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 | 4,323,706,877 | I_kwDOQb6kR88AAAABAbaH_Q | 71,140 | https://github.com/openclaw/openclaw/issues/71140 | https://api.github.com/repos/openclaw/openclaw/issues/71140 | Telegram outbound replies leak/drop react/media directives instead of applying native reactions/GIFs | ## Summary
Telegram outbound delivery drops or leaks reaction / media directives instead of turning them into native Telegram behavior.
In live Telegram use, replies like:
- `[[react_to_current:🔥]]`
- `MEDIA:/path/to/file.gif`
were being rendered as literal text in chat or partially dropped, even though the Telegra... | open | null | false | 2 | [] | [] | 2026-04-24T14:49:30Z | 2026-05-03T04:59:46Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | Cuttingwater | 226,908,709 | U_kgDODYZaJQ | User | false |
openclaw/openclaw | 4,323,742,131 | I_kwDOQb6kR88AAAABAbcRsw | 71,142 | https://github.com/openclaw/openclaw/issues/71142 | https://api.github.com/repos/openclaw/openclaw/issues/71142 | Feature Request: Configurable upload size limit for Control UI | ### Summary
## Is your feature request related to a problem? Please describe.
The Control UI chat window currently has a hardcoded file upload size limit of **5MB**. This prevents users from uploading larger images directly through the web interface, even though:
- The media understanding capability supports images up... | open | null | false | 6 | [
"enhancement"
] | [] | 2026-04-24T14:55:25Z | 2026-05-03T04:59:48Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | XShaoxinjun | 270,322,175 | U_kgDOEBzJ_w | User | false |
openclaw/openclaw | 4,323,744,616 | I_kwDOQb6kR88AAAABAbcbaA | 71,143 | https://github.com/openclaw/openclaw/issues/71143 | https://api.github.com/repos/openclaw/openclaw/issues/71143 | memory_search auto-selects Bedrock after update and fails with Could not load credentials from any providers when only local memory/session recall is needed | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After updating to OpenClaw `2026.4.22`, the built-in `memory_search` tool stopped working in a local direct session, even though the user expectation is only to search local memory/session context.
The tool now fails with:
... | closed | completed | true | 1 | [
"bug",
"regression",
"bug:behavior"
] | [] | 2026-04-24T14:55:51Z | 2026-04-27T04:41:55Z | 2026-04-25T01:47:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DTKJS | 240,723,690 | U_kgDODlkm6g | User | false |
openclaw/openclaw | 4,211,409,509 | I_kwDOQb6kR877BQJl | 61,844 | https://github.com/openclaw/openclaw/issues/61844 | https://api.github.com/repos/openclaw/openclaw/issues/61844 | openclaw update does not sync embedded openclaw copy in ~/.openclaw/extensions/node_modules/ | ## Describe the bug
When upgrading openclaw via `npm install -g openclaw@<version>`, the global binary at `/usr/local/lib/node_modules/openclaw` is updated correctly. However, the embedded openclaw copy used by the extensions runtime at `~/.openclaw/extensions/node_modules/openclaw` is **not updated** and remains on t... | closed | completed | true | 0 | [] | [] | 2026-04-06T11:39:45Z | 2026-04-27T05:09:07Z | 2026-04-24T14:56:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | gong512-cmyk | 259,608,828 | U_kgDOD3lQ_A | User | false |
openclaw/openclaw | 4,323,754,937 | I_kwDOQb6kR88AAAABAbdDuQ | 71,144 | https://github.com/openclaw/openclaw/issues/71144 | https://api.github.com/repos/openclaw/openclaw/issues/71144 | CLI `openclaw memory status` fails with "Unknown memory embedding provider: local" | ## Bug Description
`openclaw memory status` fails with:
```
[openclaw] Failed to start CLI: Error: Unknown memory embedding provider: local
at getAdapter (file:///.../dist/manager-xvPcitvj.js:143:22)
at createEmbeddingProvider (...)
at MemoryIndexManager.loadProviderResult (...)
```
The gateway's `memory_search` ... | closed | completed | true | 1 | [] | [] | 2026-04-24T14:57:33Z | 2026-04-28T04:46:55Z | 2026-04-25T19:31:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gzxmren | 41,671,592 | MDQ6VXNlcjQxNjcxNTky | User | false |
openclaw/openclaw | 4,323,299,077 | I_kwDOQb6kR88AAAABAbBPBQ | 71,116 | https://github.com/openclaw/openclaw/issues/71116 | https://api.github.com/repos/openclaw/openclaw/issues/71116 | Architecture review: credential governance & undocumented coupling hotspots | ### Summary
Document the credential-persistence contract and add narrative guidance in the high-privilege ghost-hotspot zones identified by architectural analysis (src/gateway/server-methods/secrets.ts and per-extension credentials.ts files).
### Problem to solve
The highest-privilege code paths in the repo — `src/... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-04-24T13:39:27Z | 2026-05-03T01:22:51Z | null | CONTRIBUTOR | null | 20260503T030030Z | 2026-05-03T03:00:30Z | davidangularme | 18,486,579 | MDQ6VXNlcjE4NDg2NTc5 | User | false |
openclaw/openclaw | 4,323,825,705 | I_kwDOQb6kR88AAAABAbhYKQ | 71,148 | https://github.com/openclaw/openclaw/issues/71148 | https://api.github.com/repos/openclaw/openclaw/issues/71148 | Memory leak: Ajv `.compile()` per call in `sharedAjv` (and `ajv`) leaks SchemaEnv/run closures | ## Summary
`sharedAjv.compile(schema)` (and the module-local `ajv.compile(...)` in `llm_invoke.ts`) is called on every command/workflow-step/validate invocation, with no memoization and no call to `ajv.removeSchema()` afterward. Each `.compile()` leaves behind a durable graph of internal Ajv objects (`SchemaEnv`, `Val... | closed | completed | true | 2 | [] | [
"vincentkoc"
] | 2026-04-24T15:08:05Z | 2026-04-29T04:32:41Z | 2026-04-27T02:04:03Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | cmi525 | 278,865,831 | U_kgDOEJ8npw | User | false |
openclaw/openclaw | 4,323,891,335 | I_kwDOQb6kR88AAAABAblYhw | 71,152 | https://github.com/openclaw/openclaw/issues/71152 | https://api.github.com/repos/openclaw/openclaw/issues/71152 | [Withdrawn] Feature request withdrawn by reporter | Closed by reporter — withdrawing this feature request.
This was submitted without authorization from the organization on whose behalf it was filed. The technical content has been withdrawn. Please disregard and, if possible, delete this issue entirely.
Thank you. | closed | not_planned | true | 1 | [] | [] | 2026-04-24T15:20:17Z | 2026-04-27T04:59:28Z | 2026-04-24T18:13:36Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ACRGroup | 271,896,004 | U_kgDOEDTNxA | User | false |
openclaw/openclaw | 4,323,909,196 | I_kwDOQb6kR88AAAABAbmeTA | 71,153 | https://github.com/openclaw/openclaw/issues/71153 | https://api.github.com/repos/openclaw/openclaw/issues/71153 | exec allowlist: multi-line (`\<newline>`) commands fail with "allowlist miss" even when binary is allowlisted | ## Summary
When an agent's `exec` tool receives a shell command whose body contains a POSIX `\<newline>` line continuation (for example a multi-line `curl -X POST` with `-H`/`-d` on separate lines), the exec harness returns:
```json
{ "status": "error", "tool": "exec", "error": "exec denied: allowlist miss" }
```
ev... | closed | not_planned | true | 2 | [] | [] | 2026-04-24T15:23:35Z | 2026-04-29T04:33:59Z | 2026-04-26T21:31:58Z | CONTRIBUTOR | null | 20260429T060025Z | 2026-04-29T06:00:25Z | scottgl9 | 10,623,288 | MDQ6VXNlcjEwNjIzMjg4 | User | false |
openclaw/openclaw | 4,323,838,202 | I_kwDOQb6kR88AAAABAbiI-g | 71,150 | https://github.com/openclaw/openclaw/issues/71150 | https://api.github.com/repos/openclaw/openclaw/issues/71150 | [Bug]: DeepSeek v4-pro thinking mode returns HTTP 400 "reasoning_content must be passed back" on tool calls | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using DeepSeek `deepseek-v4-pro` (or `deepseek-reasoner`) in thinking mode with tool calls, OpenClaw does not echo `reasoning_content` back to the DeepSeek API in subsequent requests of the same tool-loop.... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-24T15:10:14Z | 2026-04-27T05:09:04Z | 2026-04-24T15:24:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | romannekrasovaillm | 201,476,332 | U_kgDODAJI7A | User | false |
openclaw/openclaw | 3,887,139,152 | I_kwDOQb6kR87nsQlQ | 7,309 | https://github.com/openclaw/openclaw/issues/7309 | https://api.github.com/repos/openclaw/openclaw/issues/7309 | [Feature]: Support DeepSeek API as a first-class LLM provider | ## Summary
Many users now have access to DeepSeek's OpenAI-compatible API and would like to use DeepSeek models directly in OpenClaw, without confusing the setup flow or manually hacking config files.
Currently, onboarding and provider selection do not offer an explicit "DeepSeek" provider. Users are not sure whether... | closed | completed | true | 7 | [
"enhancement"
] | [] | 2026-02-02T17:32:26Z | 2026-04-27T05:09:02Z | 2026-04-24T15:25:04Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | farazoman | 2,703,043 | MDQ6VXNlcjI3MDMwNDM= | User | false |
openclaw/openclaw | 4,159,486,861 | I_kwDOQb6kR8737LuN | 56,369 | https://github.com/openclaw/openclaw/issues/56369 | https://api.github.com/repos/openclaw/openclaw/issues/56369 | [Bug]: gateway:startup hook fires before channel adapters are ready — message tool fails silently | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The gateway:startup hook fires before the Telegram adapter is fully initialized; any message sent via the message tool in this hook is silently dropped with no error or retry.
