there is no logging or feedback about which subsystems/hooks successfully initialized and which silently failed. This makes it very difficult to diagnose post-restart issues in production.
### Specific example: gmail-watcher
The gmail-watcher hook requires:
... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-12T18:32:04Z | 2026-05-05T02:00:54Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | slucasmyer | 50,333,570 | MDQ6VXNlcjUwMzMzNTcw | User | false |
openclaw/openclaw | 3,933,392,794 | I_kwDOQb6kR87qcs-a | 14,874 | https://github.com/openclaw/openclaw/issues/14874 | https://api.github.com/repos/openclaw/openclaw/issues/14874 | [Feature]: Proposal: optional SIE verification in skill loader (warn/strict modes) | I propose an optional, feature-flagged SIE verification gate in the skill loader.
Behavior:
- unsigned + warn => allow + warning
- unsigned + strict => reject
- signed + valid => allow
- signed + invalid => reject
Goals:
- reduce trust in unsigned/tampered skill instructions
- preserve backward compatibility when fea... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-12T18:40:03Z | 2026-05-05T02:55:32Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | palxislabs | 258,656,192 | U_kgDOD2rHwA | User | false |
openclaw/openclaw | 3,933,931,317 | I_kwDOQb6kR87qewc1 | 14,968 | https://github.com/openclaw/openclaw/issues/14968 | https://api.github.com/repos/openclaw/openclaw/issues/14968 | [Feature]: Rename Heartbeat.md to Proactive.md | ## Summary
Keep things straight forward
## Proposed solution
It makes things much easier for the newcomers.
## Alternatives considered
Keep it the current way.
## Additional context
-
| open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-12T20:41:18Z | 2026-05-05T02:00:55Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | a-bormark | 9,692,981 | MDQ6VXNlcjk2OTI5ODE= | User | false |
openclaw/openclaw | 3,934,223,408 | I_kwDOQb6kR87qf3ww | 15,022 | https://github.com/openclaw/openclaw/issues/15022 | https://api.github.com/repos/openclaw/openclaw/issues/15022 | Feature request: Coalesce interleaved text blocks into single outbound message | ## Problem
When the model produces interleaved text and tool_use content blocks in a single turn:
```
[text: "Let me check..."] [tool_use: read_file] [text: "Now fixing..."] [tool_use: edit_file] [text: "Done!"]
```
Each text block is sent as a **separate outbound message** to Telegram (and presumably other channels... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-12T21:56:01Z | 2026-05-05T03:09:59Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | dp-vibes | 253,849,379 | U_kgDODyFvIw | User | false |
openclaw/openclaw | 3,934,281,971 | I_kwDOQb6kR87qgGDz | 15,032 | https://github.com/openclaw/openclaw/issues/15032 | https://api.github.com/repos/openclaw/openclaw/issues/15032 | Feature: Per-spawn tool restrictions for sub-agents | ## Problem
When spawning sub-agents via `sessions_spawn`, the child inherits the parent's full tool access. There's no way to restrict which tools a spawned agent can use.
## Use Case: DMZ Web Search (Prompt Injection Defense)
I'm building a three-zone isolated web search pipeline to prevent prompt injection from we... | open | null | false | 4 | [
"enhancement"
] | [] | 2026-02-12T22:13:22Z | 2026-05-05T02:00:56Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | smist37 | 66,137,970 | MDQ6VXNlcjY2MTM3OTcw | User | false |
openclaw/openclaw | 3,934,648,352 | I_kwDOQb6kR87qhfgg | 15,073 | https://github.com/openclaw/openclaw/issues/15073 | https://api.github.com/repos/openclaw/openclaw/issues/15073 | Feature Request: Per-agent context/workspace on model fallback | ## Problem
When a primary model (e.g., Claude via GitHub Copilot) hits rate limits or timeouts, OpenClaw falls back to configured fallback models. However, the fallback model receives the **same system prompt and workspace context** that was built for the primary model.
This creates a problem when:
- Primary agent ha... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-13T00:06:17Z | 2026-05-05T02:13:12Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | lucca-alma | 258,741,130 | U_kgDOD2wTig | User | false |
openclaw/openclaw | 3,938,296,982 | I_kwDOQb6kR87qvaSW | 15,634 | https://github.com/openclaw/openclaw/issues/15634 | https://api.github.com/repos/openclaw/openclaw/issues/15634 | [Feature]: Mattermost — Parity improvements (formatting, reactions, rich interactions) | ## Feature Request
Mattermost channel support lags significantly behind Discord, Telegram, and WhatsApp in terms of interaction quality. Several related issues exist individually, but there's no umbrella tracking for overall Mattermost parity.
### Key gaps (some have individual issues)
- **Markdown formatting degrade... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-13T17:44:29Z | 2026-05-05T02:55:34Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | gamersalpha | 32,466,950 | MDQ6VXNlcjMyNDY2OTUw | User | false |
openclaw/openclaw | 3,939,697,595 | I_kwDOQb6kR87q0wO7 | 15,828 | https://github.com/openclaw/openclaw/issues/15828 | https://api.github.com/repos/openclaw/openclaw/issues/15828 | [Feature]: feat(session-memory): Add auto-trigger on idle, shutdown, and periodic checkpoints | The bundled session-memory hook is great but only triggers on /new command. This means memory is lost if:
• The session times out or rotates automatically
• The gateway restarts (update, crash, reboot)
• The user forgets to run /new
• Long conversations just keep going without a checkpoint
Proposed Triggers
| Trigger... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-13T23:13:51Z | 2026-05-05T03:10:01Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | sepivip | 72,915,556 | MDQ6VXNlcjcyOTE1NTU2 | User | false |
openclaw/openclaw | 3,939,746,289 | I_kwDOQb6kR87q08Hx | 15,840 | https://github.com/openclaw/openclaw/issues/15840 | https://api.github.com/repos/openclaw/openclaw/issues/15840 | [Feature]: Option to suppress `<media:audio>` placeholder text | ## Summary
Add a config option to suppress the automatic `<media:audio>` placeholder text when sending voice/audio messages via messaging channels (Signal, Telegram, etc.).
## Current Behavior
When sending audio/voice media through OpenClaw's message tool, the core automatically adds `<media:audio>` as fallback text:
... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-13T23:35:20Z | 2026-05-05T03:10:03Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | dannymcc | 179,630 | MDQ6VXNlcjE3OTYzMA== | User | false |
openclaw/openclaw | 3,940,563,186 | I_kwDOQb6kR87q4Djy | 16,035 | https://github.com/openclaw/openclaw/issues/16035 | https://api.github.com/repos/openclaw/openclaw/issues/16035 | [Feature]: Make OpenClaw agents to identify themselves as agent, clearly, in Github profile | ## Summary
Jaws dropped when I saw this: https://github.com/matplotlib/matplotlib/pull/31132#issuecomment-3882314886
I urge you to tune OpenClaw to disclose the use of/the behavior of agent in their github profile by default. It is really getting hard to ddistinguish whether the code/text is written by the human fles... | closed | not_planned | true | 5 | [
"enhancement"
] | [] | 2026-02-14T06:03:15Z | 2026-04-29T10:47:10Z | 2026-04-29T10:47:02Z | NONE | null | 20260429T120040Z | 2026-04-29T12:00:40Z | ghost | 10,137 | MDQ6VXNlcjEwMTM3 | User | false |
openclaw/openclaw | 3,940,861,180 | I_kwDOQb6kR87q5MT8 | 16,121 | https://github.com/openclaw/openclaw/issues/16121 | https://api.github.com/repos/openclaw/openclaw/issues/16121 | [Feature]: Support for Custom Emojis in Latest Telegram Bot API Update | ## Summary
Telegram recently introduced custom emojis in the latest Bot API update (see formatting options documentation). The current bot doesn’t support rendering or handling these custom emojis correctly. I need Clawdbot to recognize and properly handle custom emojis in messages, captions, and replies according to ... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-14T09:00:42Z | 2026-05-05T02:01:02Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | maaduukaar | 38,440,668 | MDQ6VXNlcjM4NDQwNjY4 | User | false |
openclaw/openclaw | 3,940,987,742 | I_kwDOQb6kR87q5rNe | 16,142 | https://github.com/openclaw/openclaw/issues/16142 | https://api.github.com/repos/openclaw/openclaw/issues/16142 | [Feature]: You are spaming ChatGPT like crazy | ## Summary
Describe the problem you are trying to solve or the opportunity you see.
You are using ChatGPT API as a string formatter causing high API usage and hits API throttle limits.
