OpenClaw Newsletter - 2026-06-15
OpenClaw Newsletter
Monday, June 15, 2026Top Stories
- iMessage: inbound message silently dropped when sent immediately after agent outbound reply (same group chat) — Environment - OpenClaw: 2026.5.22 (a374c3a) - imsg: 0.9.0 - macOS: Tahoe (26), M4 MacBook Air - iMessage plugin config: debounceMs: 3000 Problem User messages sent to a group chat immediately after...
- fix(compaction): forward abortSignal through delegateCompactionToRuntime bridge — Summary When a user presses the stop button during auto-compaction, the abort signal cancels the wait but the compaction LLM call continues in-flight and writes its result to the session transcript....
- fix(agents): guard tool definition schema reads — Summary - guard agent tool definition metadata reads so a hostile or broken tool schema getter cannot crash definition adaptation - represent unreadable parameter schemas as an invalid schema...
- fix(qqbot): keep markdown table chunks valid — Summary - Add a QQ Bot markdown chunker that keeps table chunks self-contained across message boundaries. - Preserve table state across block deliveries and flush unfinished table-row fragments as...
- fix(telegram): rotate in-flight answer preamble at tool boundary — Summary Fixes a Telegram bug where an in-flight answer preamble is overwritten instead of preserved when a tool runs mid-stream and post-tool text resumes. When the model streams a short preamble...
- Feature Request: Display inbound media attachments as image previews in WebUI — Summary When an image is sent to OpenClaw via a channel (e.g. Telegram or iMessage), the image is correctly saved to media://inbound/ and processed by the AI model. However, the WebUI chat only...
- feat(auto-reply,codex): lightweight reply lane and Codex watchdog — Summary - add a conservative lightweight reply lane for obvious low-risk text-only chat replies - keep uncertain, tool/action, media/link, high-stakes, code/config, and unresolved reply-target...
- fix(agents): forward turn-source for embedded plugin approvals (supersedes #84205) — Summary Forwards turn-source routing fields through the embedded plugin before_tool_call approval chain so plugin approvals can be delivered back to the originating chat (Telegram, Discord, Slack,...
- State directory permissions can be relaxed back to 0775 by update-check JSON writes — Hello, I noticed a recurring security audit warning where the OpenClaw state directory is reported as group-writable again after it was manually hardened to 0700. This does not seem to be a one-off...
- feat(memory-core): optional cross-encoder rerank stage for memory search — Summary OpenClaw's builtin (sqlite-vec) memory search does first-stage hybrid retrieval (dense vec0 ∥ FTS5 BM25, fused) → temporal decay → MMR diversity, but has no second-stage relevance reranker....
- fix(llm-core): wrap unreadable tool schemas — Summary - wrap unreadable tool parameter schemas in controlled validation errors - keep ordinary argument validation failures on the existing Validation failed path - add regression coverage for a...
- fix(doctor): skip top-level allowlist scan when active per-account scopes exist — Summary The empty-allowlist doctor scanner treated the top-level channel config as its own standalone account, producing false 'empty allowlist' warnings when every active account under that channel...
- docs(logging): document exact model.usage event field names in diagnostic catalog (fixes #49046) — What Expand the model.usage entry in the OpenTelemetry diagnostic event catalog with the exact JSON field names emitted by the agent runner. Add a cross-reference from the logging docs to the...
- fix(agents): harden tool search schema catalog — Summary - make Tool Search catalog fingerprinting tolerate unreadable schema objects - snapshot catalog parameters into JSON-safe metadata so describe/search cannot be poisoned by hostile schema...
- fix: prevent memory flush failure from aborting user reply (#85645) — Summary When memoryFlush fails during the reply pipeline (provider timeout, rate limit, network error), the visible error payload was treated as the final reply via replyOperation.fail("run_failed",...
- perf: structuredClone instead of JSON.parse(stringify) — Summary This PR implements the highest-impact, lowest-effort performance improvement identified in a comprehensive codebase analysis. Replaces JSON.parse(JSON.stringify()) with native...
- fix(browser): bring tab to foreground before screenshot capture — Summary What problem does this PR solve? browser screenshot (and the screenshot-backed snapshot path) hang until the CDP socket times out (~20s) whenever the target tab is backgrounded. Why does this...