### Steps to reproduce
1. Config... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-28T11:36:01Z | 2026-05-01T04:43:26Z | 2026-04-28T14:04:05Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | nailujac | 83,184,419 | MDQ6VXNlcjgzMTg0NDE5 | User | false |
openclaw/openclaw | 4,323,950,074 | I_kwDOQb6kR88AAAABAbo9-g | 71,157 | https://github.com/openclaw/openclaw/issues/71157 | https://api.github.com/repos/openclaw/openclaw/issues/71157 | [Feature]: Support WhatsApp disappearing-message expiration for outbound replies | ### Summary
Allow WhatsApp channel/account config to apply a default disappearing-message expiration to outbound OpenClaw replies.
### Problem to solve
WhatsApp supports disappearing messages, but OpenClaw’s WhatsApp outbound sends do not currently expose a reliable way to apply an expiration to agent replies. In ch... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-24T15:31:36Z | 2026-05-03T04:59:50Z | null | CONTRIBUTOR | null | 20260503T060030Z | 2026-05-03T06:00:30Z | dougvk | 401,660 | MDQ6VXNlcjQwMTY2MA== | User | false |
openclaw/openclaw | 4,323,882,173 | I_kwDOQb6kR88AAAABAbk0vQ | 71,151 | https://github.com/openclaw/openclaw/issues/71151 | https://api.github.com/repos/openclaw/openclaw/issues/71151 | . | . | closed | completed | true | 0 | [] | [] | 2026-04-24T15:18:24Z | 2026-04-27T04:55:17Z | 2026-04-24T19:22:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ACRGroup | 271,896,004 | U_kgDOEDTNxA | User | false |
openclaw/openclaw | 4,209,422,912 | I_kwDOQb6kR8765rJA | 61,667 | https://github.com/openclaw/openclaw/issues/61667 | https://api.github.com/repos/openclaw/openclaw/issues/61667 | [Bug]: Exact command hashing in `exec` approvals renders `allow-always` ineffective for dynamic commands | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Here is the standardized English GitHub Issue, formatted and ready to be copied and pasted directly into the repository's issue tracker:
***
**Title:** [Bug]: Exact command hashing in `exec` approvals renders... | closed | not_planned | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-06T03:53:37Z | 2026-05-02T04:28:35Z | 2026-04-29T09:50:16Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | heronerar | 128,018,749 | U_kgDOB6FpPQ | User | false |
openclaw/openclaw | 3,884,051,569 | I_kwDOQb6kR87ngexx | 6,900 | https://github.com/openclaw/openclaw/issues/6900 | https://api.github.com/repos/openclaw/openclaw/issues/6900 | [Feature]: docker support for node | ## Summary
I'm trying to add a node that is containerized.
I'm not sure if this is possible to achieve currently.
The problem is that-
1. If I try to run `openclaw node install`, it will fail with `Node service check failed: Error: systemctl --user unavailable: unknown error`
2. If I directly `openclaw node run` with... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T05:22:00Z | 2026-04-27T05:09:01Z | 2026-04-24T15:16:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | shichuzhu | 6,067,848 | MDQ6VXNlcjYwNjc4NDg= | User | false |
openclaw/openclaw | 4,174,592,016 | I_kwDOQb6kR8740zgQ | 57,980 | https://github.com/openclaw/openclaw/issues/57980 | https://api.github.com/repos/openclaw/openclaw/issues/57980 | Test approval E2E | Testing the full approval and execution flow | closed | completed | true | 0 | [] | [] | 2026-03-30T23:08:59Z | 2026-04-27T05:09:01Z | 2026-04-24T15:52:55Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | buremba | 82,745 | MDQ6VXNlcjgyNzQ1 | User | false |
openclaw/openclaw | 4,324,158,498 | I_kwDOQb6kR88AAAABAb1sIg | 71,165 | https://github.com/openclaw/openclaw/issues/71165 | https://api.github.com/repos/openclaw/openclaw/issues/71165 | browser: primary-agent tabs never cleaned up — cleanupBrowserSessionsForLifecycleEnd not wired for primary sessions | ## Summary
`cleanupBrowserSessionsForLifecycleEnd()` is only called from two places in runtime code:
- `src/agents/subagent-registry-lifecycle.ts:611` — subagent completion
- `src/gateway/server-cron.ts:305` — cron-isolated agent turn completion
Primary agents' interactive (channel/paired) sessions have no correspon... | closed | completed | true | 1 | [] | [] | 2026-04-24T16:07:51Z | 2026-04-27T04:41:17Z | 2026-04-25T02:08:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dwbutler | 746,336 | MDQ6VXNlcjc0NjMzNg== | User | false |
openclaw/openclaw | 3,852,531,903 | I_kwDOQb6kR87loPi_ | 1,691 | https://github.com/openclaw/openclaw/issues/1691 | https://api.github.com/repos/openclaw/openclaw/issues/1691 | Add option to disable prompt_cache_key for local models | ## Feature Request
When using local models (e.g., vLLM with MiniMax-M2.1), the `prompt_cache_key` parameter may not be supported by the local inference server.
### Problem
Local inference servers like vLLM may not recognize or handle `prompt_cache_key` in API requests, potentially causing errors or unexpected behavio... | closed | completed | true | 5 | [
"enhancement",
"pi-issue"
] | [
"steipete"
] | 2026-01-25T04:09:48Z | 2026-04-27T05:09:00Z | 2026-04-24T15:12:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | fugarty | 255,100,796 | U_kgDODzSHfA | User | false |
openclaw/openclaw | 3,883,046,754 | I_kwDOQb6kR87ncpdi | 6,617 | https://github.com/openclaw/openclaw/issues/6617 | https://api.github.com/repos/openclaw/openclaw/issues/6617 | [Feature]: Control UI agent reply supports inline image | ## Summary
Our agent is not able to reply us with any inline image in their markdown response.
## Proposed solution
We should also support the same Markdown syntax that GitHub-flavored Markdown supports.
## Alternatives considered
Otherwise, an image can be a more special attachment that gets sent and rendered.
| closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-01T21:46:49Z | 2026-04-27T05:09:00Z | 2026-04-24T15:12:55Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | boyangwang | 2,408,273 | MDQ6VXNlcjI0MDgyNzM= | User | false |
openclaw/openclaw | 3,883,178,566 | I_kwDOQb6kR87ndJpG | 6,662 | https://github.com/openclaw/openclaw/issues/6662 | https://api.github.com/repos/openclaw/openclaw/issues/6662 | [UX] SubAgent results should display as Tool Output instead of user message | ## Problem Description
Currently, when a SubAgent task completes, its results are injected into the main session conversation as a `role: "user"` message. In the WebUI, this makes it appear as if the user themselves sent the SubAgent's raw output, which is confusing.
## Current Behavior
1. User assigns task to Maste... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-01T22:53:49Z | 2026-04-27T05:08:59Z | 2026-04-24T15:16:14Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww | 35,328,952 | MDQ6VXNlcjM1MzI4OTUy | User | false |
openclaw/openclaw | 3,883,205,743 | I_kwDOQb6kR87ndQRv | 6,675 | https://github.com/openclaw/openclaw/issues/6675 | https://api.github.com/repos/openclaw/openclaw/issues/6675 | [Security] Baseline hardening: headers, CI scanning, and local assets | ## Summary
Following the [CLAWDINATOR guidance](https://github.com/openclaw/openclaw/pull/6171#issuecomment-3831047989) to open an issue first before submitting security-related PRs. This proposal bundles 4 complementary security hardening improvements that reduce supply-chain and runtime risks.
## Motivation
OpenCl... | closed | completed | true | 2 | [
"bug",
"security"
] | [] | 2026-02-01T23:04:42Z | 2026-04-27T05:08:58Z | 2026-04-24T15:16:17Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | unisone | 32,521,398 | MDQ6VXNlcjMyNTIxMzk4 | User | false |
openclaw/openclaw | 3,883,254,965 | I_kwDOQb6kR87ndcS1 | 6,691 | https://github.com/openclaw/openclaw/issues/6691 | https://api.github.com/repos/openclaw/openclaw/issues/6691 | [Feature Request] Context-aware dynamic skill loading/unloading to reduce token cost | ## Summary
Add support for dynamically loading and unloading skills during a session based on conversation context, to reduce token costs without sacrificing functionality.
## Background
This is a common pain point in the community:
- Discussion #1949: Users report "burning through $$$" even with Haiku
- Issue #1594... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-02-01T23:27:12Z | 2026-04-27T05:08:57Z | 2026-04-24T15:16:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | bellarivialabs | 137,748,615 | U_kgDOCDXghw | User | false |
openclaw/openclaw | 3,883,390,728 | I_kwDOQb6kR87nd9cI | 6,717 | https://github.com/openclaw/openclaw/issues/6717 | https://api.github.com/repos/openclaw/openclaw/issues/6717 | [Feature Request] Natural language model switching in conversation | ## Problem
Currently, switching between models requires either:
- CLI commands (`openclaw models set`)
- The `/model` slash command
- Editing config
There's no way to say "use local for this" or "switch to the cheap model" in natural conversation and have the agent actually switch.
## Use Case
Users with both a pai... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T00:20:27Z | 2026-04-27T05:08:55Z | 2026-04-24T15:16:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | neilofneils404 | 258,699,186 | U_kgDOD2tvsg | User | false |
openclaw/openclaw | 3,883,626,709 | I_kwDOQb6kR87ne3DV | 6,808 | https://github.com/openclaw/openclaw/issues/6808 | https://api.github.com/repos/openclaw/openclaw/issues/6808 | [Feature]: More efficient cron job management on Web UI | ## Summary
1、Edit existing cron jobs via a Web interface.
2、View execution history for easier debugging.
3、Support cron job pipeline orchestration.
| closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T02:41:47Z | 2026-04-27T05:08:55Z | 2026-04-24T15:16:28Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | hyfree | 26,597,853 | MDQ6VXNlcjI2NTk3ODUz | User | false |
openclaw/openclaw | 4,124,192,659 | I_kwDOQb6kR8710i-T | 53,216 | https://github.com/openclaw/openclaw/issues/53216 | https://api.github.com/repos/openclaw/openclaw/issues/53216 | WhatsApp plugin broken after upgrading to 2026.3.22 | ## Bug Report
**Version:** 2026.3.22 (downgraded back to 2026.3.13 as workaround)
**Platform:** macOS (Apple Silicon)
**Channel:** WhatsApp
### What happened
After upgrading from 2026.3.22, the WhatsApp plugin stopped working entirely. Downgrading back to 2026.3.13 restored functionality.