I added the missing logger for ChatGPT to capture all this traffic, interesting
[openai.zip](https://github.com/user-attachments/f... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-14T10:01:38Z | 2026-05-05T02:13:17Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | devzer01 | 4,328,569 | MDQ6VXNlcjQzMjg1Njk= | User | false |
openclaw/openclaw | 3,941,898,624 | I_kwDOQb6kR87q9JmA | 16,351 | https://github.com/openclaw/openclaw/issues/16351 | https://api.github.com/repos/openclaw/openclaw/issues/16351 | [Feature]: Improve `openclaw config` to avoid secrets and invalid files | ### Summary
I would like to propose changes to the `openclaw config` CLI
### Problem to solve
- an explaratory `openclaw config` immediately going into dangerous interactive onboard-mode
- agent not able to simply validate his own config
- agent constantly including resolved ENV-placeholders in the config instead o... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-14T17:22:47Z | 2026-05-05T03:10:04Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | hotzen | 159,837 | MDQ6VXNlcjE1OTgzNw== | User | false |
openclaw/openclaw | 3,942,008,761 | I_kwDOQb6kR87q9ke5 | 16,387 | https://github.com/openclaw/openclaw/issues/16387 | https://api.github.com/repos/openclaw/openclaw/issues/16387 | Feature: Dynamic header templating for custom LLM providers (enables Fireworks.ai prompt caching) | ## Feature Request: Dynamic Header Templating for Provider Config
### Problem
When using **Fireworks.ai** as an LLM provider, prompt caching requires an `x-session-affinity` header to pin requests to the same replica. Without this header, requests are load-balanced across replicas with isolated caches, resulting in *... | open | null | false | 4 | [] | [
"Takhoffman"
] | 2026-02-14T18:17:19Z | 2026-05-05T02:01:04Z | null | CONTRIBUTOR | null | 20260505T030055Z | 2026-05-05T03:00:55Z | cash-echo-bot | 252,747,386 | U_kgDODxCeeg | User | false |
openclaw/openclaw | 3,942,371,028 | I_kwDOQb6kR87q-87U | 16,555 | https://github.com/openclaw/openclaw/issues/16555 | https://api.github.com/repos/openclaw/openclaw/issues/16555 | [Feature]: Add TTL/Expiry for Delivery Queue Messages | ### Summary
Add configurable TTL for delivery queue messages to prevent stale/orphaned entries from flooding channels on gateway restart
### Problem to solve
The delivery queue (introduced in v2026.2.13) persists outbound messages to disk indefinitely. When the gateway restarts, `recoverPendingDeliveries()` attempts... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-14T21:21:44Z | 2026-05-05T02:13:19Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | rohan-ixlayer | 106,705,719 | U_kgDOBlwzNw | User | false |
openclaw/openclaw | 3,942,684,812 | I_kwDOQb6kR87rAJiM | 16,670 | https://github.com/openclaw/openclaw/issues/16670 | https://api.github.com/repos/openclaw/openclaw/issues/16670 | Onboarding Wizard should include Memory/Embedding setup as a mandatory step | ## Problem
Memory is one of OpenClaw's most important features — it's what makes a bot persistent across sessions. Yet the onboarding wizard (`openclaw setup`) doesn't mention embedding provider configuration at all.
Without configuring `memorySearch` (embedding provider + API key), `memory_search` silently fails and... | open | null | false | 7 | [] | [
"Takhoffman"
] | 2026-02-15T00:33:45Z | 2026-05-05T02:01:07Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | actwebmaster | 47,089,798 | MDQ6VXNlcjQ3MDg5Nzk4 | User | false |
openclaw/openclaw | 3,942,830,315 | I_kwDOQb6kR87rAtDr | 16,711 | https://github.com/openclaw/openclaw/issues/16711 | https://api.github.com/repos/openclaw/openclaw/issues/16711 | [Feature]: Chat composer “Expand” button/modal (for long prompts) | ### Summary
Chat composer could use an “Expand” button/modal for long prompts.
### Problem to solve
When prompts get long, the current composer feels cramped.
### Proposed solution
Could we add a small Expand button/icon next to the chat input actions (near Send/New session) that opens a larger modal editor?
###... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T01:47:22Z | 2026-05-05T02:01:08Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | dragidude | 38,094,665 | MDQ6VXNlcjM4MDk0NjY1 | User | false |
openclaw/openclaw | 3,943,307,556 | I_kwDOQb6kR87rChkk | 16,896 | https://github.com/openclaw/openclaw/issues/16896 | https://api.github.com/repos/openclaw/openclaw/issues/16896 | [Feature] Add right-click reply in Dashboard webchat | ### Summary
Add right-click context menu to reply to specific messages in Dashboard webchat
### Problem to solve
Users cannot reply to specific messages in the Dashboard webchat. Currently, to respond to a message, users must manually copy the message content and create a new message referencing it. This is inconven... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T06:51:31Z | 2026-05-05T02:55:35Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | magicyyyy | 32,222,639 | MDQ6VXNlcjMyMjIyNjM5 | User | false |
openclaw/openclaw | 3,944,200,619 | I_kwDOQb6kR87rF7mr | 17,211 | https://github.com/openclaw/openclaw/issues/17211 | https://api.github.com/repos/openclaw/openclaw/issues/17211 | [Feature]: Add native checkpoint engine with deterministic resume and diff | ### Summary
Standardize task-boundary checkpoints and deterministic replay to reduce context drift after compaction.
### Problem to solve
Checkpointing is ad-hoc and replay is not deterministic. Users cannot reliably resume complex work with confidence that critical task state was preserved.
### Proposed solution... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T14:32:35Z | 2026-05-05T02:01:11Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Ar3ss12 | 150,044,748 | U_kgDOCPGATA | User | false |
openclaw/openclaw | 3,944,202,710 | I_kwDOQb6kR87rF8HW | 17,213 | https://github.com/openclaw/openclaw/issues/17213 | https://api.github.com/repos/openclaw/openclaw/issues/17213 | [Reliability]: Enforce atomic intent→execution flow to prevent orphan pre-action notices | ### Summary
Guarantee that announced actions always end in explicit commit/abort status.
### Problem to solve
Pre-action notices can be emitted without guaranteed execution under interruptions. This creates false progress signals and operational confusion.
### Proposed solution
Transaction-style run lifecycle:
-... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T14:33:47Z | 2026-05-05T02:01:12Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Ar3ss12 | 150,044,748 | U_kgDOCPGATA | User | false |
openclaw/openclaw | 3,944,203,806 | I_kwDOQb6kR87rF8Ye | 17,215 | https://github.com/openclaw/openclaw/issues/17215 | https://api.github.com/repos/openclaw/openclaw/issues/17215 | [Tooling]: Add pre/post-update compatibility scanner with risk report and smoke recommendations | ### Summary
Provide a native upgrade scanner to detect schema/behavior drift before and after updates.
### Problem to solve
Updates can silently break configs or runtime behavior. Issues are often discovered only after rollout, increasing downtime and rollback cost.
### Proposed solution
Add built-in `doctor --u... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T14:34:24Z | 2026-05-05T02:01:14Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Ar3ss12 | 150,044,748 | U_kgDOCPGATA | User | false |
openclaw/openclaw | 3,944,207,986 | I_kwDOQb6kR87rF9Zy | 17,217 | https://github.com/openclaw/openclaw/issues/17217 | https://api.github.com/repos/openclaw/openclaw/issues/17217 | [Feature]: | ### Summary
## FEATURE REQUEST: Define Session Boundaries Explicitly
### Current Problem
"Session" is undefined:
- Is it from connection start to end?
- Does context pruning happen mid-session?
- What triggers a new session?
- How do users control session boundaries?
### Requested Behavior
1. **Session definition:**... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-15T14:36:45Z | 2026-05-05T02:13:21Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | BispensGipsGebis | 2,768,883 | MDQ6VXNlcjI3Njg4ODM= | User | false |
openclaw/openclaw | 3,944,692,539 | I_kwDOQb6kR87rHzs7 | 17,424 | https://github.com/openclaw/openclaw/issues/17424 | https://api.github.com/repos/openclaw/openclaw/issues/17424 | [Feature]: Configurable Embedding Dimensions for Memory Search | ### Summary
**Is your feature request related to a problem? Please describe.**
Currently, OpenClaw uses the default dimension size for embedding models. For `text-embedding-3-large`, this is 3072 dimensions, which is unnecessarily large for most use cases and increases
### Problem to solve
- API costs (more tokens... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-15T18:56:03Z | 2026-05-05T02:13:22Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | clawdykorting-stack | 261,796,896 | U_kgDOD5q0IA | User | false |
openclaw/openclaw | 3,945,605,987 | I_kwDOQb6kR87rLStj | 17,683 | https://github.com/openclaw/openclaw/issues/17683 | https://api.github.com/repos/openclaw/openclaw/issues/17683 | Feature Request: Scoped / Script-Limited Agent Mode (Cost-Controlled + Predictable Execution) | ### Summary
I’d like to propose a “Scoped Agent Mode” that limits autonomy and focuses on predictable, cost-controlled execution patterns (cron + vetted scripts + LLM augmentation).
After running OpenClaw on a dedicated VPS for ~48 hours, I found the persistent agent pattern powerful — but difficult to control from b... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-16T02:48:29Z | 2026-05-05T03:10:06Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | stevewoz1234567890 | 87,572,213 | MDQ6VXNlcjg3NTcyMjEz | User | false |
openclaw/openclaw | 3,945,900,988 | I_kwDOQb6kR87rMau8 | 17,761 | https://github.com/openclaw/openclaw/issues/17761 | https://api.github.com/repos/openclaw/openclaw/issues/17761 | [Bug]: Gateway auth dispatcher blocks all internal services when mode=trusted-proxy (no shared-secret fallback) | ### Summary
`authorizeGatewayConnect` in `src/gateway/auth.ts` treats `trusted-proxy` as a single-mode gate. When proxy auth fails (i.e. the request didn't come from a trusted proxy), the function early-returns with a failure before reaching the shared-secret (token/password) or Tailscale authentication paths. This me... | closed | completed | true | 9 | [] | [
"vincentkoc"
] | 2026-02-16T05:03:51Z | 2026-04-30T04:32:57Z | 2026-04-27T22:10:24Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dashed | 139,499 | MDQ6VXNlcjEzOTQ5OQ== | User | false |
openclaw/openclaw | 3,946,106,293 | I_kwDOQb6kR87rNM21 | 17,810 | https://github.com/openclaw/openclaw/issues/17810 | https://api.github.com/repos/openclaw/openclaw/issues/17810 | [Feature]: Stop Burying Your Agent's Soul: Configurable Context File Elevation | ### Summary
Every OpenClaw user with a` SOUL.md` is fighting a losing battle. Your carefully crafted persona, identity, and memory files are being flattened into a generic pile of "project context" — ranked equal to `README.md` and random source files. This feature gives architectural weight to the files that actually... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-16T06:17:10Z | 2026-05-05T03:10:07Z | null | CONTRIBUTOR | null | 20260505T060034Z | 2026-05-05T06:00:34Z | sar618 | 214,745,104 | U_kgDODMzAEA | User | false |
openclaw/openclaw | 3,946,228,004 | I_kwDOQb6kR87rNqkk | 17,840 | https://github.com/openclaw/openclaw/issues/17840 | https://api.github.com/repos/openclaw/openclaw/issues/17840 | [Feature]: opt-in reaction-triggered agent turns | ### Summary
Reaction events are currently passive system events that don't trigger agent turns. Request an opt-in mechanism so reactions can wake the agent for interactive patterns.