- Telegram poll_answer updates are not routed to agent turns — Summary Telegram polls can be created from OpenClaw, but non-anonymous poll votes are not routed to the agent as actionable turns/events. This makes poll-based approvals look usable in Telegram while...
- fix(diagnostics): prune non-idle ghost entries after failed stuck-session recovery — Summary When stuck-session recovery fails (status: "failed"), recoveryOutcomeMutatesSessionState() returns false → the diagnostic entry stays in "processing" / "waiting" state permanently. The...
- fix(cron): resolve lastRunStatus in cron list/show human output — Summary openclaw cron list and openclaw cron show print idle for a job whose real status is ok / error / skipped whenever the job's state carries only lastRunStatus (the primary field) and not the...
- fix(agents): honor disabled envelope timestamps at model boundary — Opened on behalf of Onur Solmaz (osolmaz). AI-assisted contribution. Summary OpenClaw could still send timestamp bytes to model input even when agents.defaults.envelopeTimestamp was "off". That...
- fix(provider): harden provider tool schema hooks — Summary - Harden provider tool schema normalization against plugin-supplied tools with unreadable parameters fields. - Share guarded tool-field reads across Gemini, DeepSeek, and native OpenAI schema...
- Feature request: Route Telegram group topics to different agents — Problem Currently, OpenClaw routes Telegram messages by channel + accountId (bot). When a bot joins a group with multiple topics (forum topics), all messages from all topics go to the same agent....
- #92664: [Bug]: read tool fails to read GBK-encoded text files on Chinese Windows (displays garbled text) — Summary - Fix garbled text when reading non-UTF-8 files (GBK on Chinese Windows, Shift_JIS on Japanese Windows) by reusing the existing Windows active- codepage decoder (decodeWindowsOutputBuffer) in...
- fix(agents): harden system prompt tool reporting — Summary - Harden system prompt report generation against tool metadata getters that throw while reading name, description, label, or parameters. - Keep schema size/property telemetry best-effort when...
- Voice-call plugin: hangup aborts in-flight tool calls (sessions_spawn killed before child initializes) — Problem When a voice call ends, the gateway aborts the embedded agent run via lane management for the voice lane. All in-flight tool calls — including sessions_spawn — receive the abort signal...
- fix(plugins): harden plugin tool shape reads — Summary - Harden plugin tool runtime shape checks against hostile getters on tool names, descriptions, execute functions, and parameter schemas. - Keep malformed plugin tools isolated so a bad tool...
- fix(status): avoid stale session context windows — Summary Fix /status context-window reporting when persisted session runtime metadata is stale. The status message can display the currently selected/configured model while still using an old...
- docs: document embedded compaction context contracts — Summary - Document the embedded compaction runtime context contract. - Clarify how compaction target resolution applies model overrides, auth-profile safety, and Codex-runtime OpenAI routing. - Keep...
- [Bug]: session-memory hook model config ignored for LLM slug generation — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary The model field in hooks.internal.entries.session-memory is not used when generating LLM-based slugs....
- OpenClaw repo stats for 2026-06-15 — 378788 stars, 79241 forks, 2570 contributors
- OpenClaw sponsors (110 total) — 110 sponsors supporting OpenClaw
- fix: allow guarded transfer of deleted skills (#2647)
- feat: add plugin versions tab (#2646) — feat: add package versions API wrapper fix: omit empty package version cursor feat: add plugin versions tab fix: paginate plugin version history fix: remove unsafe plugin version downloads fix:...
- revert: retire API key required skill feature (#2644)
- fix: serialize registry backup index writes
- fix: normalize publisher recovery handle comparisons (#2643) — Normalize mixed-case stored user handles in publisher recovery guards and retire the previous mixed-case handle before handoff.
- fix: serialize registry backup retry drains (#2641)
- fix: add guarded personal publisher recovery (#2642) — Summary: - The PR adds an admin-only personal publisher recovery flow with HTTP API, admin CLI support, shared response schema, docs/spec notes, and tests. - Reproducibility: yes. Source inspection...