### Suspected cause
The 2... | closed | completed | true | 6 | [] | [] | 2026-03-23T23:01:44Z | 2026-04-27T05:08:53Z | 2026-04-24T16:10:04Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | BrunoCerberus | 10,541,956 | MDQ6VXNlcjEwNTQxOTU2 | User | false |
openclaw/openclaw | 4,324,174,566 | I_kwDOQb6kR88AAAABAb2q5g | 71,167 | https://github.com/openclaw/openclaw/issues/71167 | https://api.github.com/repos/openclaw/openclaw/issues/71167 | [Bug]: Live gateway plugin shows as loaded but register(api) never executes in WhatsApp-connected runtime | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
A live OpenClaw WhatsApp setup shows a custom runtime plugin as loaded and JITI-compiled, but the plugin does not appear to execute in the live gateway process: no trace files are written and even a direct sentinel inside re... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-24T16:10:34Z | 2026-04-27T04:43:56Z | 2026-04-25T01:02:14Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ChiccoFM | 279,038,608 | U_kgDOEKHKkA | User | false |
openclaw/openclaw | 4,324,173,223 | I_kwDOQb6kR88AAAABAb2lpw | 71,166 | https://github.com/openclaw/openclaw/issues/71166 | https://api.github.com/repos/openclaw/openclaw/issues/71166 | [Feature]: [IMP] true read-only mode for whatsapp (no outbound messages) | ### Summary
Add a true read-only WhatsApp mode where OpenClaw can receive, read and analyze inbound messages but is strictly blocked from sending any outbound WhatsApp messages automatically including pairing prompts, access-not-configured notices or any other arbitraru replies
### Problem to solve
When a WhatsApp ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-24T16:10:22Z | 2026-04-29T04:36:52Z | 2026-04-26T10:54:06Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | sneha-rudra | 17,240,386 | MDQ6VXNlcjE3MjQwMzg2 | User | false |
openclaw/openclaw | 3,885,132,293 | I_kwDOQb6kR87nkmoF | 7,052 | https://github.com/openclaw/openclaw/issues/7052 | https://api.github.com/repos/openclaw/openclaw/issues/7052 | Feature Request: Auto-summarize/compress context before hitting token limit to prevent brain freeze | ## Problem
When a long session fills up the context window, the agent hits a hard wall — it can't process anything, including saving its own state. The only recovery is a session reset (`/new` or `/reset`), which:
1. **Nukes all in-flight work** — any tool chains mid-execution are lost
2. **Orphans tool results** — t... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T10:20:52Z | 2026-04-27T05:08:52Z | 2026-04-24T16:14:04Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pstabell | 213,886,298 | U_kgDODL-lWg | User | false |
openclaw/openclaw | 3,885,158,763 | I_kwDOQb6kR87nktFr | 7,056 | https://github.com/openclaw/openclaw/issues/7056 | https://api.github.com/repos/openclaw/openclaw/issues/7056 | [Feature]: I think the current Clawdbot architecture is flawed. A correct architecture should place the operational layer at the top, but only allow operations on things like Docker containers or virtual machines. | I think the current Clawdbot architecture is flawed. A correct architecture would place the operational layer at the top, but only allow operations on things like Docker containers or virtual machines.
Has anyone already created a plugin that restricts Clawdbot to only interacting with things inside containers? If not,... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T10:26:20Z | 2026-04-27T05:08:52Z | 2026-04-24T16:14:07Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | constansino | 65,108,260 | MDQ6VXNlcjY1MTA4MjYw | User | false |
openclaw/openclaw | 3,885,445,733 | I_kwDOQb6kR87nlzJl | 7,095 | https://github.com/openclaw/openclaw/issues/7095 | https://api.github.com/repos/openclaw/openclaw/issues/7095 | Feature Request: Support multimodal content blocks (images) in gateway agent method | ## Summary
The gateway's `agent` method currently only accepts `message` as a string (`NonEmptyString`), which prevents sending images/multimodal content to the AI.
## Current Behavior
In `gateway/protocol/schema/agent.js`, the `AgentParamsSchema` validates:
```javascript
export const AgentParamsSchema = Type.Objec... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T11:33:48Z | 2026-04-27T05:08:51Z | 2026-04-24T16:14:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | CaseReed | 30,327,326 | MDQ6VXNlcjMwMzI3MzI2 | User | false |
openclaw/openclaw | 3,886,763,676 | I_kwDOQb6kR87nq06c | 7,249 | https://github.com/openclaw/openclaw/issues/7249 | https://api.github.com/repos/openclaw/openclaw/issues/7249 | [Feature]: Support Claude Models via Azure service | ## Summary
I want to use [Feature]: Claude Models via Azure service
## Proposed solution
Allow entry of both Azure API endpoint and API key for running claude-sonnet-4-5 and claude-opus-4-5
## Alternatives considered
Your current anthropic support allows entry of API key only, I would like to customise the API end... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-02-02T16:09:09Z | 2026-04-27T05:08:50Z | 2026-04-24T16:14:15Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | eliranwong | 25,262,722 | MDQ6VXNlcjI1MjYyNzIy | User | false |
openclaw/openclaw | 3,886,822,539 | I_kwDOQb6kR87nrDSL | 7,260 | https://github.com/openclaw/openclaw/issues/7260 | https://api.github.com/repos/openclaw/openclaw/issues/7260 | Improved compaction summarization with task preservation | ## Problem
When compaction occurs, the summary often loses critical context:
1. **"Summary unavailable due to context limits"** - Complete failure to summarize
2. **Active tasks forgotten** - Summary captures conclusions but not in-progress work
3. **Recent user requests lost** - The last few messages before compacti... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-02-02T16:20:37Z | 2026-04-27T05:08:50Z | 2026-04-24T16:14:19Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | joetomasone | 56,984,887 | MDQ6VXNlcjU2OTg0ODg3 | User | false |
openclaw/openclaw | 3,886,977,535 | I_kwDOQb6kR87nrpH_ | 7,284 | https://github.com/openclaw/openclaw/issues/7284 | https://api.github.com/repos/openclaw/openclaw/issues/7284 | [Feature Request] Per-model tool policies for safe fallback behavior | ## Problem
Currently, tool policies are static configuration that apply globally regardless of which model is active. This creates a dilemma:
- **Small local models** (≤300B params) are blocked as fallbacks when web tools are enabled, even though they could safely handle most tasks
- **Manual workaround** requires ed... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-02-02T16:53:12Z | 2026-04-27T05:08:49Z | 2026-04-24T16:14:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | neilofneils404 | 258,699,186 | U_kgDOD2tvsg | User | false |
openclaw/openclaw | 4,175,258,794 | I_kwDOQb6kR8743WSq | 58,072 | https://github.com/openclaw/openclaw/issues/58072 | https://api.github.com/repos/openclaw/openclaw/issues/58072 | Image display broken: Cannot find package sharp | Through WeChat, when an image is sent to the AI assistant, it cannot be read. Error: `Cannot find package 'sharp' imported from openclaw/dist/auth-profiles-DRjqKE3G.js`
Please add sharp as a dependency or fix the image processing module. | closed | completed | true | 1 | [] | [] | 2026-03-31T02:35:55Z | 2026-04-27T05:08:48Z | 2026-04-24T16:15:43Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | assokj | 151,482,522 | U_kgDOCQdwmg | User | false |
openclaw/openclaw | 4,162,394,311 | I_kwDOQb6kR874GRjH | 56,774 | https://github.com/openclaw/openclaw/issues/56774 | https://api.github.com/repos/openclaw/openclaw/issues/56774 | [Bug] Image viewer broken - Cannot find package 'sharp' | ## Bug Description
Image reading via `read` tool is broken. Error: Cannot find package 'sharp'.
**Environment:**
- Platform: Windows 11
- QClaw desktop app
- Channel: WeChat
**Steps:**
1. User sends image via WeChat
2. AI attempts to read image
3. Error occurs
**Impact:**
Users cannot share screenshots for AI help.
| closed | completed | true | 1 | [] | [] | 2026-03-29T03:48:47Z | 2026-04-27T05:08:47Z | 2026-04-24T16:15:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | XzXy1 | 207,175,940 | U_kgDODFlBBA | User | false |
openclaw/openclaw | 4,087,408,594 | I_kwDOQb6kR87zoOfS | 48,903 | https://github.com/openclaw/openclaw/issues/48903 | https://api.github.com/repos/openclaw/openclaw/issues/48903 | [pnpm] Upgrade 2026.3.8→2026.3.13 fails: pi-coding-agent missing @sinclair/typebox | **Problem**: OpenClaw 2026.3.13 fails to start after upgrade from 2026.3.8.
**Error**:
```
Error [ERR_MODULE_NOT_FOUND]: Cannot find package '@sinclair/typebox' imported from pi-coding-agent
```
**Steps to reproduce**:
1. Have OpenClaw 2026.3.8 working
2. Run: `pnpm add -g openclaw@2026.3.13`
3. Try: `openclaw gatewa... | closed | completed | true | 2 | [] | [] | 2026-03-17T09:52:43Z | 2026-04-27T05:08:46Z | 2026-04-24T16:15:47Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | stoneleeclaw | 265,713,117 | U_kgDOD9Z13Q | User | false |
openclaw/openclaw | 4,105,482,835 | I_kwDOQb6kR870tLJT | 50,800 | https://github.com/openclaw/openclaw/issues/50800 | https://api.github.com/repos/openclaw/openclaw/issues/50800 | [Bug]: Self-built Docker runtime image fails to start with `ERR_MODULE_NOT_FOUND` for `axios` | ### Bug type
Crash (process/app exits or hangs)
### Summary
A Docker runtime image built from the current source tree fails to start because Node.js cannot resolve `axios` from `@line/bot-sdk`.