### Problem to solve
**Use cases:**
- Emoji-choice polling (agent pre-reacts with options, user picks one)
- Lazy reactions as resp... | open | null | false | 5 | [
"enhancement"
] | [] | 2026-02-16T06:57:58Z | 2026-05-05T02:13:24Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Jackman3005 | 1,906,988 | MDQ6VXNlcjE5MDY5ODg= | User | false |
openclaw/openclaw | 3,946,432,089 | I_kwDOQb6kR87rOcZZ | 17,876 | https://github.com/openclaw/openclaw/issues/17876 | https://api.github.com/repos/openclaw/openclaw/issues/17876 | [Feature]: | ### Summary
Model switching UX: gradual rollout instead of immediate swap
### Problem to solve
Currently, when users request a model change (e.g., "switch from Opus to Haiku"), the change is applied immediately. If the new primary model has compatibility issues, bugs, or unexpected behavior, the user is left in a br... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-16T08:04:14Z | 2026-05-05T02:01:19Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | watalsalleskan | 152,772,147 | U_kgDOCRseMw | User | false |
openclaw/openclaw | 3,946,756,193 | I_kwDOQb6kR87rPrhh | 17,925 | https://github.com/openclaw/openclaw/issues/17925 | https://api.github.com/repos/openclaw/openclaw/issues/17925 | [Feature]: Support native web_search passthrough for ZAI (GLM) and Google (Gemini) providers | ### Summary
### Feature Description
OpenClaw already supports native `web_search` tool passthrough for xAI Grok via the Responses API (see `runGrokSearch` in the source code, which sends `tools: [{ type: "web_search" }]`).
**Request:** Please add the same native web search passthrough support for **ZAI (Zhipu GLM)*... | open | null | false | 5 | [
"enhancement"
] | [] | 2026-02-16T09:24:50Z | 2026-05-05T02:13:25Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | cedar202510-dotcom | 236,872,706 | U_kgDODh5kAg | User | false |
openclaw/openclaw | 3,946,810,783 | I_kwDOQb6kR87rP42f | 17,931 | https://github.com/openclaw/openclaw/issues/17931 | https://api.github.com/repos/openclaw/openclaw/issues/17931 | [Feature]: Mount skill directories read-only in sandbox containers | ### Summary
Skill directories are copied into the writable sandbox workspace, allowing agents to modify skill files; they should be bind-mounted read-only instead.
### Problem to solve
Skills are instructions, they never themselves should need to be writable. It is impossible to enforce any runtime security if the v... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-16T09:37:07Z | 2026-05-05T02:01:21Z | null | CONTRIBUTOR | null | 20260505T030055Z | 2026-05-05T03:00:55Z | tasaankaeris | 260,025,173 | U_kgDOD3-rVQ | User | false |
openclaw/openclaw | 3,948,289,971 | I_kwDOQb6kR87rVh-z | 18,160 | https://github.com/openclaw/openclaw/issues/18160 | https://api.github.com/repos/openclaw/openclaw/issues/18160 | [Feature]: Direct Exec Mode for Cron Jobs | ### Summary
Current cron jobs require `agentTurn` execution, which can timeout and jobs don't run.
### Problem to solve
## Problem
Current cron jobs require `agentTurn` execution, which:
- Forces LLM interpretation of simple commands
- Adds unnecessary API calls and latency
- Creates reliability issues (LLM timeout... | open | null | false | 10 | [
"enhancement"
] | [] | 2026-02-16T15:22:15Z | 2026-05-05T02:13:27Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | robertvanstedum | 65,718,279 | MDQ6VXNlcjY1NzE4Mjc5 | User | false |
openclaw/openclaw | 3,949,480,235 | I_kwDOQb6kR87raEkr | 18,548 | https://github.com/openclaw/openclaw/issues/18548 | https://api.github.com/repos/openclaw/openclaw/issues/18548 | [Feature]: 3 architecture improvements — bootstrap relevance, incremental compaction, observation extraction | OpenClaw's bootstrap file injection uses a positional split (70% head / 20% tail / 10% marker) that cuts the middle of files regardless of content relevance. For large workspace files, the most relevant section may be in the truncated middle. Confirmed in source: BOOTSTRAP_HEAD_RATIO = 0.7, BOOTSTRAP_TAIL_RATIO = 0.2 i... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-16T20:37:25Z | 2026-05-05T02:01:24Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | nnbiech95-cpu | 234,278,239 | U_kgDODfbNXw | User | false |
openclaw/openclaw | 3,949,603,522 | I_kwDOQb6kR87rairC | 18,571 | https://github.com/openclaw/openclaw/issues/18571 | https://api.github.com/repos/openclaw/openclaw/issues/18571 | [Feature]: Expose sessionKey/runId to outbound message hook for deterministic verification | ### Summary
Add an outbound_payloads_sending hook that fires before a message is delivered. Needs to read the message text, access the session's tool call history, annotate or block.
### Problem to solve
Why: LLM's lie their pants off. I built two deterministic verification scripts (zero LLM) that catch it after th... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-16T21:15:58Z | 2026-05-05T02:01:25Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | glen-carlson | 261,998,086 | U_kgDOD53GBg | User | false |
openclaw/openclaw | 3,951,183,888 | I_kwDOQb6kR87rgkgQ | 18,967 | https://github.com/openclaw/openclaw/issues/18967 | https://api.github.com/repos/openclaw/openclaw/issues/18967 | [Feature]: Parent-scoped sessions_send for sub-agents | ### Summary
Allow sub-agents to send mid-run updates and questions to their direct parent session only.
### Problem to solve
Currently `sessions_send` is in `SUBAGENT_TOOL_DENY_ALWAYS` — intentionally blocked because "subagents communicate through announce chain". This means sub-agents have no way to send mid-run pr... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-17T06:59:57Z | 2026-05-05T02:13:30Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | miloishot | 89,070,480 | MDQ6VXNlcjg5MDcwNDgw | User | false |
openclaw/openclaw | 3,951,256,053 | I_kwDOQb6kR87rg2H1 | 18,985 | https://github.com/openclaw/openclaw/issues/18985 | https://api.github.com/repos/openclaw/openclaw/issues/18985 | [Feature]: Supports Windows 11 MSYS environment and Fishshell. | ### Summary
Windows 11 msys2 is a Unix-like environment that supports tools such as bash and fish.
```
curl -fsSL https://openclaw.ai/install.sh | bash
🦞 OpenClaw Installer
End-to-end encrypted, Zuck-to-Zuck excluded.
· gum skipped (unsupported os/arch (unsupported/x86_64))
✗ Unsupported operating system
This... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-17T07:22:30Z | 2026-05-05T02:13:33Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | ahaoboy | 19,884,146 | MDQ6VXNlcjE5ODg0MTQ2 | User | false |
openclaw/openclaw | 3,951,923,414 | I_kwDOQb6kR87rjZDW | 19,075 | https://github.com/openclaw/openclaw/issues/19075 | https://api.github.com/repos/openclaw/openclaw/issues/19075 | [Feature]: Add baseUrl option for Brave Search provider | ### Summary
Currently, the Brave Search provider in tools.web.search has no baseUrl option — the endpoint is hardcoded. The Perplexity provider already supports baseUrl, so there's precedent in the codebase.
### Problem to solve
Why this matters:
In self-hosted and enterprise deployments, it's common to route exter... | closed | completed | true | 6 | [
"enhancement"
] | [] | 2026-02-17T10:05:40Z | 2026-05-05T04:28:20Z | 2026-05-02T07:27:53Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | jkoprax | 14,879,141 | MDQ6VXNlcjE0ODc5MTQx | User | false |
openclaw/openclaw | 3,953,691,963 | I_kwDOQb6kR87rqI07 | 19,289 | https://github.com/openclaw/openclaw/issues/19289 | https://api.github.com/repos/openclaw/openclaw/issues/19289 | Browser backend plugin interface (support third-party drivers via MCP/custom adapter) | ### Summary
Pluggable browser backend interface for registering third-party drivers without core modifications.
### Problem to solve
OpenClaw currently supports two browser driver types (extension/relay and managed). Users cannot register alternative backends (e.g., MCP engines) without forks or brittle patches.
##... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-17T16:45:03Z | 2026-05-05T03:10:08Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | G7CNF | 226,691,162 | U_kgDODYMIWg | User | false |
openclaw/openclaw | 3,955,611,053 | I_kwDOQb6kR87rxdWt | 19,680 | https://github.com/openclaw/openclaw/issues/19680 | https://api.github.com/repos/openclaw/openclaw/issues/19680 | [Feature]: Proposal: Restructure VISION.md as a decision framework for humans and AI PR review | ### Summary
Restructure VISION.md from a static rejection list into a decision framework that both human contributors and an AI PR auditor can use to evaluate any PR against the product thesis.
### Problem to solve
VISION.md currently tells the reader what to reject with a static list. That works for humans who read... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-18T02:50:56Z | 2026-05-05T02:13:40Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Emuthmartinez | 27,366,738 | MDQ6VXNlcjI3MzY2NzM4 | User | false |
openclaw/openclaw | 3,956,434,020 | I_kwDOQb6kR87r0mRk | 19,859 | https://github.com/openclaw/openclaw/issues/19859 | https://api.github.com/repos/openclaw/openclaw/issues/19859 | [Feature Request] WebChat: Add option to control Enter key sending behavior | ### Summary
Add a setting to control whether Enter key sends the message immediately or inserts a newline (with Shift+Enter to send).