- fix: compact theme controls in profile menu (#2640)
- fix: skip indexed registry backup retries (#2639)
- test: cover skill summary digest sync (#2638)
- fix: preserve legacy package ZIP hashes (#2636) — test: fix indexed row helper lint fix: preserve legacy package zip hashes
- fix: bulk repair registry backup indexes
- fix: drain registry backup retries safely (#2634)
- fix: preserve single-row ClawHub header (#2632)
- docs: restore OpenClaw-hosted ClawHub docs (#2631) — Revert "fix: serve docs subdomain root (#2630)" This reverts commit 9ef33726b8119e8f63d89e05e648497c3674156c. Revert "feat: move docs to canonical subdomain" This reverts commit...
- fix: serve docs subdomain root (#2630)
- feat: move docs to canonical subdomain
- update coderabbit
- mradermacher/Medina-Qwen3.5-27B-OpenClaw-Uncensored-i1-GGUF
- OpenClaw on DigitalOcean — Deploy your own personal AI assistant with OpenClaw on a DigitalOcean Droplet® server, a powerful open-source platform that runs entirely on your infrastructure. OpenClaw connects to the messaging...
- Claude Code refuses requests or charges extra if your commits mention "OpenClaw"
- OpenClaw – Moltbot Renamed Again
- OpenClaw’s memory is unreliable, and you don’t know when it will break
- How I connected OpenClaw to a Chinese frontier model and cut agent costs by 90% — I've been running OpenClaw for a few months. Love the agent workflow. Not so much the monthly...
- I think I found the first real reason to build AI agent workflows in OpenClaw — A practical case for using OpenClaw to automate receipt-to-ledger workflows: bounded inputs and outputs, clear human review gates, and a better split
- Building a Cross-Framework AI Agent Network Using XMPP (OpenClaw Hermes) — 🧩 The Problem: AI Agents Are Still Isolated Most AI agent frameworks today are powerful —...
- I thought the cheap model would save my OpenClaw bill, then I watched $100 disappear in 2 days — A practical DEV post for OpenClaw users: why using one cheap model everywhere often increases total cost, how to route by task risk instead, and why c
- Building a consent-preserving auth.md skill for OpenClaw agents — Signup forms were designed for humans with browsers. AI agents get a 401 and have nowhere to...
- How to Build a Multi Channel AI Appointment Agent Using AgenDuet, Bedrock Nova Sonic and OpenClaw — Ever had a conversation with a voice AI that suddenly froze up mid sentence? You ask it to change your appointment, and there is this…Continue reading on Medium »
- No GPU, No API, No Problem — Everyone says you need a $3,000 GPU to run AI locally. I had a 4GB laptop and a problem the cloud couldn’t solve. Here’s what actually…Continue reading on Level Up Coding »
- (PLR) OpenClaw AI Expertise Review: 2026 Pros & Cons — A smart, ready-to-sell PLR package that turns AI skills into revenue.Continue reading on Medium »
- OpenClaw and Virtualisation Infrastructure — Talking to Harvester HCI Like a Human — I have a two-node Harvester cluster and It runs VMs, manages storage and handles whatever workloads I throw at it.Continue reading on Medium »
- Is OpenClaw the Secret Weapon Against AI Risks? (What Silicon Valley Isn’t Telling You) — Subject: “We have a problem.”Continue reading on Medium »
- **How to Run OpenClaw på Raspberry Pi E35 Hardware i 2024 — Hemmeligheden Bag Lynhurtig… — Det lyder som science fiction, ikke? Men i 2024 er det ikke længere en drøm. Det er en realitet, som hundredvis af udviklere og hobbyister…Continue reading on Medium »
- How I deployed my openclaw assistant (and what I use it for) — Firstly.. what is openclaw?Continue reading on Medium »
- Giving Your AI Assistant a Voice and Eyes in Microsoft Teams — Giving Your AI Assistant a Voice — and Eyes — in Microsoft TeamsContinue reading on Medium »
- Kimi — AI assistant supports complex math, coding, image understanding, and reasoning with up to 128k token processing, short and long chain-of-thought tasks, visual and text input, accurate benchmarks,...