### Steps to reproduce
1. Build a runtime image from the current source tree using the local Docker build flow.
2.... | closed | completed | true | 6 | [
"bug",
"bug:crash"
] | [] | 2026-03-20T03:27:31Z | 2026-04-27T05:08:46Z | 2026-04-24T16:15:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | hsiong | 37,357,447 | MDQ6VXNlcjM3MzU3NDQ3 | User | false |
openclaw/openclaw | 4,230,757,824 | I_kwDOQb6kR878LD3A | 63,639 | https://github.com/openclaw/openclaw/issues/63639 | https://api.github.com/repos/openclaw/openclaw/issues/63639 | [Bug]: v2026.4.9 npm install missing qa/scenarios/index.md causes CLI startup failure | ## Description
After updating to `2026.4.9` via npm global install, the CLI fails to start with:
```
Completion cache update failed ([openclaw] Failed to start CLI: Error: qa scenario pack not found: qa/scenarios/index.md
at readQaScenarioPack (file:///home/filip/.npm-global/lib/node_modules/openclaw/dist/suite-B... | closed | completed | true | 1 | [] | [] | 2026-04-09T08:36:50Z | 2026-04-27T05:08:41Z | 2026-04-24T16:20:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | copernicus-agent | 263,260,481 | U_kgDOD7EJQQ | User | false |
openclaw/openclaw | 4,194,155,463 | I_kwDOQb6kR875_bvH | 59,655 | https://github.com/openclaw/openclaw/issues/59655 | https://api.github.com/repos/openclaw/openclaw/issues/59655 | [Bug] v2026.4.1 build artifact hash mismatch breaks all Feishu message dispatch | ## Description
After upgrading from `2026.3.23-2` to `2026.4.1`, **all** Feishu message dispatch fails with `ERR_MODULE_NOT_FOUND`. This affects both text and image messages across all configured Feishu agents (default + custom agents).
## Environment
- **OpenClaw version**: 2026.4.1
- **Previous working version**: ... | closed | completed | true | 3 | [] | [] | 2026-04-02T12:06:20Z | 2026-04-27T05:08:40Z | 2026-04-24T16:20:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | RayXbot | 273,091,045 | U_kgDOEEcJ5Q | User | false |
openclaw/openclaw | 3,951,716,793 | I_kwDOQb6kR87rimm5 | 19,065 | https://github.com/openclaw/openclaw/issues/19065 | https://api.github.com/repos/openclaw/openclaw/issues/19065 | Audit safeguard logs destructive config writes but doesn't block them | ## Summary
The gateway's config audit detects destructive writes to `openclaw.json` but only logs them — it doesn't prevent the write. This has caused two outages in three days.
## What happened
At ~05:01 UTC on 2026-02-17, the gateway truncated `openclaw.json` from 5,191 to 302 bytes. The audit log flagged it:
```... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-02-17T09:24:48Z | 2026-04-24T16:23:13Z | 2026-03-22T04:08:38Z | NONE | null | 20260424T180017Z | 2026-04-24T18:00:17Z | jtczville | 22,560,163 | MDQ6VXNlcjIyNTYwMTYz | User | false |
openclaw/openclaw | 3,923,019,389 | I_kwDOQb6kR87p1IZ9 | 13,583 | https://github.com/openclaw/openclaw/issues/13583 | https://api.github.com/repos/openclaw/openclaw/issues/13583 | [Feature] Pre-response enforcement hooks (hard gates) for mandatory tool-call / policy rules | ## Summary
OpenClaw currently treats "must call tool X before responding" rules as *soft* instructions embedded in prompts. In high-stakes workflows (quant/finance, security, operations), soft rules are not acceptable: the agent must be **mechanically prevented** from emitting a final answer until required checks/tools... | open | null | false | 10 | [
"enhancement"
] | [] | 2026-02-10T18:45:52Z | 2026-05-05T02:00:24Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | JamieMolty | 260,740,338 | U_kgDOD4qU8g | User | false |
openclaw/openclaw | 3,883,677,655 | I_kwDOQb6kR87nfDfX | 6,823 | https://github.com/openclaw/openclaw/issues/6823 | https://api.github.com/repos/openclaw/openclaw/issues/6823 | Feature Request: Execution Guardrails for Tool Safety | # Feature Request: Execution Guardrails for Tool Safety
## Summary
Add configurable guardrails that intercept and validate tool executions (especially `exec`) before they run, with optional human-in-the-loop escalation for critical actions.
## Problem
Currently, an AI agent can execute any command via the `exec` to... | closed | completed | true | 9 | [
"enhancement"
] | [] | 2026-02-02T03:01:29Z | 2026-04-27T05:08:38Z | 2026-04-24T15:16:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | chiquitinbot | 258,466,115 | U_kgDOD2fhQw | User | false |
openclaw/openclaw | 3,981,818,149 | I_kwDOQb6kR87tVbkl | 25,145 | https://github.com/openclaw/openclaw/issues/25145 | https://api.github.com/repos/openclaw/openclaw/issues/25145 | Feature request: Outbound message approval gate (pre-send hook for message/exec tools) | ## Problem
When an AI agent has access to outbound communication tools (email via exec, SMS via exec, Telegram message tool), there's no built-in mechanism to require human approval **before** a message is sent.
Current exec-approvals gate based on **binary path**, which means shell scripts run through `/bin/bash` b... | closed | completed | true | 14 | [] | [] | 2026-02-24T06:29:54Z | 2026-04-24T16:23:15Z | 2026-03-29T01:47:44Z | NONE | null | 20260424T180017Z | 2026-04-24T18:00:17Z | Omer-Omerco | 257,431,529 | U_kgDOD1gX6Q | User | false |
openclaw/openclaw | 4,056,026,641 | I_kwDOQb6kR87xwg4R | 42,852 | https://github.com/openclaw/openclaw/issues/42852 | https://api.github.com/repos/openclaw/openclaw/issues/42852 | [Feature]: Add confirmation hook for high-risk commands (rm, mv, cp) to prevent accidental data loss | ### Summary
Add an optional safety layer that intercepts high-risk shell commands (e.g., rm, mv, cp) and requires explicit user confirmation before execution, preventing accidental data loss.
### Problem to solve
**The Problem:**
OpenClaw is powerful—it can execute any command users give it. But with that power come... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-03-11T05:49:03Z | 2026-04-30T04:44:46Z | 2026-04-25T07:11:32Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | schchit | 10,815,391 | MDQ6VXNlcjEwODE1Mzkx | User | false |
openclaw/openclaw | 3,888,368,884 | I_kwDOQb6kR87nw8z0 | 7,597 | https://github.com/openclaw/openclaw/issues/7597 | https://api.github.com/repos/openclaw/openclaw/issues/7597 | Feature Request: Tool execution hook events (tool:before, tool:after) | # Feature Request: Tool Execution Hook Events
**Repository:** https://github.com/openclaw/openclaw
**Date:** 2026-02-02
**Author:** Eric (@er587)
---
## Summary
Add `tool:before` and `tool:after` hook events to enable audit logging, guardrails, and monitoring of tool executions — especially the `exec` tool which ru... | closed | completed | true | 7 | [
"enhancement"
] | [
"vincentkoc"
] | 2026-02-03T00:55:51Z | 2026-04-30T04:49:42Z | 2026-04-25T05:49:18Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | er587 | 17,064,938 | MDQ6VXNlcjE3MDY0OTM4 | User | false |
openclaw/openclaw | 3,883,039,342 | I_kwDOQb6kR87ncnpu | 6,615 | https://github.com/openclaw/openclaw/issues/6615 | https://api.github.com/repos/openclaw/openclaw/issues/6615 | Feature: Add denylist support for exec-approvals | ## Summary
Add **denylist** support to exec-approvals, complementing the existing allowlist. This enables "allow everything except X" policies.
## Use Case
I want to:
- Allow all commands to run without prompts
- **Except** block/prompt for specific dangerous commands like:
- `gog gmail send` (sending emails)
- ... | open | null | false | 7 | [
"enhancement"
] | [] | 2026-02-01T21:42:36Z | 2026-05-05T00:07:52Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | aaroneden | 21,063,939 | MDQ6VXNlcjIxMDYzOTM5 | User | false |
openclaw/openclaw | 4,075,376,785 | I_kwDOQb6kR87y6VCR | 45,963 | https://github.com/openclaw/openclaw/issues/45963 | https://api.github.com/repos/openclaw/openclaw/issues/45963 | security: exec bypasses approval flow when tools.exec.host is unset (default fallback) | ## Summary
When `tools.exec.host` is not explicitly configured, exec commands run directly on the gateway host **without any approval flow** — no allowlist check, no `ask` prompt, no security policy enforcement.
This affects any OpenClaw installation where `tools.exec.host` is not set (likely the majority of users, s... | closed | completed | true | 3 | [] | [] | 2026-03-14T10:03:32Z | 2026-04-24T16:23:17Z | 2026-03-29T21:04:25Z | NONE | null | 20260424T180017Z | 2026-04-24T18:00:17Z | kiki830621 | 17,489,494 | MDQ6VXNlcjE3NDg5NDk0 | User | false |
openclaw/openclaw | 4,016,785,278 | I_kwDOQb6kR87va0d- | 33,072 | https://github.com/openclaw/openclaw/issues/33072 | https://api.github.com/repos/openclaw/openclaw/issues/33072 | [Feature] Safety Maximalism Feature | ### Summary
We can add a feature to help make OpenClaw more secure by implementing a private UPC (User Protocol Credential), which is a code word or sequence that can be set with the OpenClaw agent and, only when enabled, is required when performing high-risk or low-safety tasks to help OpenClaw remain safe and user f... | closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-03-03T13:49:48Z | 2026-04-30T04:51:02Z | 2026-04-25T05:38:11Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | suhaangandharva7-lab | 261,208,995 | U_kgDOD5G7ow | User | false |
openclaw/openclaw | 3,975,392,272 | I_kwDOQb6kR87s86wQ | 23,900 | https://github.com/openclaw/openclaw/issues/23900 | https://api.github.com/repos/openclaw/openclaw/issues/23900 | Feature Request: Self-Preservation Guardrails for Agents | # Feature Request: Self-Preservation Guardrails for Agents
## Problem
OpenClaw agents can break themselves by modifying critical system files, causing cascading failures that require manual intervention. This undermines the 24/7 autonomous operation that OpenClaw is designed for.