### Problem to solve
Currently in WebChat UI, pressing Enter immediately sends the message. Users who write longer messages often accidentally send incomplete messages when pressing E... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-02-18T07:43:02Z | 2026-05-02T04:27:17Z | 2026-04-29T20:14:32Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | nothingp | 8,275,285 | MDQ6VXNlcjgyNzUyODU= | User | false |
openclaw/openclaw | 3,960,310,122 | I_kwDOQb6kR87sDYlq | 20,460 | https://github.com/openclaw/openclaw/issues/20460 | https://api.github.com/repos/openclaw/openclaw/issues/20460 | [Feature]: Store configuration backup files in their own directory, with datetime name convention | ### Summary
Store the configuration backup files in their own directory, with datetime name convention (ex: `.openclaw/config-backup/openclaw.json.bak.2026-02-18T14-37-28.123Z`)
### Problem to solve
* Every time there is a configuration change/update, OpenClaw apparently creates a backup copy of the configuration be... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-02-18T23:59:29Z | 2026-05-05T02:01:38Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | pandurand | 3,216,556 | MDQ6VXNlcjMyMTY1NTY= | User | false |
openclaw/openclaw | 3,973,006,993 | I_kwDOQb6kR87sz0aR | 23,014 | https://github.com/openclaw/openclaw/issues/23014 | https://api.github.com/repos/openclaw/openclaw/issues/23014 | [Feature]: Add .m2a audio format support for Telegram and media handling | ### Summary
Support .m2a as an audio format so that voice/audio files sent via Telegram (and other channels) are recognized as audio, get the correct placeholder/context, like .mp3/.ogg.
### Problem to solve
When a user sends an .m2a audio file via Telegram, OpenClaw does not treat it as audio. The agent is unable t... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-21T22:55:46Z | 2026-05-05T02:55:45Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | mayonaissse | 185,061 | MDQ6VXNlcjE4NTA2MQ== | User | false |
openclaw/openclaw | 3,983,583,872 | I_kwDOQb6kR87tcKqA | 25,401 | https://github.com/openclaw/openclaw/issues/25401 | https://api.github.com/repos/openclaw/openclaw/issues/25401 | [Feature]: models.providers.*.enabled flag to disable providers without removing config | ### Summary
Allow disabling a model provider via `enabled: false` without deleting its config block.
### Problem to solve
When you need to temporarily disable a provider (cost spike, rate limits, debugging), the only option is deleting the entire provider block from openclaw.json. Tried adding `"enabled": false` but... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-24T12:55:25Z | 2026-05-05T02:55:57Z | null | CONTRIBUTOR | null | 20260505T030055Z | 2026-05-05T03:00:55Z | rrenamed | 87,486,610 | MDQ6VXNlcjg3NDg2NjEw | User | false |
openclaw/openclaw | 3,984,712,062 | I_kwDOQb6kR87tgeF- | 25,621 | https://github.com/openclaw/openclaw/issues/25621 | https://api.github.com/repos/openclaw/openclaw/issues/25621 | [Feature]: Error: Gateway service install not supported on openbsd | ### Summary
```
$ openclaw status
__ OpenClaw 2026.2.23 (b817600) _ It's not "failing," it's "discovering new ways to configure the same thing wrong."
│
_
│
_
[openclaw] Failed to start CLI: Error: Gateway service install not supported on openbsd
at resolveGatewayService (file:///usr/local/openclaw/node_modu... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-02-24T16:40:42Z | 2026-05-05T02:56:02Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | kucharskim | 4,342,029 | MDQ6VXNlcjQzNDIwMjk= | User | false |
openclaw/openclaw | 3,987,412,583 | I_kwDOQb6kR87tqxZn | 26,183 | https://github.com/openclaw/openclaw/issues/26183 | https://api.github.com/repos/openclaw/openclaw/issues/26183 | [Feature]: WhatsApp ackReaction removal after reply (parity with other channels) | ### Summary
OpenClaw supports removeAckAfterReply in messages config and remove: true as a reaction tool parameter (according to the docs), but neither works for WhatsApp's ackReaction. Once a message is replied with an emoji, it stays permanently — there is no config option to remove it after the agent replies. Other... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-02-25T06:02:37Z | 2026-04-30T04:43:06Z | 2026-04-26T04:39:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MrUnforsaken | 113,491,899 | U_kgDOBsO_uw | User | false |
openclaw/openclaw | 3,988,922,143 | I_kwDOQb6kR87twh8f | 26,428 | https://github.com/openclaw/openclaw/issues/26428 | https://api.github.com/repos/openclaw/openclaw/issues/26428 | [Feature]: install complitions openclaw.fish into directory ~/.config/fish/completions | ### Summary
instead of sourcing directly in ~/.config/fish/config.fish
### Problem to solve
.
### Proposed solution
.
### Alternatives considered
_No response_
### Impact
.
### Evidence/examples
_No response_
### Additional information
_No response_ | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-02-25T11:33:54Z | 2026-05-03T04:38:23Z | 2026-04-30T05:19:37Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | kuvaldini | 47,349,143 | MDQ6VXNlcjQ3MzQ5MTQz | User | false |
openclaw/openclaw | 4,014,592,585 | I_kwDOQb6kR87vSdJJ | 32,515 | https://github.com/openclaw/openclaw/issues/32515 | https://api.github.com/repos/openclaw/openclaw/issues/32515 | Feature Request: One-click Anthropic authentication (OAuth flow) | ### Summary
One-click Anthropic authentication (OAuth flow)
### Problem to solve
The current authentication setup requires users to understand the difference between API keys and OAuth tokens, manually obtain them, know where to paste them, and debug .env files and config. This is a DevOps workflow — not a user... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-03T04:11:07Z | 2026-05-05T02:56:47Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | andreygalizin | 36,531,202 | MDQ6VXNlcjM2NTMxMjAy | User | false |
openclaw/openclaw | 4,019,519,572 | I_kwDOQb6kR87vlQBU | 33,724 | https://github.com/openclaw/openclaw/issues/33724 | https://api.github.com/repos/openclaw/openclaw/issues/33724 | [Feature]: Allow user configuration of where *.json.bak files save. | ### Summary
Allow users to configure the path `.openclaw.json.bak` files are saved too, as well as the behavior of what triggers a backups, and how many latest versions are kept at once.
### Problem to solve
All of the following elements identified are hardcoded, and unable to be changed by the user:
1. All `.open... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-04T01:23:03Z | 2026-05-05T02:56:57Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | gray344 | 142,852,509 | U_kgDOCIPBnQ | User | false |
openclaw/openclaw | 4,020,388,168 | I_kwDOQb6kR87vokFI | 34,064 | https://github.com/openclaw/openclaw/issues/34064 | https://api.github.com/repos/openclaw/openclaw/issues/34064 | Add Feishu message history retrieval to agent tools | ### Summary
Allow agents to retrieve recent Feishu chat message history via a new `history` action in feishu_chat tool.
### Problem to solve
When an agent starts a new session on Feishu, it has no access to previous conversation history. Cross-session context continuity relies entirely on the agent manually writing ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-04T06:40:32Z | 2026-04-28T04:42:47Z | 2026-04-26T02:41:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | upmyth-328 | 264,552,966 | U_kgDOD8TCBg | User | false |
openclaw/openclaw | 4,022,033,650 | I_kwDOQb6kR87vu1zy | 34,474 | https://github.com/openclaw/openclaw/issues/34474 | https://api.github.com/repos/openclaw/openclaw/issues/34474 | [Feature]: Allow subscribe to repository releases inside Slack | ### Summary
I'd like to subscribe to the repository and watch for new releases, allowing Slack to notify me whenever a new release comes out.
### Problem to solve
In Slack, I use GitHub app to get notifications. But when I try to subscribe to openclaw, I get the following error:
```
Either the app isn't installed on... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-04T13:31:18Z | 2026-04-28T04:42:46Z | 2026-04-26T02:41:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mateusscheper | 43,916,794 | MDQ6VXNlcjQzOTE2Nzk0 | User | false |
openclaw/openclaw | 4,022,127,070 | I_kwDOQb6kR87vvMne | 34,500 | https://github.com/openclaw/openclaw/issues/34500 | https://api.github.com/repos/openclaw/openclaw/issues/34500 | Feature Request: Include Model Call Count in /usage Command | ### Summary
Add a counter for the number of LLM calls to the /usage command output to provide greater clarity on token consumption.
### Problem to solve
The current `/usage` command aggregates token counts from all LLM calls within a single turn, which creates ambiguity. A user cannot distinguish between one very la... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-03-04T13:52:13Z | 2026-04-28T04:42:45Z | 2026-04-26T02:41:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rpm2205 | 248,671,524 | U_kgDODtJtJA | User | false |
openclaw/openclaw | 4,026,116,679 | I_kwDOQb6kR87v-apH | 35,573 | https://github.com/openclaw/openclaw/issues/35573 | https://api.github.com/repos/openclaw/openclaw/issues/35573 | [Feature]: Native fal.ai provider support | ### Summary
Add native fal.ai as a first-class auth provider so users can connect OpenClaw to fal.ai's OpenRouter-compatible LLM gateway without manual config file editing.
### Problem to solve
fal.ai offers an OpenRouter-compatible API (https://fal.run/openrouter/router) with access to all major models (Claude, Gem... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-05T06:24:10Z | 2026-04-28T04:42:44Z | 2026-04-26T02:41:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | egeaydin1 | 67,596,263 | MDQ6VXNlcjY3NTk2MjYz | User | false |
openclaw/openclaw | 4,041,197,234 | I_kwDOQb6kR87w38ay | 39,821 | https://github.com/openclaw/openclaw/issues/39821 | https://api.github.com/repos/openclaw/openclaw/issues/39821 | [Feature]: Add Cmd+Enter (Mac) / Ctrl+Enter (Windows) shortcut to send messages in Control UI | ### Summary
Add keyboard shortcut (Cmd/Ctrl+Enter) to send messages in Control UI chat, with Enter for new line.