- openclaw called it on WCUP again. the model doesn't miss ┃🧫┃ $WCUP ┃💰┃ MC : $14.3K ┃🎯┃ TARGET MC : $195.5K ┃📈┃ POTE
- openclaw flagged HYRE early. price is moving exactly as predicted ┃🧫┃ $HYRE ┃💰┃ MC : $187.6K ┃🎯┃ TARGET MC : $917.3K
- openclaw flagged WOOLLY as a strong hold. not surprised ┃🧫┃ $WOOLLY ┃💰┃ MC : $46.7K ┃🎯┃ TARGET MC : $741.9K ┃📈┃ POT
- We built 16 AI agent personas so you don't have to. Introducing Soul Pack — pre-built SOUL.md files for Hermes, OpenCla
- openclaw called it on Gusic again. the model doesn't miss ┃🧫┃ $Gusic ┃💰┃ MC : $21.2K ┃🎯┃ TARGET MC : $245.7K ┃📈┃ PO
- openclaw on DSACHI —signals all pointing up ┃🧫┃ $DSACHI ┃💰┃ MC : $102.1K ┃🎯┃ TARGET MC : $547.5K ┃📈┃ POTENTIAL : +4
- PUMPVILLE is looking good on openclaw right now ngl ┃🧫┃ $PUMPVILLE ┃💰┃ MC : $99.4K ┃🎯┃ TARGET MC : $521.8K ┃📈┃ POTE
- BabyT signals are clean on openclaw. this one's on my radar 🦞 ┃🧫┃ $BabyT ┃💰┃ MC : $40.4K ┃🎯┃ TARGET MC : $638.3K ┃📈
- Comment AIDY and I’ll give you the link. So Hermes agent got a new feature. What is Hermes Agent? What is OpenClaw? How
- openclaw called ACE bullish. i'm listening ┃🧫┃ $ACE ┃💰┃ MC : $435.9K ┃🎯┃ TARGET MC : $1.4M ┃📈┃ POTENTIAL : +211% ┃
- For some weird reasons after been away for AI world it seems I miss tinkering with AI agents, openclaw, and hermes even
- openclaw called KANE bullish. i'm listening ┃🧫┃ $KANE ┃💰┃ MC : $17.5K ┃🎯┃ TARGET MC : $201.4K ┃📈┃ POTENTIAL : +1051
- BLOXX is looking good on openclaw right now ngl ┃🧫┃ $BLOXX ┃💰┃ MC : $563.5K ┃🎯┃ TARGET MC : $2.0M ┃📈┃ POTENTIAL : +
- RT @kevincodex: Earning OpenGateway inference tokens integrated now to OpenClaude. currently testing in-house ads rn
- openclaw flagged CHATOSHI as a strong hold. not surprised ┃🧫┃ $CHATOSHI ┃💰┃ MC : $504.1K ┃🎯┃ TARGET MC : $1.5M ┃📈┃
- PVE looking healthy according to openclaw. adding more ┃🧫┃ $PVE ┃💰┃ MC : $11.3K ┃🎯┃ TARGET MC : $143.8K ┃📈┃ POTENTI
- just ran NEMO through openclaw. risk is low, upside is real ┃🧫┃ $NEMO ┃💰┃ MC : $27.8K ┃🎯┃ TARGET MC : $285.6K ┃📈┃ P
- openclaw on Machi —momentum strong, risk low. easy ┃🧫┃ $Machi ┃💰┃ MC : $162.1K ┃🎯┃ TARGET MC : $839.8K ┃📈┃ POTENTIA
- SIGMA looking like a clean entry according to openclaw ┃🧫┃ $SIGMA ┃💰┃ MC : $1.5M ┃🎯┃ TARGET MC : $5.9M ┃📈┃ POTENTIA
- just ran ISO through openclaw. risk is low, upside is real ┃🧫┃ $ISO ┃💰┃ MC : $252.9K ┃🎯┃ TARGET MC : $1.8M ┃📈┃ POTE
Trending on X
- openclaw called it on WCUP again. the model doesn't miss ┃🧫┃ $WCUP ┃💰┃ MC : $14.3K ┃🎯┃ TARGET MC : $195.5K ┃📈┃ POTE
- openclaw flagged HYRE early. price is moving exactly as predicted ┃🧫┃ $HYRE ┃💰┃ MC : $187.6K ┃🎯┃ TARGET MC : $917.3K