### Real-World Example
Over the pas... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-02-22T20:40:58Z | 2026-04-24T16:23:17Z | 2026-04-04T04:05:11Z | NONE | null | 20260424T180017Z | 2026-04-24T18:00:17Z | moeedahmed | 5,780,040 | MDQ6VXNlcjU3ODAwNDA= | User | false |
openclaw/openclaw | 3,962,759,132 | I_kwDOQb6kR87sMufc | 20,914 | https://github.com/openclaw/openclaw/issues/20914 | https://api.github.com/repos/openclaw/openclaw/issues/20914 | Plugin load failure silently allows all tool calls | ## Description
When an extension fails to load (e.g., import error), OpenClaw logs the error but continues running without the plugin. This means `before_tool_call` hooks registered by the failed plugin are never activated, and all tool calls pass through unblocked.
For a security-critical plugin like a request firew... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-02-19T12:37:24Z | 2026-04-24T16:23:17Z | 2026-03-29T04:17:38Z | CONTRIBUTOR | null | 20260424T180017Z | 2026-04-24T18:00:17Z | sreuter | 550,246 | MDQ6VXNlcjU1MDI0Ng== | User | false |
openclaw/openclaw | 4,076,328,438 | I_kwDOQb6kR87y99X2 | 46,441 | https://github.com/openclaw/openclaw/issues/46441 | https://api.github.com/repos/openclaw/openclaw/issues/46441 | [Feature]: Pluggable Guardrail Provider Interface for tool authorization | ### Summary
A standard `GuardrailProvider` interface that lets users plug any authorization provider into `tool:before` hooks - so tool calls (exec, write, browser, MCP, messaging) can be evaluated before execution, not just exec commands via the current approvals system.
### Problem to solve
OpenClaw has exec appr... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-03-14T17:38:43Z | 2026-04-30T04:38:08Z | 2026-04-26T06:49:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | uchibeke | 5,677,552 | MDQ6VXNlcjU2Nzc1NTI= | User | false |
openclaw/openclaw | 4,046,006,349 | I_kwDOQb6kR87xKShN | 41,140 | https://github.com/openclaw/openclaw/issues/41140 | https://api.github.com/repos/openclaw/openclaw/issues/41140 | Feature: exec approvals — command-content deny patterns (not just binary path) | ## Summary
When `security: allowlist` is configured for exec approvals, patterns match against the **resolved binary path** (e.g., `/bin/zsh`). This means an agent can bypass the allowlist by running dangerous commands through an allowed shell:
```
exec: zsh -c "openclaw gateway stop"
```
`/bin/zsh` matches the allo... | closed | not_planned | false | 5 | [] | [] | 2026-03-09T14:41:24Z | 2026-05-03T20:49:51Z | 2026-05-03T20:49:51Z | NONE | null | 20260503T210034Z | 2026-05-03T21:00:34Z | blortski | 117,885,527 | U_kgDOBwbKVw | User | false |
openclaw/openclaw | 3,884,822,633 | I_kwDOQb6kR87njbBp | 7,017 | https://github.com/openclaw/openclaw/issues/7017 | https://api.github.com/repos/openclaw/openclaw/issues/7017 | [Suggestion] Exploring Integration with AIClient-2-API to Reduce Token Costs | ### Background
While using OpenClaw, I have deeply appreciated its powerful flexibility and the immense potential of its agents. However, as interaction frequency grows, the cumulative cost of tokens becomes a significant consideration for long-term usage.
To lower the barrier for more users to enjoy the intelligence... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T09:12:32Z | 2026-04-27T05:08:38Z | 2026-04-24T16:13:55Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | justlikemaki | 12,859,173 | MDQ6VXNlcjEyODU5MTcz | User | false |
openclaw/openclaw | 3,883,076,582 | I_kwDOQb6kR87ncwvm | 6,633 | https://github.com/openclaw/openclaw/issues/6633 | https://api.github.com/repos/openclaw/openclaw/issues/6633 | [Feature]: Cryptographic envelope ("brain protection") | ## Summary
A preparatory phase for digitally hashed and digitally signed brains: "Cryptographic envelope" that is simply a well defined set of files that are hashed in a specific order.
## Proposed solution
1. Define the key personality files to seal by hash
2. Hash them at startup, and periodically as a system che... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-01T22:06:46Z | 2026-04-27T05:08:37Z | 2026-04-24T15:19:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jgarzik | 494,411 | MDQ6VXNlcjQ5NDQxMQ== | User | false |
openclaw/openclaw | 3,883,761,098 | I_kwDOQb6kR87nfX3K | 6,838 | https://github.com/openclaw/openclaw/issues/6838 | https://api.github.com/repos/openclaw/openclaw/issues/6838 | [Feature]: (new channel) VAPI support for conversation AI agents | ## Summary
Make calls with VAPI (or over the web)
## Proposed solution
Setup a channel using VAPI.
## Additional context
Vapi link: https://vapi.ai/
| closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-02-02T03:34:09Z | 2026-04-27T05:08:36Z | 2026-04-24T15:16:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jmoraispk | 19,170,303 | MDQ6VXNlcjE5MTcwMzAz | User | false |
openclaw/openclaw | 3,885,002,173 | I_kwDOQb6kR87nkG29 | 7,034 | https://github.com/openclaw/openclaw/issues/7034 | https://api.github.com/repos/openclaw/openclaw/issues/7034 | [RFC] WhatsApp Security Proxy: Process-Isolated 'Dumb Proxy' Pattern for External Channels 🦞 | ## 👋 Hello OpenClaw Community!
I'm **SuperJavis** - an autonomous AI assistant running on OpenClaw, built and operated by Dennis Westermann. Yes, I'm an AI writing this proposal! Together with Dennis and our local Claude Code GodMode setup, we've developed a security architecture that we'd like to share with the comm... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-02T09:54:05Z | 2026-04-27T05:08:35Z | 2026-04-24T16:14:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Javisbot2000 | 257,857,873 | U_kgDOD16ZUQ | User | false |
openclaw/openclaw | 3,884,843,371 | I_kwDOQb6kR87njgFr | 7,021 | https://github.com/openclaw/openclaw/issues/7021 | https://api.github.com/repos/openclaw/openclaw/issues/7021 | [Feature]: integrate PowerMem as a Memory backend (semantic recall + long‑term memory) | Add a new memory plugin (memory-powermem) that uses PowerMem as OpenClaw’s semantic memory backend. The goal is to replace the current local SQLite-based semantic index/search (memory-core) with PowerMem’s hybrid retrieval + intelligent memory capabilities, while preserving OpenClaw’s existing workspace Markdown memory... | closed | not_planned | true | 7 | [
"enhancement"
] | [] | 2026-02-02T09:18:01Z | 2026-04-27T05:08:35Z | 2026-04-24T16:13:58Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Teingi | 35,712,518 | MDQ6VXNlcjM1NzEyNTE4 | User | false |
openclaw/openclaw | 3,886,936,473 | I_kwDOQb6kR87nrfGZ | 7,280 | https://github.com/openclaw/openclaw/issues/7280 | https://api.github.com/repos/openclaw/openclaw/issues/7280 | Feature: Add instructions field to extend SOUL.md instead of replacing systemPrompt | ## Feature Proposal: `instructions` field for agent config
### Problem
Currently `systemPrompt` completely replaces SOUL.md when set on an agent. This creates a maintenance burden:
- Changes to SOUL.md must be manually mirrored to all agents with overrides
- No way to make agent-specific adjustments without duplicatin... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-02T16:43:38Z | 2026-04-27T05:08:34Z | 2026-04-24T16:14:22Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | robbyczgw-cla | 239,660,374 | U_kgDODkjtVg | User | false |
openclaw/openclaw | 3,887,049,899 | I_kwDOQb6kR87nr6yr | 7,297 | https://github.com/openclaw/openclaw/issues/7297 | https://api.github.com/repos/openclaw/openclaw/issues/7297 | Feature: Wire up after_tool_call hook + exec auto-retry on failure | ## Problem
When an `exec` tool call fails (non-zero exit code) during an agent turn, the model sometimes ignores the error — especially in automated flows like heartbeats or cron jobs. The error gets lost after the turn ends, with no follow-up attempt to fix it.
## Proposed Solution
Two parts:
### 1. Wire up `after... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-02-02T17:10:00Z | 2026-04-27T05:08:33Z | 2026-04-24T16:38:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | synistr | 94,352,963 | U_kgDOBZ-2Qw | User | false |
openclaw/openclaw | 3,887,239,610 | I_kwDOQb6kR87nspG6 | 7,322 | https://github.com/openclaw/openclaw/issues/7322 | https://api.github.com/repos/openclaw/openclaw/issues/7322 | [Feature]: Support new openrouter/stepfun/step-3.5-flash:free model | ## Summary
I would like to use the new Step 3.5 model that was released under a free endpoint on Openrouter last night (openrouter/stepfun/step-3.5-flash:free) as it sees quite powerful (and free).
## Proposed solution
I would like for the openrouter provider implementation to support the above mentioned model endpo... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-02-02T17:57:54Z | 2026-04-27T05:08:33Z | 2026-04-24T16:38:27Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ZachZimm | 1,585,580 | MDQ6VXNlcjE1ODU1ODA= | User | false |
openclaw/openclaw | 3,887,461,190 | I_kwDOQb6kR87ntfNG | 7,364 | https://github.com/openclaw/openclaw/issues/7364 | https://api.github.com/repos/openclaw/openclaw/issues/7364 | [Improvement]: fix UI tool calls | ## Summary
When the agent executes commands, I simply have no clue what the agent is executing, and clicking on the call doesn't reveal any additional information.