### Problem to solve
Currently, the Control UI chat input sends messages when pressing Enter alone. This makes it difficult to:
1. Type multi-line messages without accidentally sending them
2. Edit mess... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-03-08T12:25:55Z | 2026-04-28T04:42:38Z | 2026-04-26T02:42:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Liutingxun5548 | 49,804,792 | MDQ6VXNlcjQ5ODA0Nzky | User | false |
openclaw/openclaw | 4,041,523,914 | I_kwDOQb6kR87w5MLK | 40,011 | https://github.com/openclaw/openclaw/issues/40011 | https://api.github.com/repos/openclaw/openclaw/issues/40011 | [Feature]: Support OTLP environment variables specifically for traces, logs, and metrics, in addition to the global one | ### Summary
Currently, OpenClaw supports the OpenTelemetry (OTLP/HTTP) exporter with **a single URL**, configured via the `diagnostics.otel.endpoint` parameter or the environment variable `OTEL_EXPORTER_OTLP_ENDPOINT`.
### Problem to solve
In some cases, users may have different backends for different telemetry sig... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-08T15:35:39Z | 2026-04-28T04:42:37Z | 2026-04-26T02:42:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jiekun | 30,280,396 | MDQ6VXNlcjMwMjgwMzk2 | User | false |
openclaw/openclaw | 4,042,027,022 | I_kwDOQb6kR87w7HAO | 40,208 | https://github.com/openclaw/openclaw/issues/40208 | https://api.github.com/repos/openclaw/openclaw/issues/40208 | [Feature]: Per-agent plugin scoping (disable plugin background services for specific agents) | ### Summary
Requested feature
Core support for per-agent plugin scoping / lifecycle, e.g. something like:
per-agent allow/deny list for plugins, or
ability for plugins to register “agent-scoped services” that only run for selected agents.
### Problem to solve
I’m running multiple agents on one G... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-08T19:43:20Z | 2026-04-28T04:42:34Z | 2026-04-26T02:42:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ruizcrp | 12,788,884 | MDQ6VXNlcjEyNzg4ODg0 | User | false |
openclaw/openclaw | 4,045,231,282 | I_kwDOQb6kR87xHVSy | 40,955 | https://github.com/openclaw/openclaw/issues/40955 | https://api.github.com/repos/openclaw/openclaw/issues/40955 | [Feature]: Unable to reproduce the OpenClaw vulnerability | ### Summary
Many of the vulnerabilities in [security ](https://github.com/openclaw/openclaw/security)are non-reproducible. Do you happen to have any reproduction scripts
Specifically, I’ve looked into:
[1][GHSA-qcc4-p59m-p54m](https://github.com/openclaw/openclaw/security/advisories/GHSA-qcc4-p59m-p54m)
[2] [GHSA-c... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T12:10:56Z | 2026-04-28T04:42:31Z | 2026-04-26T02:44:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Subway2023 | 147,013,944 | U_kgDOCMNBOA | User | false |
openclaw/openclaw | 4,045,384,833 | I_kwDOQb6kR87xH6yB | 40,986 | https://github.com/openclaw/openclaw/issues/40986 | https://api.github.com/repos/openclaw/openclaw/issues/40986 | [Feature] Share memory between different sessions of the same agent | ### Summary
Allow sharing memory between different sessions of the same agent (e.g., DM and group chat)
### Problem to solve
Currently, the same agent (e.g., Luka) has independent sessions for DM (direct message) and group chat. Each session has its own memory/context, so when a user switches from DM to a group chat... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-09T12:43:22Z | 2026-04-28T04:42:31Z | 2026-04-26T02:44:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | alecgu | 30,724,675 | MDQ6VXNlcjMwNzI0Njc1 | User | false |
openclaw/openclaw | 4,045,518,117 | I_kwDOQb6kR87xIbUl | 41,024 | https://github.com/openclaw/openclaw/issues/41024 | https://api.github.com/repos/openclaw/openclaw/issues/41024 | feat: Native Computer Use UI Integration (GPT-5.4 Style) | # Feature Request: Native Computer Use UI Integration (GPT-5.4 Style)
## Problem Description
Currently, OpenClaw primarily controls computers through CLI-based tools (`exec`, `browser`, `canvas`). With the release of GPT-5.4's native Computer Use capability, users expect OpenClaw to support visual UI-based computer o... | closed | completed | true | 3 | [] | [] | 2026-03-09T13:09:46Z | 2026-04-28T04:42:30Z | 2026-04-26T02:44:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | lileilei-camera | 7,010,843 | MDQ6VXNlcjcwMTA4NDM= | User | false |
openclaw/openclaw | 4,046,113,377 | I_kwDOQb6kR87xKsph | 41,163 | https://github.com/openclaw/openclaw/issues/41163 | https://api.github.com/repos/openclaw/openclaw/issues/41163 | 飞书多机器人无法收发信息 | ### Bug type
Regression (worked before, now fails)
### Summary
我有2台计算机，一个飞书账号，A计算机部署openclaw并且配置了A机器人，这里我只填写了key和id，它正常工作，当我B计算机配置B机器人，也是只填写key和id时，它无法正常收发信息，是哪里出现了问题吗？
I have two computers and one Lark account. On computer A, I deployed OpenClaw and configured Robot A. I only filled in the key and id, and it worke... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-09T14:59:46Z | 2026-04-28T04:42:29Z | 2026-04-26T02:44:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | StephenJose-Dai | 36,736,149 | MDQ6VXNlcjM2NzM2MTQ5 | User | false |
openclaw/openclaw | 4,046,319,064 | I_kwDOQb6kR87xLe3Y | 41,194 | https://github.com/openclaw/openclaw/issues/41194 | https://api.github.com/repos/openclaw/openclaw/issues/41194 | update: gateway service unexpectedly started on non-gateway nodes after update | ## Summary
Running `openclaw update` on a **non-gateway node** (a node where the gateway service was never installed) unexpectedly starts the gateway service after the update completes.
## Steps to reproduce
1. On a machine that uses OpenClaw as a client only (no `openclaw gateway install` has been run).
2. Run `ope... | closed | completed | true | 2 | [] | [] | 2026-03-09T15:38:21Z | 2026-04-28T04:42:29Z | 2026-04-26T02:44:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zhengbingxian | 33,149,547 | MDQ6VXNlcjMzMTQ5NTQ3 | User | false |
openclaw/openclaw | 4,046,503,294 | I_kwDOQb6kR87xML1- | 41,226 | https://github.com/openclaw/openclaw/issues/41226 | https://api.github.com/repos/openclaw/openclaw/issues/41226 | [Bug]: Behavior bug (incorrect output/state without crash) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw fails to authenticate with OpenRouter and Venice AI despite valid API keys, returning HTTP 401/400 errors.
### Steps to reproduce
1. Set OPENROUTER_API_KEY in ~/.openclaw/.env
2. Run openclaw gateway
3. Send Telegram message to b... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-09T16:10:44Z | 2026-04-28T04:42:28Z | 2026-04-26T02:44:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Dara-Hatamti | 221,226,321 | U_kgDODS-lUQ | User | false |
openclaw/openclaw | 4,046,558,149 | I_kwDOQb6kR87xMZPF | 41,240 | https://github.com/openclaw/openclaw/issues/41240 | https://api.github.com/repos/openclaw/openclaw/issues/41240 | Bug: heartbeat intervalMs set to ~30 days (2,592,000,000ms) after 3.8 update — Node.js 32-bit overflow causes 500MB log spam | ## Summary
After updating to OpenClaw 3.8 and restarting, the heartbeat subsystem sets `intervalMs` to **2,592,000,000ms** (~30 days) regardless of the configured value. Node.js cannot fit this into a 32-bit signed integer, clamps the timeout to 1ms, and logs a `TimeoutOverflowWarning` error on every iteration — gener... | closed | completed | true | 2 | [] | [] | 2026-03-09T16:21:11Z | 2026-04-28T04:42:27Z | 2026-04-26T02:45:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mrsergiopascal | 258,002,037 | U_kgDOD2DMdQ | User | false |
openclaw/openclaw | 4,046,619,805 | I_kwDOQb6kR87xMoSd | 41,257 | https://github.com/openclaw/openclaw/issues/41257 | https://api.github.com/repos/openclaw/openclaw/issues/41257 | [Bug]: MiniMax models' tool calls are stripped instead of executed | ### Bug type
<img width="495" height="199" alt="Image" src="https://github.com/user-attachments/assets/34313e8a-1f60-453d-881a-1473b5c6d6c2" />
Regression (worked before, now fails)
### Summary
## Problem
MiniMax M2.5 models output tool calls in format:
```xml
<minimax:tool_call>
<invoke name="exec">
... | closed | not_planned | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-09T16:33:26Z | 2026-04-28T04:42:26Z | 2026-04-26T02:45:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KingsYR123 | 48,837,228 | MDQ6VXNlcjQ4ODM3MjI4 | User | false |
openclaw/openclaw | 4,046,665,411 | I_kwDOQb6kR87xMzbD | 41,270 | https://github.com/openclaw/openclaw/issues/41270 | https://api.github.com/repos/openclaw/openclaw/issues/41270 | Config mutation commands have inconsistent restart semantics (config.set vs config.patch) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
`config.patch` and `config.apply` both write config to disk **and** schedule an automatic gateway restart via `scheduleGatewaySigusr1Restart()`. However, `config.set` only writes config to disk and prints "Restart the gateway to apply" — it... | closed | completed | true | 2 | [] | [] | 2026-03-09T16:41:49Z | 2026-04-28T04:42:26Z | 2026-04-26T02:45:11Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Yaxuan42 | 184,813,557 | U_kgDOCwQH9Q | User | false |
openclaw/openclaw | 4,046,768,029 | I_kwDOQb6kR87xNMed | 41,285 | https://github.com/openclaw/openclaw/issues/41285 | https://api.github.com/repos/openclaw/openclaw/issues/41285 | [Bug]: error 4008 | ### Bug type
Regression (worked before, now fails)
### Summary
ystem say unauthorized: gateway token missing (open the dashboard URL and paste the token in Control UI settings)
此网关需要身份验证。添加令牌或密码，然后点击连接。
openclaw dashboard --no-open → tokenized URL
openclaw doctor --generate-gateway-token → set token
but i run opencl... | closed | completed | true | 9 | [
"bug",
"regression"
] | [] | 2026-03-09T17:00:44Z | 2026-04-28T04:42:25Z | 2026-04-26T02:45:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | panchangtian-hash | 266,802,806 | U_kgDOD-cWdg | User | false |
openclaw/openclaw | 4,046,966,043 | I_kwDOQb6kR87xN80b | 41,312 | https://github.com/openclaw/openclaw/issues/41312 | https://api.github.com/repos/openclaw/openclaw/issues/41312 | [Feature]: Add cacheRead guardrails for large retained tool results in long sessions | ## Summary
Add built-in guardrails to reduce cacheRead-dominant token waste in long tool-use sessions (especially when large tool results are repeatedly replayed in prompt context).