- openclaw flagged WOOLLY as a strong hold. not surprised ┃🧫┃ $WOOLLY ┃💰┃ MC : $46.7K ┃🎯┃ TARGET MC : $741.9K ┃📈┃ POT
- We built 16 AI agent personas so you don't have to. Introducing Soul Pack — pre-built SOUL.md files for Hermes, OpenCla
- openclaw called it on Gusic again. the model doesn't miss ┃🧫┃ $Gusic ┃💰┃ MC : $21.2K ┃🎯┃ TARGET MC : $245.7K ┃📈┃ PO
- openclaw on DSACHI —signals all pointing up ┃🧫┃ $DSACHI ┃💰┃ MC : $102.1K ┃🎯┃ TARGET MC : $547.5K ┃📈┃ POTENTIAL : +4
- PUMPVILLE is looking good on openclaw right now ngl ┃🧫┃ $PUMPVILLE ┃💰┃ MC : $99.4K ┃🎯┃ TARGET MC : $521.8K ┃📈┃ POTE
- BabyT signals are clean on openclaw. this one's on my radar 🦞 ┃🧫┃ $BabyT ┃💰┃ MC : $40.4K ┃🎯┃ TARGET MC : $638.3K ┃📈
- Comment AIDY and I’ll give you the link. So Hermes agent got a new feature. What is Hermes Agent? What is OpenClaw? How
- openclaw called ACE bullish. i'm listening ┃🧫┃ $ACE ┃💰┃ MC : $435.9K ┃🎯┃ TARGET MC : $1.4M ┃📈┃ POTENTIAL : +211% ┃
- For some weird reasons after been away for AI world it seems I miss tinkering with AI agents, openclaw, and hermes even
- openclaw called KANE bullish. i'm listening ┃🧫┃ $KANE ┃💰┃ MC : $17.5K ┃🎯┃ TARGET MC : $201.4K ┃📈┃ POTENTIAL : +1051
- BLOXX is looking good on openclaw right now ngl ┃🧫┃ $BLOXX ┃💰┃ MC : $563.5K ┃🎯┃ TARGET MC : $2.0M ┃📈┃ POTENTIAL : +
- RT @kevincodex: Earning OpenGateway inference tokens integrated now to OpenClaude. currently testing in-house ads rn
- openclaw flagged CHATOSHI as a strong hold. not surprised ┃🧫┃ $CHATOSHI ┃💰┃ MC : $504.1K ┃🎯┃ TARGET MC : $1.5M ┃📈┃
- PVE looking healthy according to openclaw. adding more ┃🧫┃ $PVE ┃💰┃ MC : $11.3K ┃🎯┃ TARGET MC : $143.8K ┃📈┃ POTENTI
- just ran NEMO through openclaw. risk is low, upside is real ┃🧫┃ $NEMO ┃💰┃ MC : $27.8K ┃🎯┃ TARGET MC : $285.6K ┃📈┃ P
- openclaw on Machi —momentum strong, risk low. easy ┃🧫┃ $Machi ┃💰┃ MC : $162.1K ┃🎯┃ TARGET MC : $839.8K ┃📈┃ POTENTIA
- SIGMA looking like a clean entry according to openclaw ┃🧫┃ $SIGMA ┃💰┃ MC : $1.5M ┃🎯┃ TARGET MC : $5.9M ┃📈┃ POTENTIA
- just ran ISO through openclaw. risk is low, upside is real ┃🧫┃ $ISO ┃💰┃ MC : $252.9K ┃🎯┃ TARGET MC : $1.8M ┃📈┃ POTE
Releases
- OpenClaw repo stats for 2026-06-15 — 378788 stars, 79241 forks, 2570 contributors
Community
- iMessage: inbound message silently dropped when sent immediately after agent outbound reply (same group chat) — Environment - OpenClaw: 2026.5.22 (a374c3a) - imsg: 0.9.0 - macOS: Tahoe (26), M4 MacBook Air - iMessage plugin config: debounceMs: 3000 Problem User messages sent to a group chat immediately after...