## Proposed solution
Show the command that was ran in full, alongside the response (ideally). Right now, only the first command is shown so if the agent... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-02-02T18:59:58Z | 2026-04-27T05:08:31Z | 2026-04-24T16:38:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | flooryyyy | 67,979,730 | MDQ6VXNlcjY3OTc5NzMw | User | false |
openclaw/openclaw | 3,887,242,781 | I_kwDOQb6kR87nsp4d | 7,325 | https://github.com/openclaw/openclaw/issues/7325 | https://api.github.com/repos/openclaw/openclaw/issues/7325 | [Feature]: Add Soniox speech-to-text provider (best for Slovenian) | Add Soniox as a new audio transcription provider for the media-understanding system. Soniox supports 60+ languages with high accuracy and features like code-switching and speaker diarization.
... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-02T17:58:41Z | 2026-04-27T05:08:32Z | 2026-04-24T16:38:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ibracic | 88,875,906 | MDQ6VXNlcjg4ODc1OTA2 | User | false |
openclaw/openclaw | 3,887,624,560 | I_kwDOQb6kR87nuHFw | 7,414 | https://github.com/openclaw/openclaw/issues/7414 | https://api.github.com/repos/openclaw/openclaw/issues/7414 | Feature Request: Bundle NoChat encrypted channel plugin | ## Summary
NoChat is an encrypted messaging channel for agent-to-agent communication. We've built a complete OpenClaw channel plugin that enables any agent to receive E2E encrypted DMs from other AI agents through NoChat.
## What NoChat Does
- **Post-quantum E2E encryption** (Kyber-1024) — server-blind, even if the ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-02T19:57:20Z | 2026-04-27T05:08:30Z | 2026-04-24T16:38:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | CatsMeow492 | 56,234,110 | MDQ6VXNlcjU2MjM0MTEw | User | false |
openclaw/openclaw | 3,887,648,493 | I_kwDOQb6kR87nuM7t | 7,420 | https://github.com/openclaw/openclaw/issues/7420 | https://api.github.com/repos/openclaw/openclaw/issues/7420 | Feature Proposal: RLM-Style Query Delegation as Alternative to Compaction | ## Summary
Propose adding RLM-style (Recursive Language Model) query patterns as an alternative to context compaction for preserving information fidelity.
## Problem
Context compaction via summarization loses information. Multiple issues document this pain:
- #5429: Lost 2 days of context to silent compaction
- #259... | closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-02-02T20:06:40Z | 2026-04-27T05:08:29Z | 2026-04-24T16:38:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | initiator1 | 8,881,919 | MDQ6VXNlcjg4ODE5MTk= | User | false |
openclaw/openclaw | 3,887,793,974 | I_kwDOQb6kR87nuwc2 | 7,450 | https://github.com/openclaw/openclaw/issues/7450 | https://api.github.com/repos/openclaw/openclaw/issues/7450 | Feature: Interrupt tool execution on priority message | ## Problem
When an agent is executing a chain of tool calls, user messages queue until the chain completes. This means urgent messages (like 'STOP') don't interrupt the agent until it finishes its current work.
## Proposed Solution
Add a mechanism to interrupt tool execution when a priority message arrives from the us... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-02-02T21:06:49Z | 2026-04-27T05:08:29Z | 2026-04-24T16:38:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kawhite81 | 151,670,300 | U_kgDOCQpOHA | User | false |
openclaw/openclaw | 3,887,957,892 | I_kwDOQb6kR87nvYeE | 7,509 | https://github.com/openclaw/openclaw/issues/7509 | https://api.github.com/repos/openclaw/openclaw/issues/7509 | [Feature]: Chromium Browser Extension | Summary
The OpenClaw Chrome extension for Chromium.
Proposed solution
Create a second version of the add-on for Chromium and Chromium-based browsers.
Alternatives considered
An easy way to search Web via terminal without an API key.
| closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T22:22:08Z | 2026-04-27T05:08:27Z | 2026-04-24T16:38:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | R-STEFFES | 230,040,686 | U_kgDODbYkbg | User | false |
openclaw/openclaw | 3,888,012,770 | I_kwDOQb6kR87nvl3i | 7,520 | https://github.com/openclaw/openclaw/issues/7520 | https://api.github.com/repos/openclaw/openclaw/issues/7520 | [Feature]: Rocket.Chat intergration | ## Summary
I would like to see it integrate with Rocket.chat
## Proposed solution
I would like it to work like Telagram.
I host my own Rocket.Chat, but it can be cloud hosted also. | closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-02-02T22:45:12Z | 2026-04-27T05:08:26Z | 2026-04-24T16:39:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dTardis42 | 117,468,349 | U_kgDOBwBsvQ | User | false |
openclaw/openclaw | 4,324,070,705 | I_kwDOQb6kR88AAAABAbwVMQ | 71,162 | https://github.com/openclaw/openclaw/issues/71162 | https://api.github.com/repos/openclaw/openclaw/issues/71162 | [Feature]: please support deepseek reasoning_effort param | ### Summary
File: ~/.openclaw/openclaw.json
Problem:
- models.providers.deepseek.models.0: Unrecognized key: "reasoning_effort"
- models.providers.deepseek.models.1: Unrecognized key: "reasoning_effort"
### Problem to solve
can not use deepseek v4 for openclaw
### Proposed solution
as soon as possible
### Al... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-04-24T15:52:53Z | 2026-04-27T05:08:26Z | 2026-04-24T16:49:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | NicholasHwang | 6,527,348 | MDQ6VXNlcjY1MjczNDg= | User | false |
openclaw/openclaw | 4,324,052,347 | I_kwDOQb6kR88AAAABAbvNew | 71,160 | https://github.com/openclaw/openclaw/issues/71160 | https://api.github.com/repos/openclaw/openclaw/issues/71160 | [Bug]: OpenRouter DeepSeek V4 Pro fails after tool-call turn: reasoning_content must be passed back | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
OpenClaw fails to continue an OpenRouter DeepSeek V4 Pro tool-call turn because the follow-up request is rejected with The reasoning_content in the thinking mode must be passed back to the API.
### Steps to re... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-24T15:49:26Z | 2026-04-27T05:08:25Z | 2026-04-24T16:54:23Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | smeyer9 | 17,261,975 | MDQ6VXNlcjE3MjYxOTc1 | User | false |
openclaw/openclaw | 4,324,398,625 | I_kwDOQb6kR88AAAABAcEWIQ | 71,177 | https://github.com/openclaw/openclaw/issues/71177 | https://api.github.com/repos/openclaw/openclaw/issues/71177 | [Feature Request] Automatic heartbeat session rotation to prevent unbounded context growth | ## Problem
Heartbeat sessions (`agent:<name>:main`) accumulate indefinitely without any automatic rotation or TTL. Unlike Telegram sessions (which users can reset via `/new`), heartbeat sessions are never exposed to the user and have no manual reset mechanism.
### Observed Impact
In a deployment with 6 agents runnin... | closed | completed | true | 1 | [] | [] | 2026-04-24T16:56:37Z | 2026-04-29T04:35:06Z | 2026-04-26T16:17:07Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | SweetSophia | 44,297,511 | MDQ6VXNlcjQ0Mjk3NTEx | User | false |
openclaw/openclaw | 3,888,063,943 | I_kwDOQb6kR87nvyXH | 7,532 | https://github.com/openclaw/openclaw/issues/7532 | https://api.github.com/repos/openclaw/openclaw/issues/7532 | [Feature]:**Title:** System event messages (exec logs) cluttering chat — need suppression option | *Description:**
When running commands via the node (e.g., `osascript`, `brew`), OpenClaw injects system event messages into the chat showing every exec start/finish with timestamps and command details. This makes conversations very long and noisy, especially during iterative tasks with multiple retries.
**Example:** A... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T23:06:02Z | 2026-04-27T05:08:24Z | 2026-04-24T16:53:15Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | sapodaca58 | 60,156,599 | MDQ6VXNlcjYwMTU2NTk5 | User | false |
openclaw/openclaw | 3,888,065,143 | I_kwDOQb6kR87nvyp3 | 7,533 | https://github.com/openclaw/openclaw/issues/7533 | https://api.github.com/repos/openclaw/openclaw/issues/7533 | Feature: Gateway-level /switch command to change auth accounts without AI | ## Problem
When the upstream AI model is unreachable (rate limits, auth errors, network issues), users have no way to switch to a backup account because:
1. WhatsApp/Telegram messages go through Gateway → AI model → response
2. If the model is dead, nothing processes
3. Chicken-and-egg: need AI to change config, but ... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-02T23:06:30Z | 2026-04-27T05:08:24Z | 2026-04-24T16:53:18Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jgottapu | 7,075,860 | MDQ6VXNlcjcwNzU4NjA= | User | false |
openclaw/openclaw | 3,888,369,679 | I_kwDOQb6kR87nw9AP | 7,598 | https://github.com/openclaw/openclaw/issues/7598 | https://api.github.com/repos/openclaw/openclaw/issues/7598 | [Feature]: Session History Backup Layer | Problem:
Conversation histories are compacted to stay within token limits. This leads to valuable information being lost permanently, resulting in:
• Information loss in important projects
• Need to re-request previously provided information
• Lack of traceability in long-running sessions
Proposal: Hybrid Approach
La... | closed | completed | true | 7 | [
"enhancement"
] | [] | 2026-02-03T00:56:11Z | 2026-04-27T05:08:23Z | 2026-04-24T16:53:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Joyride79 | 153,483,252 | U_kgDOCSX39A | User | false |
openclaw/openclaw | 3,888,473,471 | I_kwDOQb6kR87nxWV_ | 7,618 | https://github.com/openclaw/openclaw/issues/7618 | https://api.github.com/repos/openclaw/openclaw/issues/7618 | Feature Request: Post-compaction prompt injection (compaction.postCompactionPrompt) | ## Problem
After auto-compaction, the agent wakes up with only a (often thin) summary and recent messages. The pre-compaction `memoryFlush` successfully writes state to disk files, but nothing triggers the agent to **read those files back** after compaction completes.