## Problem to solve
In a recent real-world session analysis, token usage was heavily dominated by replayed cached context instead of new ... | closed | completed | true | 1 | [] | [] | 2026-03-09T17:38:48Z | 2026-04-28T04:42:23Z | 2026-04-26T02:45:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Moshiii | 7,553,188 | MDQ6VXNlcjc1NTMxODg= | User | false |
openclaw/openclaw | 4,047,132,016 | I_kwDOQb6kR87xOlVw | 41,344 | https://github.com/openclaw/openclaw/issues/41344 | https://api.github.com/repos/openclaw/openclaw/issues/41344 | [Bug]: Control UI Agent form shows wrong Primary model and Save fails with GatewayRequestError: invalid config | ### Bug type
Regression (worked before, now fails)
### Summary
On OpenClaw 2026.3.8, the Control UI Agent page has two related problems:
Model Selection shows the wrong Primary model (default) value
Saving valid edits from the Agent page fails with:
GatewayRequestError: invalid config
This affects at least:
Model ... | closed | completed | true | 7 | [
"bug",
"regression"
] | [] | 2026-03-09T18:12:37Z | 2026-04-28T04:42:22Z | 2026-04-26T02:45:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Ezri-Lin | 100,393,071 | U_kgDOBfvgbw | User | false |
openclaw/openclaw | 4,047,192,427 | I_kwDOQb6kR87xO0Fr | 41,354 | https://github.com/openclaw/openclaw/issues/41354 | https://api.github.com/repos/openclaw/openclaw/issues/41354 | Discord plugin disconnects every ~10-15 minutes (health-monitor restart loop) | ## Description
The Discord provider disconnects every ~10-15 minutes, requiring the health-monitor to detect and restart it. This creates a recurring cycle where messages sent during the ~2 second reconnect window can be missed.
## Environment
- **OpenClaw version:** 2026.3.8 (3caab92)
- **Node:** v22.22.0
- **OS:**... | closed | completed | true | 4 | [] | [] | 2026-03-09T18:26:17Z | 2026-04-28T04:42:22Z | 2026-04-26T02:45:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | russellweiss | 44,983,860 | MDQ6VXNlcjQ0OTgzODYw | User | false |
openclaw/openclaw | 4,047,776,609 | I_kwDOQb6kR87xRCth | 41,417 | https://github.com/openclaw/openclaw/issues/41417 | https://api.github.com/repos/openclaw/openclaw/issues/41417 | feat(cli): add `openclaw agents view-system-prompt` command | ## Summary
Add a CLI command to print the fully assembled system prompt for any agent without making an LLM call.
```
openclaw agents view-system-prompt [agentId]
```
### Options
- `--model <id>` — model id to show in the Runtime line (provider/model format)
- `--channel <channel>` — channel to simulate for messagin... | closed | not_planned | true | 3 | [] | [] | 2026-03-09T20:35:14Z | 2026-04-28T04:42:21Z | 2026-04-26T02:45:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bOsowski | 22,472,628 | MDQ6VXNlcjIyNDcyNjI4 | User | false |
openclaw/openclaw | 4,047,998,882 | I_kwDOQb6kR87xR4-i | 41,451 | https://github.com/openclaw/openclaw/issues/41451 | https://api.github.com/repos/openclaw/openclaw/issues/41451 | memory-lancedb: Add agent-level namespace isolation | ## Problem
All agents share the same LanceDB table (`memories`) at `~/.openclaw/memory/lancedb`. There's no agent ID field on memory entries, so `memory_recall` returns memories from ALL agents regardless of which agent is querying.
**Example:** Family agent (talking to mum) gets back memories about "Raphael game eng... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T21:27:37Z | 2026-04-28T04:42:19Z | 2026-04-26T02:45:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kgeesawor | 4,641,828 | MDQ6VXNlcjQ2NDE4Mjg= | User | false |
openclaw/openclaw | 4,048,122,422 | I_kwDOQb6kR87xSXI2 | 41,469 | https://github.com/openclaw/openclaw/issues/41469 | https://api.github.com/repos/openclaw/openclaw/issues/41469 | Signal channel: attachments fail with LocalMediaAccessError for agent workspaces | ## Bug
Signal channel replies that include local file attachments from agent workspaces fail with:
```
LocalMediaAccessError: Local media path is not under an allowed directory: /Users/.../.openclaw/workspace-bookkeeper/books_20260309.zip
```
## Root Cause
In `deliverReplies` for Signal (the reply delivery path, no... | closed | completed | true | 1 | [] | [] | 2026-03-09T21:58:44Z | 2026-04-28T04:42:19Z | 2026-04-26T02:45:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | skillz-xx | 837,936 | MDQ6VXNlcjgzNzkzNg== | User | false |
openclaw/openclaw | 4,048,232,984 | I_kwDOQb6kR87xSyIY | 41,481 | https://github.com/openclaw/openclaw/issues/41481 | https://api.github.com/repos/openclaw/openclaw/issues/41481 | openai-completions: assistant content sent as array causes recursive nesting on some models | ## Bug Description
When using the `openai-completions` provider (pi-ai), assistant message content is sent as an array of `{type: "text", text: "..."}` objects instead of a plain string. While most models tolerate this, some models — notably **DeepSeek V3.2 via NVIDIA NIM** — mirror the content-block structure literal... | closed | completed | true | 1 | [] | [] | 2026-03-09T22:25:00Z | 2026-04-28T04:42:18Z | 2026-04-26T02:45:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | geraldoaax | 35,141,135 | MDQ6VXNlcjM1MTQxMTM1 | User | false |
openclaw/openclaw | 4,329,761,362 | I_kwDOQb6kR88AAAABAhLqUg | 71,878 | https://github.com/openclaw/openclaw/issues/71878 | https://api.github.com/repos/openclaw/openclaw/issues/71878 | Control UI: mc() history merge discards local messages when server returns empty | ## Summary
The `mc()` message-merge function in the Control UI frontend discards locally-streamed messages when the server's `chat.history` response returns an empty array. This causes the first assistant response in a new session (or after `/new`) to vanish immediately after streaming completes.
This is a **separate... | closed | completed | true | 1 | [] | [] | 2026-04-26T01:40:17Z | 2026-04-28T04:42:17Z | 2026-04-26T02:45:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | WolvenRA | 35,547,932 | MDQ6VXNlcjM1NTQ3OTMy | User | false |
openclaw/openclaw | 4,048,422,139 | I_kwDOQb6kR87xTgT7 | 41,516 | https://github.com/openclaw/openclaw/issues/41516 | https://api.github.com/repos/openclaw/openclaw/issues/41516 | Make typing indicator TTL (typingTtlMs) configurable | ## Problem
The typing indicator TTL (`typingTtlMs`) is hardcoded to 2 minutes in the codebase.
## Impact
When responses take longer than 2 minutes to generate, users experience "ghost typing" — the typing indicator stops even though the agent is still working on a reply.
## Current Behavior
The typing indicator st... | closed | not_planned | true | 1 | [] | [] | 2026-03-09T23:16:34Z | 2026-04-28T04:42:16Z | 2026-04-26T02:45:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | felipematos | 775,850 | MDQ6VXNlcjc3NTg1MA== | User | false |
openclaw/openclaw | 4,048,587,942 | I_kwDOQb6kR87xUIym | 41,553 | https://github.com/openclaw/openclaw/issues/41553 | https://api.github.com/repos/openclaw/openclaw/issues/41553 | [UX] Surface layer-specific diagnostics for Control UI auth/origin vs remote CDP reachability in WSL2 + Windows setups | ### Summary
Improve layer-specific diagnostics for Control UI origin/auth state versus remote Chrome CDP reachability in WSL2 + Windows setups, so users can identify the failing layer faster and avoid misdirected troubleshooting.
### Problem to solve
In WSL2 + Windows + remote Chrome CDP setups, multiple independent ... | closed | not_planned | true | 11 | [
"enhancement"
] | [] | 2026-03-10T00:11:04Z | 2026-04-28T04:42:15Z | 2026-04-26T02:45:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Owlock | 135,854,806 | U_kgDOCBj61g | User | false |
openclaw/openclaw | 4,048,681,182 | I_kwDOQb6kR87xUfje | 41,568 | https://github.com/openclaw/openclaw/issues/41568 | https://api.github.com/repos/openclaw/openclaw/issues/41568 | Process: keep PR scope tight when CI fails for unrelated baseline issues | ## Problem
When a focused PR fails required CI because of pre-existing baseline issues (for example dependency audit state from `main`), contributors end up expanding that PR with unrelated fixes just to make checks green.
That causes scope creep and makes reviews harder.
## Proposed rule
For feature/fix PRs, keep sc... | closed | completed | true | 1 | [] | [] | 2026-03-10T00:43:05Z | 2026-04-28T04:42:15Z | 2026-04-26T02:45:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wanningwei | 4,141,553 | MDQ6VXNlcjQxNDE1NTM= | User | false |
openclaw/openclaw | 4,048,811,396 | I_kwDOQb6kR87xU_WE | 41,587 | https://github.com/openclaw/openclaw/issues/41587 | https://api.github.com/repos/openclaw/openclaw/issues/41587 | [Feature]: Hot Reload Config when included files have changed | ### Summary
Have the config watcher watch an optional settings folder for changes to files so included file changes will also trigger the hot reload process.