- fix(compaction): forward abortSignal through delegateCompactionToRuntime bridge — Summary When a user presses the stop button during auto-compaction, the abort signal cancels the wait but the compaction LLM call continues in-flight and writes its result to the session transcript....
- fix(agents): guard tool definition schema reads — Summary - guard agent tool definition metadata reads so a hostile or broken tool schema getter cannot crash definition adaptation - represent unreadable parameter schemas as an invalid schema...
- fix(qqbot): keep markdown table chunks valid — Summary - Add a QQ Bot markdown chunker that keeps table chunks self-contained across message boundaries. - Preserve table state across block deliveries and flush unfinished table-row fragments as...
- fix(telegram): rotate in-flight answer preamble at tool boundary — Summary Fixes a Telegram bug where an in-flight answer preamble is overwritten instead of preserved when a tool runs mid-stream and post-tool text resumes. When the model streams a short preamble...
- Feature Request: Display inbound media attachments as image previews in WebUI — Summary When an image is sent to OpenClaw via a channel (e.g. Telegram or iMessage), the image is correctly saved to media://inbound/ and processed by the AI model. However, the WebUI chat only...
- feat(auto-reply,codex): lightweight reply lane and Codex watchdog — Summary - add a conservative lightweight reply lane for obvious low-risk text-only chat replies - keep uncertain, tool/action, media/link, high-stakes, code/config, and unresolved reply-target...
- fix(agents): forward turn-source for embedded plugin approvals (supersedes #84205) — Summary Forwards turn-source routing fields through the embedded plugin before_tool_call approval chain so plugin approvals can be delivered back to the originating chat (Telegram, Discord, Slack,...
- State directory permissions can be relaxed back to 0775 by update-check JSON writes — Hello, I noticed a recurring security audit warning where the OpenClaw state directory is reported as group-writable again after it was manually hardened to 0700. This does not seem to be a one-off...
- feat(memory-core): optional cross-encoder rerank stage for memory search — Summary OpenClaw's builtin (sqlite-vec) memory search does first-stage hybrid retrieval (dense vec0 ∥ FTS5 BM25, fused) → temporal decay → MMR diversity, but has no second-stage relevance reranker....
- fix(llm-core): wrap unreadable tool schemas — Summary - wrap unreadable tool parameter schemas in controlled validation errors - keep ordinary argument validation failures on the existing Validation failed path - add regression coverage for a...
- fix(doctor): skip top-level allowlist scan when active per-account scopes exist — Summary The empty-allowlist doctor scanner treated the top-level channel config as its own standalone account, producing false 'empty allowlist' warnings when every active account under that channel...
- docs(logging): document exact model.usage event field names in diagnostic catalog (fixes #49046) — What Expand the model.usage entry in the OpenTelemetry diagnostic event catalog with the exact JSON field names emitted by the agent runner. Add a cross-reference from the logging docs to the...
- fix(agents): harden tool search schema catalog — Summary - make Tool Search catalog fingerprinting tolerate unreadable schema objects - snapshot catalog parameters into JSON-safe metadata so describe/search cannot be poisoned by hostile schema...
- fix: prevent memory flush failure from aborting user reply (#85645) — Summary When memoryFlush fails during the reply pipeline (provider timeout, rate limit, network error), the visible error payload was treated as the final reply via replyOperation.fail("run_failed",...
- perf: structuredClone instead of JSON.parse(stringify) — Summary This PR implements the highest-impact, lowest-effort performance improvement identified in a comprehensive codebase analysis. Replaces JSON.parse(JSON.stringify()) with native...
- fix(browser): bring tab to foreground before screenshot capture — Summary What problem does this PR solve? browser screenshot (and the screenshot-backed snapshot path) hang until the CDP socket times out (~20s) whenever the target tab is backgrounded. Why does this...
- Telegram poll_answer updates are not routed to agent turns — Summary Telegram polls can be created from OpenClaw, but non-anonymous poll votes are not routed to the agent as actionable turns/events. This makes poll-based approvals look usable in Telegram while...
- fix(diagnostics): prune non-idle ghost entries after failed stuck-session recovery — Summary When stuck-session recovery fails (status: "failed"), recoveryOutcomeMutatesSessionState() returns false → the diagnostic entry stays in "processing" / "waiting" state permanently. The...