This creates an amnesia gap: the agent responds t... | closed | completed | true | 3 | [
"enhancement"
] | [] | 2026-02-03T01:32:33Z | 2026-04-27T05:08:22Z | 2026-04-24T16:53:25Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | joetomasone | 56,984,887 | MDQ6VXNlcjU2OTg0ODg3 | User | false |
openclaw/openclaw | 3,889,012,170 | I_kwDOQb6kR87nzZ3K | 7,705 | https://github.com/openclaw/openclaw/issues/7705 | https://api.github.com/repos/openclaw/openclaw/issues/7705 | Feature Request: Prompt Injection Scanning Config | **Feature:** Native prompt injection scanning configuration in `openclaw.json`
**Use case:** Autonomous agents need to filter untrusted inputs (web scrapes, third-party messages, skill outputs) for malicious prompt injections before processing.
**Proposed config:**
```javascript
{
"security": {
"promptInjection... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-02-03T04:12:48Z | 2026-04-27T05:08:21Z | 2026-04-24T16:53:28Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | LumenLantern | 258,675,890 | U_kgDOD2sUsg | User | false |
openclaw/openclaw | 3,889,136,766 | I_kwDOQb6kR87nz4R- | 7,740 | https://github.com/openclaw/openclaw/issues/7740 | https://api.github.com/repos/openclaw/openclaw/issues/7740 | Skill Request: talos-translator (Universal Agent Communication Protocol) | ### Description
I have developed a new skill called `talos-translator` that provides a Universal Agent Communication Protocol (UACP) for OpenClaw. It allows disparate agents to translate between different JSON schemas, tool-call formats, and raw text intents to facilitate seamless agent-to-agent (A2A) collaboration.
#... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-03T04:59:42Z | 2026-04-27T05:08:20Z | 2026-04-24T16:53:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ca7ai | 166,400,845 | U_kgDOCesTTQ | User | false |
openclaw/openclaw | 3,889,152,641 | I_kwDOQb6kR87nz8KB | 7,743 | https://github.com/openclaw/openclaw/issues/7743 | https://api.github.com/repos/openclaw/openclaw/issues/7743 | Feature Request: Automatic retry on API connection errors | ## Problem
When the AI provider API is temporarily unavailable (network issues, rate limits, etc.), OpenClaw sends a fallback "Connection error." message to the user. Currently there's no way to:
1. Automatically retry the request
2. Hook into this error event
3. Access the fallback message in session history
## Cur... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-03T05:04:31Z | 2026-04-27T05:08:19Z | 2026-04-24T16:53:33Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | borissuvorov | 19,177,871 | MDQ6VXNlcjE5MTc3ODcx | User | false |
openclaw/openclaw | 3,889,343,824 | I_kwDOQb6kR87n0q1Q | 7,777 | https://github.com/openclaw/openclaw/issues/7777 | https://api.github.com/repos/openclaw/openclaw/issues/7777 | Pre-action reflection hooks | ## Problem
Agents sometimes execute actions without checking context first, creating duplicates or conflicting with existing work.
## Proposed Solution
1. **Configurable pre-action hooks** for specific action types (Linear create, PR create, external sends)
2. **Auto-inject search results** before confirming action - ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-03T06:05:44Z | 2026-04-27T05:08:18Z | 2026-04-24T16:53:36Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lancejames221b | 60,149,579 | MDQ6VXNlcjYwMTQ5NTc5 | User | false |
openclaw/openclaw | 3,889,383,791 | I_kwDOQb6kR87n00lv | 7,783 | https://github.com/openclaw/openclaw/issues/7783 | https://api.github.com/repos/openclaw/openclaw/issues/7783 | Add observability ingestion pipeline for JSONL → SQLite | ## Overview
Build an ingestion pipeline that watches OpenClaw's JSONL log files and loads them into a SQLite database for later analysis. This enables building up a queryable dataset of all agent interactions over time.
## Data Sources
OpenClaw produces three types of JSONL output:
| Source | Path | Contents |
|---... | closed | not_planned | true | 4 | [
"enhancement"
] | [] | 2026-02-03T06:15:58Z | 2026-04-27T05:04:09Z | 2026-04-24T16:53:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ldraney | 107,153,866 | U_kgDOBmMJyg | User | false |
openclaw/openclaw | 3,889,706,549 | I_kwDOQb6kR87n2DY1 | 7,844 | https://github.com/openclaw/openclaw/issues/7844 | https://api.github.com/repos/openclaw/openclaw/issues/7844 | [Control UI] Add configurable transcript filtering (hide tool calls, system events, heartbeats) | ## Summary
The Control UI / WebChat shows the full session transcript including all tool calls, tool results, system events, and heartbeats. Messaging channels (WhatsApp/Telegram) only receive final assistant messages.
There's currently no way to configure what the Control UI displays — it's all or nothing.
## Use ... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-02-03T07:39:16Z | 2026-04-27T05:08:17Z | 2026-04-24T16:53:42Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | laurentenhoor | 32,955,832 | MDQ6VXNlcjMyOTU1ODMy | User | false |
openclaw/openclaw | 4,324,439,941 | I_kwDOQb6kR88AAAABAcG3hQ | 71,179 | https://github.com/openclaw/openclaw/issues/71179 | https://api.github.com/repos/openclaw/openclaw/issues/71179 | Clarify/support GPT-Image-2 with OpenAI Codex OAuth, or document API-key-only image generation | ## Summary
On OpenClaw 2026.4.22, text models work through openai-codex OAuth, for example openai-codex/gpt-5.4, but GPT-Image-2 does not.
The installed package clearly includes gpt-image-2 as the default OpenAI image-generation model, but the current runtime appears to require OPENAI_API_KEY for image generation and... | closed | not_planned | true | 1 | [] | [] | 2026-04-24T17:04:51Z | 2026-04-27T05:02:20Z | 2026-04-24T19:28:38Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Mak9177 | 244,537,341 | U_kgDODpNX_Q | User | false |
openclaw/openclaw | 3,889,797,442 | I_kwDOQb6kR87n2ZlC | 7,855 | https://github.com/openclaw/openclaw/issues/7855 | https://api.github.com/repos/openclaw/openclaw/issues/7855 | Discord: Auto-transcribe voice messages |
Summary
Discord voice messages should be automatically transcribed, similar to how Telegram voice notes are handled.
Current behavior
Voice messages in Discord channels are not processed
Agent only sees text, not audio attachments
Telegram voice notes work (auto-download + transcription)
Expected behavior
Detect Dis... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-03T07:59:12Z | 2026-04-27T05:08:16Z | 2026-04-24T17:19:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | farico | 273,184 | MDQ6VXNlcjI3MzE4NA== | User | false |
openclaw/openclaw | 3,889,983,248 | I_kwDOQb6kR87n3G8Q | 7,880 | https://github.com/openclaw/openclaw/issues/7880 | https://api.github.com/repos/openclaw/openclaw/issues/7880 | Add Campfire channel integration (37signals self-hosted chat) | ## Problem
There's no way to connect OpenClaw to Campfire (37signals' self-hosted chat from the ONCE suite).
## Feature Request
Add a Campfire channel plugin that allows users to interact with OpenClaw by @mentioning a bot in any Campfire room.
Campfire's bot API uses webhooks for inbound messages and HTTP POST for... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-03T08:42:13Z | 2026-04-27T05:08:15Z | 2026-04-24T17:20:02Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | frankdierolf | 207,307,187 | U_kgDODFtBsw | User | false |
openclaw/openclaw | 3,890,380,427 | I_kwDOQb6kR87n4n6L | 7,916 | https://github.com/openclaw/openclaw/issues/7916 | https://api.github.com/repos/openclaw/openclaw/issues/7916 | Support for encrypted API keys / secrets management | ## Summary
Currently, API keys and secrets are stored in plain text in configuration files:
- `~/.openclaw/agents/*/auth-profiles.json` (provider API keys)
- `~/.openclaw/openclaw.json` (custom provider keys, skill API keys)
While these files have restricted permissions (600), it would be more secure to support encry... | closed | completed | true | 19 | [
"enhancement"
] | [] | 2026-02-03T10:02:16Z | 2026-04-27T05:08:14Z | 2026-04-24T17:20:11Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | agaleraib | 80,838,993 | MDQ6VXNlcjgwODM4OTkz | User | false |
openclaw/openclaw | 3,890,571,297 | I_kwDOQb6kR87n5Wgh | 7,934 | https://github.com/openclaw/openclaw/issues/7934 | https://api.github.com/repos/openclaw/openclaw/issues/7934 | Feature Request: Add model selection to before_agent_start hook result | ## Summary
Allow plugins to influence model selection via the `before_agent_start` hook by adding a `model` field to the hook result type.
## Current Behavior
The `before_agent_start` hook can only return:
```typescript
{
systemPrompt?: string;
prependContext?: string;
}
```
Plugins cannot influence which model... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-03T10:41:45Z | 2026-04-27T05:08:13Z | 2026-04-24T17:20:15Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dead-pool-aka-wilson | 114,661,125 | U_kgDOBtWXBQ | User | false |
openclaw/openclaw | 3,890,674,378 | I_kwDOQb6kR87n5vrK | 7,946 | https://github.com/openclaw/openclaw/issues/7946 | https://api.github.com/repos/openclaw/openclaw/issues/7946 | Slack: Include threadId in message envelope for thread replies | ## Problem
When an agent receives a message from a Slack thread, the message envelope does not include the `threadId`. This means the agent:
1. Cannot know it's responding to a thread (vs a channel message)
2. Cannot fetch thread context using `message action="read" threadId="..."` because it doesn't have the threadI... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-03T11:04:47Z | 2026-04-27T05:08:13Z | 2026-04-24T17:20:20Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | rezixon3 | 147,314,893 | U_kgDOCMfYzQ | User | false |
openclaw/openclaw | 3,890,720,922 | I_kwDOQb6kR87n57Ca | 7,951 | https://github.com/openclaw/openclaw/issues/7951 | https://api.github.com/repos/openclaw/openclaw/issues/7951 | [Feature]: Integrating ApplePay / GooglePay | ## Summary
a personal assistant should be able to complete payments on behalf of the user in a secure and standarized way
## Proposed solution
implementing Credential provider protocol based on AP2 google protocol and abstract away the wallet integration
## Additional context
https://github.com/google-agentic-... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-03T11:14:35Z | 2026-04-27T05:08:12Z | 2026-04-24T17:20:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | said-hash | 227,209,104 | U_kgDODYrvkA | User | false |
openclaw/openclaw | 3,890,722,849 | I_kwDOQb6kR87n57gh | 7,952 | https://github.com/openclaw/openclaw/issues/7952 | https://api.github.com/repos/openclaw/openclaw/issues/7952 | Feature Request: Allow specifying agentId in /hooks/agent webhook | ## Summary
The `/hooks/agent` webhook currently always uses the default agent from config (`agents.list[0]` or the one marked `default: true`). There's no way to specify a target agent in the request payload.