### Problem to solve
As the openclaw.json gets longer and longer, it makes sense to break it up into individual files for easier readability. However, the pro... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-10T01:29:25Z | 2026-04-28T04:42:13Z | 2026-04-26T02:47:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | christopherflora | 62,209,728 | MDQ6VXNlcjYyMjA5NzI4 | User | false |
openclaw/openclaw | 4,048,815,802 | I_kwDOQb6kR87xVAa6 | 41,589 | https://github.com/openclaw/openclaw/issues/41589 | https://api.github.com/repos/openclaw/openclaw/issues/41589 | [Feature Request] 支持在 Skill 配置中指定模型 | ## 描述
希期在 Skill 配置中指定使用的模型，类似于 agents 的模型配置方式。
## 背景
目剋剋前 OpenClaw 支持为 agent 指定主模型和备用模型（fallbacks)，但对于单个 Skill 无法指定特定的模型。
## 期望的功能
在 skills.entries.<skill-name> 配置中添加 model 字段，例如：
```json
{
"skills": {
"entries": {
"my-skill": {
"enabled": true,
"model": "deepseek/deepseek-chat"
}... | closed | not_planned | true | 2 | [] | [] | 2026-03-10T01:31:15Z | 2026-04-28T04:42:12Z | 2026-04-26T02:47:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | niuwenkui | 62,731,201 | MDQ6VXNlcjYyNzMxMjAx | User | false |
openclaw/openclaw | 4,048,899,616 | I_kwDOQb6kR87xVU4g | 41,605 | https://github.com/openclaw/openclaw/issues/41605 | https://api.github.com/repos/openclaw/openclaw/issues/41605 | [Bug]: Control UI Skills panel: clicking anywhere triggers filter and hides all skills | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Control UI Skills panel triggers filter event on any click, causing skills list to disappear and show "No skills found"
### Steps to reproduce
1. Open Control UI
2. Switch to the Skills tab
3. Wait for skills list to load (displays normal... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T02:00:45Z | 2026-04-28T04:42:12Z | 2026-04-26T02:47:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dearjnana | 37,565,658 | MDQ6VXNlcjM3NTY1NjU4 | User | false |
openclaw/openclaw | 4,048,943,747 | I_kwDOQb6kR87xVfqD | 41,627 | https://github.com/openclaw/openclaw/issues/41627 | https://api.github.com/repos/openclaw/openclaw/issues/41627 | linux上安装OpenClaw 完整版“云端大脑 + 本地windowns手脚” 通过ssh通讯,如何操作本地文件？ | ### Summary
有没有人在远端linux上部署 ，然后在 win上调用，
我看windos上不支持安装技能的，只要上liunx才能装所有技能放远端，
那问题来了，它就是操作的是liux上的系统如
<img width="2022" height="1007" alt="Image" src="https://github.com/user-attachments/assets/07bc1dd6-c1fb-4832-922c-63bd7d91afb8" />
并不是操作的是本机啊。
--------------------------------------------------
我是通过Windows 建立 SS... | closed | completed | true | 5 | [
"enhancement"
] | [] | 2026-03-10T02:17:59Z | 2026-04-28T04:42:11Z | 2026-04-26T02:47:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gg22mm | 6,648,329 | MDQ6VXNlcjY2NDgzMjk= | User | false |
openclaw/openclaw | 4,048,951,498 | I_kwDOQb6kR87xVhjK | 41,632 | https://github.com/openclaw/openclaw/issues/41632 | https://api.github.com/repos/openclaw/openclaw/issues/41632 | [Feature]: image 工具支持配置图片压缩质量 | ### Summary
image 工具调用视觉模型时自动压缩图片，导致文字密集型文档（如营业执照、合同）识别准确率下降。期望能够配置图片压缩参数（分辨率、质量）
### Problem to solve
当前 image 工具默认对图片进行压缩，但压缩参数不可配置。用户无法控制图片质量。同一模型（qwen3-vl-plus），直接调用 API vs 通过 image 工具调用，识别结果差异明显，后者丢失了大量细节。
### Proposed solution
在 agents.defaults.imageModel 中增加 compression 配置项，支持：
1. 详细模式：{ maxWidth: 2000, max... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-03-10T02:20:21Z | 2026-04-28T04:42:10Z | 2026-04-26T02:47:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tobfen | 6,858,207 | MDQ6VXNlcjY4NTgyMDc= | User | false |
openclaw/openclaw | 4,049,098,928 | I_kwDOQb6kR87xWFiw | 41,675 | https://github.com/openclaw/openclaw/issues/41675 | https://api.github.com/repos/openclaw/openclaw/issues/41675 | Feishu: reply/quoted image content not downloaded in group chat | ## Bug Description
When a user replies to an image message in a Feishu group chat and @mentions the bot, the bot can detect that the quoted message is an image but **cannot see the actual image content**. It only receives a text placeholder like `[image message]`.
## Reproduction Steps
1. In a Feishu group chat with... | closed | not_planned | true | 2 | [] | [] | 2026-03-10T03:13:05Z | 2026-04-28T04:42:09Z | 2026-04-26T02:47:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | CountofGlamorgan | 59,939,121 | MDQ6VXNlcjU5OTM5MTIx | User | false |
openclaw/openclaw | 4,049,116,711 | I_kwDOQb6kR87xWJ4n | 41,679 | https://github.com/openclaw/openclaw/issues/41679 | https://api.github.com/repos/openclaw/openclaw/issues/41679 | [Bug]: [tools] read failed: ENOENT | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
2026.3.8
### Steps to reproduce
不清楚
### Expected behavior
没有报错
### Actual behavior
<img width="1183" height="108" alt="Image" src="https://github.com/user-attachments/assets/865d16ae-27e4-4c46-bdba-a8a6c0619dce" />
### OpenClaw versi... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T03:20:14Z | 2026-04-28T04:42:08Z | 2026-04-26T02:47:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | sobergou | 50,661,423 | MDQ6VXNlcjUwNjYxNDIz | User | false |
openclaw/openclaw | 4,049,218,191 | I_kwDOQb6kR87xWiqP | 41,702 | https://github.com/openclaw/openclaw/issues/41702 | https://api.github.com/repos/openclaw/openclaw/issues/41702 | [Bug]: JSON parsing syntax error when processing fields containing "GPT5.4": Expected ':' after property name | ### Bug type
Regression (worked before, now fails)
### Summary
When attempting to use or submit data related to "GPT5.4" configurations, the system fails to parse the JSON and throws a syntax error. This may be due to issues during JSON serialization or deserialization when handling specific strings (such as fields ... | closed | not_planned | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-10T03:51:41Z | 2026-04-28T04:42:08Z | 2026-04-26T02:47:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Yanxingang | 17,355,243 | MDQ6VXNlcjE3MzU1MjQz | User | false |
openclaw/openclaw | 4,049,373,298 | I_kwDOQb6kR87xXIhy | 41,753 | https://github.com/openclaw/openclaw/issues/41753 | https://api.github.com/repos/openclaw/openclaw/issues/41753 | [Bug]: Intermittent Inbound Telegram DM Delivery Failure Persists in v2026.3. | ### Bug type
Regression (worked before, now fails)
### Summary
Inbound Telegram DMs are still being silently dropped in `v2026.3.8`, requiring the sender to resend the message. This issue persists despite the fix for Telegram DM duplication bugs mentioned in the v2026.3.8 changelog (potentially related to bug #25793... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-10T04:47:50Z | 2026-04-28T04:42:07Z | 2026-04-26T02:47:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | lvonmises | 197,539,581 | U_kgDOC8Y2_Q | User | false |
openclaw/openclaw | 4,049,445,682 | I_kwDOQb6kR87xXaMy | 41,771 | https://github.com/openclaw/openclaw/issues/41771 | https://api.github.com/repos/openclaw/openclaw/issues/41771 | [Feature]: Add image compression config to image tool | ### Summary
The image tool automatically compresses images when calling vision models, reducing accuracy for text-dense documents like business licenses and contracts. Users should be able to configure image compression parameters (resolution, quality).
### Problem to solve
The image tool currently compresses images... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-10T05:12:14Z | 2026-04-28T04:42:06Z | 2026-04-26T02:47:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | tobfen | 6,858,207 | MDQ6VXNlcjY4NTgyMDc= | User | false |
openclaw/openclaw | 4,049,509,652 | I_kwDOQb6kR87xXp0U | 41,790 | https://github.com/openclaw/openclaw/issues/41790 | https://api.github.com/repos/openclaw/openclaw/issues/41790 | Feature: Allow session.mainKey to designate any channel session as the main session | ## Problem
`session.mainKey` exists in the config schema with documentation stating it "overrides the canonical main session key," but it is currently a dead config — `applySessionDefaults()` in `defaults.ts` unconditionally overwrites it to `"main"`, and `resolve-route.ts` hardcodes `DEFAULT_MAIN_KEY = "main"`.
When... | closed | not_planned | true | 1 | [] | [] | 2026-03-10T05:32:25Z | 2026-04-28T04:42:05Z | 2026-04-26T02:48:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Senorsen | 4,176,091 | MDQ6VXNlcjQxNzYwOTE= | User | false |
openclaw/openclaw | 4,049,664,162 | I_kwDOQb6kR87xYPii | 41,827 | https://github.com/openclaw/openclaw/issues/41827 | https://api.github.com/repos/openclaw/openclaw/issues/41827 | [Feature]: | ### Summary
# Custom Provider Integration Fails Silently in Main Agent and Sub-Agent Sessions
## Summary
Custom provider models (LM Studio, Ollama, etc.) configured via `models.providers` fail to execute in both main agent sessions and sub-agent spawns. OpenClaw immediately falls back to cloud providers without atte... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-10T06:14:00Z | 2026-04-28T04:42:04Z | 2026-04-26T02:48:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KyudoAU | 225,513,608 | U_kgDODXEQiA | User | false |
openclaw/openclaw | 4,049,693,821 | I_kwDOQb6kR87xYWx9 | 41,839 | https://github.com/openclaw/openclaw/issues/41839 | https://api.github.com/repos/openclaw/openclaw/issues/41839 | [Bug]: MiniMax M2.5 tool calls have no effect — XML tool invocations are stripped instead of executed | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
When `opencode/minimax-m2.5` is set as the primary model, all tool calls are silently discarded. The agent never executes any tool, regardless of the request.
## Steps to reproduce
1. Configure primary model as `opencode/minimax-m2.5` in `~... | closed | not_planned | true | 2 | [] | [] | 2026-03-10T06:22:29Z | 2026-04-28T04:42:03Z | 2026-04-26T02:48:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mqyang56 | 26,771,507 | MDQ6VXNlcjI2NzcxNTA3 | User | false |
openclaw/openclaw | 4,049,877,967 | I_kwDOQb6kR87xZDvP | 41,881 | https://github.com/openclaw/openclaw/issues/41881 | https://api.github.com/repos/openclaw/openclaw/issues/41881 | [Feature]: Enable Multi-Architecture Docker Image Builds (ARM64/ARMv7) | ### Summary
Build and publish official Docker images for linux/arm64 and linux/arm/v7 architectures alongside the existing linux/amd64 images to support heterogeneous computing environments.
### Problem to solve
Currently, [Project Name] only provides Docker images for the amd64 (x86_64) architecture. This creates s... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-10T07:09:32Z | 2026-04-28T04:42:03Z | 2026-04-26T02:48:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rooshop | 207,657,382 | U_kgDODGCZpg | User | false |
openclaw/openclaw | 4,050,677,729 | I_kwDOQb6kR87xcG_h | 42,025 | https://github.com/openclaw/openclaw/issues/42025 | https://api.github.com/repos/openclaw/openclaw/issues/42025 | [Bug]: Gateway service install failed systemctl enable failed | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Encountered gateway installation issues when deploying using Amazon Linux 2023 kernel-6.1 AMI.
### Steps to reproduce
1. Installed Node.js version 25.
2. Installed libatomic (14.2.1-7) and git (2.50.1-1).
3. Installed openclaw (2026.3.8... | closed | not_planned | true | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-10T09:58:16Z | 2026-04-28T04:42:02Z | 2026-04-26T02:48:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Achai312 | 189,685,398 | U_kgDOC05elg | User | false |
openclaw/openclaw | 4,050,712,421 | I_kwDOQb6kR87xcPdl | 42,033 | https://github.com/openclaw/openclaw/issues/42033 | https://api.github.com/repos/openclaw/openclaw/issues/42033 | [i18n] Missing Chinese (zh-CN) translations in Control UI - Agents page | null | closed | completed | true | 1 | [] | [] | 2026-03-10T10:04:26Z | 2026-04-28T04:42:01Z | 2026-04-26T02:48:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gulumiji | 45,776,320 | MDQ6VXNlcjQ1Nzc2MzIw | User | false |
openclaw/openclaw | 4,050,974,483 | I_kwDOQb6kR87xdPcT | 42,062 | https://github.com/openclaw/openclaw/issues/42062 | https://api.github.com/repos/openclaw/openclaw/issues/42062 | GLM-5 via Ollama: plain-text thinking leaks to users (similar to #33242) | ## Description
Follow-up from #33242. While that issue focuses on Opus 4.6, the same problem occurs with GLM-5 via Ollama — the model "thinks out loud" in plain assistant text, which gets delivered directly to users.
## Current Behavior
The model writes internal reasoning as regular text that bypasses thinking tag s... | closed | completed | true | 1 | [] | [] | 2026-03-10T10:55:10Z | 2026-04-28T04:42:00Z | 2026-04-26T02:48:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Caio-Sc | 53,196,995 | MDQ6VXNlcjUzMTk2OTk1 | User | false |
openclaw/openclaw | 4,051,028,695 | I_kwDOQb6kR87xdcrX | 42,072 | https://github.com/openclaw/openclaw/issues/42072 | https://api.github.com/repos/openclaw/openclaw/issues/42072 | Workspace hooks not loaded for non-default agents | ## Summary
Per-agent workspace hooks (`<workspace>/hooks/`) are not loaded for agents other than the default agent (main).
## Documentation vs Behavior
The [hooks documentation](https://docs.openclaw.ai/automation/hooks) states:
> **Hook Discovery**
> 1. **Workspace hooks**: `<workspace>/hooks/` (per-agent, highest... | closed | not_planned | true | 1 | [] | [] | 2026-03-10T11:06:14Z | 2026-04-28T04:42:00Z | 2026-04-26T02:48:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | JameelHao | 232,899,707 | U_kgDODeHEew | User | false |
openclaw/openclaw | 4,051,320,799 | I_kwDOQb6kR87xej_f | 42,102 | https://github.com/openclaw/openclaw/issues/42102 | https://api.github.com/repos/openclaw/openclaw/issues/42102 | [Bug]: openclaw install hangs after Enable zsh shell completion for openclaw？ | ### Bug type
Crash (process/app exits or hangs)
### Summary
I use curl -fsSL https://openclaw.ai/install.sh | bash to install openclaw in my vps（2c2g ubuntu24.04），but after Enable zsh shell completion for openclaw？I choose yes, then the installer hangs
### Steps to reproduce
1. curl -fsSL https://openclaw.ai/insta... | closed | completed | true | 3 | [
"bug",
"bug:crash"
] | [] | 2026-03-10T12:02:15Z | 2026-04-28T04:41:59Z | 2026-04-26T02:48:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | haozhou-wong | 61,231,331 | MDQ6VXNlcjYxMjMxMzMx | User | false |
openclaw/openclaw | 4,051,560,047 | I_kwDOQb6kR87xfeZv | 42,138 | https://github.com/openclaw/openclaw/issues/42138 | https://api.github.com/repos/openclaw/openclaw/issues/42138 | Feature request: CLI-first auth pool with ordered fallback | ## Problem
OpenClaw users often keep multiple LLM routes:
- multiple providers
- multiple accounts under the same provider
- backup models for quota exhaustion or temporary outages
Today, switching between them or setting up a reliable fallback path is too manual. In practice this means raw config edits, rerunning o... | closed | completed | true | 1 | [] | [] | 2026-03-10T12:44:38Z | 2026-04-28T04:41:58Z | 2026-04-26T02:48:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fantasyengineercdream | 211,698,591 | U_kgDODJ5Dnw | User | false |
openclaw/openclaw | 4,051,748,466 | I_kwDOQb6kR87xgMZy | 42,167 | https://github.com/openclaw/openclaw/issues/42167 | https://api.github.com/repos/openclaw/openclaw/issues/42167 | [Bug] Sidebar tree: collapse/expand toggle not working |
**Description**:
In the DenchClaw/OpenClaw sidebar, clicking on folder arrows to collapse/expand doesn't work. The arrow visually rotates (down → right), but the child items remain visible and don't collapse.
**Steps to reproduce**:
1. Open the sidebar in DenchClaw web interface
2. Find any folder with files (e.g., "... | closed | not_planned | true | 2 | [] | [] | 2026-03-10T13:16:16Z | 2026-04-28T04:41:57Z | 2026-04-26T02:50:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | pepeneif | 8,365,621 | MDQ6VXNlcjgzNjU2MjE= | User | false |
openclaw/openclaw | 4,051,801,852 | I_kwDOQb6kR87xgZb8 | 42,172 | https://github.com/openclaw/openclaw/issues/42172 | https://api.github.com/repos/openclaw/openclaw/issues/42172 | agent RPC runs cannot be aborted via chat.abort | ## Summary
Runs started via the `agent` gateway RPC cannot be aborted by `chat.abort`. The `agent` handler in `src/gateway/server-methods/agent.ts` (`dispatchAgentRunFromGateway`) never registers an `AbortController` in `chatAbortControllers`, so when `chat.abort` is called it cannot find the run and returns `{ aborte... | closed | completed | true | 1 | [] | [] | 2026-03-10T13:25:17Z | 2026-04-28T04:41:57Z | 2026-04-26T02:50:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hitchhiker | 890,543 | MDQ6VXNlcjg5MDU0Mw== | User | false |
openclaw/openclaw | 4,051,832,073 | I_kwDOQb6kR87xgg0J | 42,177 | https://github.com/openclaw/openclaw/issues/42177 | https://api.github.com/repos/openclaw/openclaw/issues/42177 | [Bug]: Reply with a large amount of repetitive content | ### Bug type
Regression (worked before, now fails)
### Summary
<img width="762" height="536" alt="Image" src="https://github.com/user-attachments/assets/1340e9e1-4d5b-472b-8e19-6577a1bcb739" />
### Steps to reproduce
1. tg&feishu
2. 帮我安装某个skill，提供了一个链接，https://clawhub.ai/pskoett/self-improving-agent
### Expected ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-10T13:30:22Z | 2026-04-28T04:41:56Z | 2026-04-26T02:50:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cryptoya | 102,146,039 | U_kgDOBhaf9w | User | false |
openclaw/openclaw | 4,051,949,726 | I_kwDOQb6kR87xg9ie | 42,198 | https://github.com/openclaw/openclaw/issues/42198 | https://api.github.com/repos/openclaw/openclaw/issues/42198 | Gateway crashes on every version upgrade (npm install -g openclaw@latest) | ## Description
Every time OpenClaw is upgraded via `npm install -g openclaw@latest`, the Gateway crashes and cannot restart automatically. Manual intervention is always required to bring the Gateway back online.
## Environment
- **OS:** Ubuntu 22.04 (Linux 6.8.0-101-generic x64)
- **Node:** v22.22.1
- **Upgrade path... | closed | completed | true | 1 | [] | [] | 2026-03-10T13:50:20Z | 2026-04-28T04:41:55Z | 2026-04-26T02:50:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Vincent-Tainan | 48,849,497 | MDQ6VXNlcjQ4ODQ5NDk3 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.