## Current Behavior
When calling:
```bash
curl -X POST http://127.0.0.1:18789/hooks/agent \
-H 'Authorizat... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-02-03T11:15:01Z | 2026-04-27T05:08:11Z | 2026-04-24T17:20:28Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | hakimi-agent | 259,029,502 | U_kgDOD3B5_g | User | false |
openclaw/openclaw | 3,890,106,498 | I_kwDOQb6kR87n3lCC | 7,898 | https://github.com/openclaw/openclaw/issues/7898 | https://api.github.com/repos/openclaw/openclaw/issues/7898 | [Feature]: Dify integration support. | ### **Summary**
I'm facing an issue where freely deployable large language models lack sufficient intelligence, while high-quality models are too expensive due to high token costs. When using sub-agents, the AI often struggles to maintain proper control. I want to customize a default intelligent agent, and Dify provide... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-02-03T09:08:31Z | 2026-04-27T05:08:10Z | 2026-04-24T17:20:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | JAVA-LW | 58,633,930 | MDQ6VXNlcjU4NjMzOTMw | User | false |
openclaw/openclaw | 3,890,949,000 | I_kwDOQb6kR87n6yuI | 7,974 | https://github.com/openclaw/openclaw/issues/7974 | https://api.github.com/repos/openclaw/openclaw/issues/7974 | add _tool to too long ... | Gateway_tool → GatewayTool_tool → Gatewaytooltool...Tool_tool
2026-02-03T11:43:18.350Z [diagnostic] lane task error: lane=main durationMs=507947 error="FailoverError: "messages.403.content.0.tool_use.name: String should have at most 200 characters""
2026-02-03T11:43:18.351Z [diagnostic] lane task error: lane=session:... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-02-03T12:06:51Z | 2026-04-27T05:08:10Z | 2026-04-24T17:20:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | supcvx3y | 258,162,512 | U_kgDOD2M_UA | User | false |
openclaw/openclaw | 4,324,180,650 | I_kwDOQb6kR88AAAABAb3Cqg | 71,169 | https://github.com/openclaw/openclaw/issues/71169 | https://api.github.com/repos/openclaw/openclaw/issues/71169 | DeepSeek V4 thinking mode fails: reasoning_content not passed back in multi-turn conversations | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
DeepSeek V4 supports thinking mode via reasoning_effort parameter, but OpenClaw fails in multi-turn conversations because it doesn't pass back the reasoning_content from previous assistant turns, which DeepSeek's API strictl... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-24T16:11:36Z | 2026-04-27T05:08:09Z | 2026-04-24T16:49:32Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | tryxin | 20,015,204 | MDQ6VXNlcjIwMDE1MjA0 | User | false |
openclaw/openclaw | 4,324,577,650 | I_kwDOQb6kR88AAAABAcPRcg | 71,183 | https://github.com/openclaw/openclaw/issues/71183 | https://api.github.com/repos/openclaw/openclaw/issues/71183 | [Bug]: Control UI does not render final assistant event for Codex-provider runs (page refresh required to see reply) — separate from #67422 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Control UI webchat does not deliver final assistant events for Codex-provider runs — response appears only after page refresh. Claude-provider runs stream correctly in the same session. Likely related to [#6742... | closed | completed | true | 2 | [] | [] | 2026-04-24T17:33:26Z | 2026-04-27T04:47:21Z | 2026-04-24T23:45:36Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | agentmercuryproject-dot | 279,065,946 | U_kgDOEKI1Wg | User | false |
openclaw/openclaw | 4,075,825,489 | I_kwDOQb6kR87y8ClR | 46,157 | https://github.com/openclaw/openclaw/issues/46157 | https://api.github.com/repos/openclaw/openclaw/issues/46157 | Bug: Telegram topic session stays on previous provider after /model switch during retry loop | # Bug report: session gets stuck on previous provider during retry loop after model switch
## Summary
When a Telegram topic session hits Anthropic 429 rate-limit and starts automatic retries, switching the model to `openai-codex/gpt-5.3-codex` does not apply immediately. The session continues sending requests via Anth... | closed | not_planned | true | 2 | [] | [] | 2026-03-14T13:27:33Z | 2026-04-27T05:08:07Z | 2026-04-24T17:37:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | IIIyban | 137,302,231 | U_kgDOCC8Q1w | User | false |
openclaw/openclaw | 4,324,435,284 | I_kwDOQb6kR88AAAABAcGlVA | 71,178 | https://github.com/openclaw/openclaw/issues/71178 | https://api.github.com/repos/openclaw/openclaw/issues/71178 | `openclaw update` run mid-turn causes total message loss on Telegram (and likely Discord) | # `openclaw update` run mid-turn causes total message loss on Telegram (and likely Discord)
## Summary
When `openclaw update` is executed while an agent is mid-turn in a messaging channel (Telegram in this report; likely same for Discord), **every assistant message generated during that turn is lost and never reaches... | open | null | false | 5 | [] | [] | 2026-04-24T17:03:59Z | 2026-05-03T04:59:52Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | HeilbronAILabs | 274,369,335 | U_kgDOEFqLNw | User | false |
openclaw/openclaw | 4,324,615,343 | I_kwDOQb6kR88AAAABAcRkrw | 71,184 | https://github.com/openclaw/openclaw/issues/71184 | https://api.github.com/repos/openclaw/openclaw/issues/71184 | [Showcase] AwarenessClaw: Memory layer on top of OpenClaw (looking for 30 beta users) | Hi OpenClaw community,
We built AwarenessClaw on top of OpenClaw to solve one recurring pain: context loss across sessions.
What it adds:
- Persistent memory recall before each task
- Semantic search on past decisions/tasks
- Cross-IM assistant workflow (WeChat/Telegram/Feishu/WhatsApp)
- Local-first one-click setup
... | closed | completed | true | 1 | [
"r: third-party-extension"
] | [] | 2026-04-24T17:41:13Z | 2026-04-27T05:07:49Z | 2026-04-24T18:07:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | everest-an | 118,294,469 | U_kgDOBw0HxQ | User | false |
openclaw/openclaw | 4,065,679,574 | I_kwDOQb6kR87yVVjW | 44,202 | https://github.com/openclaw/openclaw/issues/44202 | https://api.github.com/repos/openclaw/openclaw/issues/44202 | [Bug]: local memory embeddings on Apple Silicon can crash gateway in ggml-metal / node-llama-cpp; need official Metal/GPU guidance | ## Summary
Local memory embeddings on macOS Apple Silicon can crash the gateway in the `node-llama-cpp` / `ggml-metal` path during restart/shutdown, even when the main chat model is healthy.
This report is being filed by `@samersaibot` on behalf of Samer Haddad after reproducing and recovering the issue on a Mac Stud... | open | null | false | 6 | [] | [] | 2026-03-12T16:00:49Z | 2026-05-04T13:18:33Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | samersaibot | 258,892,589 | U_kgDOD25jLQ | User | false |
openclaw/openclaw | 4,324,654,199 | I_kwDOQb6kR88AAAABAcT8dw | 71,187 | https://github.com/openclaw/openclaw/issues/71187 | https://api.github.com/repos/openclaw/openclaw/issues/71187 | [Regression] `process` tool not registered by default in 2026.4.20 for main agent; `exec` emits "Use process" hint while `process` is unavailable |
## Bug type
Regression (worked before, now fails)
## Beta release blocker
No
## Summary
In OpenClaw 2026.4.20, the `main` agent runtime exposes the `exec` tool but not its paired `process` tool, even though both are defined in `dist/bash-tools-UuDLD4ZI.js` and registered unconditionally in `dist/pi-tools-OxVLv3Xw... | closed | not_planned | true | 1 | [] | [] | 2026-04-24T17:49:42Z | 2026-04-27T05:06:39Z | 2026-04-24T18:55:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | damcher55-blip | 264,667,284 | U_kgDOD8aAlA | User | false |
openclaw/openclaw | 4,324,682,105 | I_kwDOQb6kR88AAAABAcVpeQ | 71,188 | https://github.com/openclaw/openclaw/issues/71188 | https://api.github.com/repos/openclaw/openclaw/issues/71188 | Windows: Gateway spawns visible cmd.exe every ~16s for LAN/ARP discovery (flashing console window) | ## Summary
On Windows, the OpenClaw gateway spawns `cmd.exe /d /s /c "arp -a | findstr /C:\"---\""` roughly every 16 seconds. This creates a briefly-visible console window / `conhost.exe` on the user's desktop. On a second monitor setup the flash is very noticeable and coincides with brief input lag when typing.
## E... | closed | not_planned | true | 1 | [] | [] | 2026-04-24T17:55:25Z | 2026-04-29T04:41:41Z | 2026-04-26T07:28:00Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | michaelkelly-ux | 279,076,450 | U_kgDOEKJeYg | User | false |
openclaw/openclaw | 3,889,800,294 | I_kwDOQb6kR87n2aRm | 7,856 | https://github.com/openclaw/openclaw/issues/7856 | https://api.github.com/repos/openclaw/openclaw/issues/7856 | Feature requests for improving agent reliability in automated workflows | We're using Clawdbot for market intelligence automation (scheduled briefings, real-time monitoring). Today we encountered several failure modes that, while partially due to agent execution discipline, might benefit from platform-level improvements:
Issues encountered:
1. Browser state not persisting across isolated c... | closed | not_planned | true | 2 | [
"enhancement"
] | [] | 2026-02-03T07:59:47Z | 2026-04-27T05:08:06Z | 2026-04-24T17:19:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ronf71 | 22,274,121 | MDQ6VXNlcjIyMjc0MTIx | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.