- fix(cron): resolve lastRunStatus in cron list/show human output — Summary openclaw cron list and openclaw cron show print idle for a job whose real status is ok / error / skipped whenever the job's state carries only lastRunStatus (the primary field) and not the...
- fix(agents): honor disabled envelope timestamps at model boundary — Opened on behalf of Onur Solmaz (osolmaz). AI-assisted contribution. Summary OpenClaw could still send timestamp bytes to model input even when agents.defaults.envelopeTimestamp was "off". That...
- fix(provider): harden provider tool schema hooks — Summary - Harden provider tool schema normalization against plugin-supplied tools with unreadable parameters fields. - Share guarded tool-field reads across Gemini, DeepSeek, and native OpenAI schema...
- Feature request: Route Telegram group topics to different agents — Problem Currently, OpenClaw routes Telegram messages by channel + accountId (bot). When a bot joins a group with multiple topics (forum topics), all messages from all topics go to the same agent....
- #92664: [Bug]: read tool fails to read GBK-encoded text files on Chinese Windows (displays garbled text) — Summary - Fix garbled text when reading non-UTF-8 files (GBK on Chinese Windows, Shift_JIS on Japanese Windows) by reusing the existing Windows active- codepage decoder (decodeWindowsOutputBuffer) in...
- fix(agents): harden system prompt tool reporting — Summary - Harden system prompt report generation against tool metadata getters that throw while reading name, description, label, or parameters. - Keep schema size/property telemetry best-effort when...
- Voice-call plugin: hangup aborts in-flight tool calls (sessions_spawn killed before child initializes) — Problem When a voice call ends, the gateway aborts the embedded agent run via lane management for the voice lane. All in-flight tool calls — including sessions_spawn — receive the abort signal...
- fix(plugins): harden plugin tool shape reads — Summary - Harden plugin tool runtime shape checks against hostile getters on tool names, descriptions, execute functions, and parameter schemas. - Keep malformed plugin tools isolated so a bad tool...
- fix(status): avoid stale session context windows — Summary Fix /status context-window reporting when persisted session runtime metadata is stale. The status message can display the currently selected/configured model while still using an old...
- docs: document embedded compaction context contracts — Summary - Document the embedded compaction runtime context contract. - Clarify how compaction target resolution applies model overrides, auth-profile safety, and Codex-runtime OpenAI routing. - Keep...
- [Bug]: session-memory hook model config ignored for LLM slug generation — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary The model field in hooks.internal.entries.session-memory is not used when generating LLM-based slugs....
- OpenClaw sponsors (110 total) — 110 sponsors supporting OpenClaw
- fix: allow guarded transfer of deleted skills (#2647)
- feat: add plugin versions tab (#2646) — feat: add package versions API wrapper fix: omit empty package version cursor feat: add plugin versions tab fix: paginate plugin version history fix: remove unsafe plugin version downloads fix:...
- revert: retire API key required skill feature (#2644)
- fix: serialize registry backup index writes
- fix: normalize publisher recovery handle comparisons (#2643) — Normalize mixed-case stored user handles in publisher recovery guards and retire the previous mixed-case handle before handoff.
- fix: serialize registry backup retry drains (#2641)
- fix: add guarded personal publisher recovery (#2642) — Summary: - The PR adds an admin-only personal publisher recovery flow with HTTP API, admin CLI support, shared response schema, docs/spec notes, and tests. - Reproducibility: yes. Source inspection...
- fix: compact theme controls in profile menu (#2640)
- fix: skip indexed registry backup retries (#2639)
- test: cover skill summary digest sync (#2638)
- fix: preserve legacy package ZIP hashes (#2636) — test: fix indexed row helper lint fix: preserve legacy package zip hashes
- fix: bulk repair registry backup indexes
- fix: drain registry backup retries safely (#2634)
- fix: preserve single-row ClawHub header (#2632)
- docs: restore OpenClaw-hosted ClawHub docs (#2631) — Revert "fix: serve docs subdomain root (#2630)" This reverts commit 9ef33726b8119e8f63d89e05e648497c3674156c. Revert "feat: move docs to canonical subdomain" This reverts commit...
- fix: serve docs subdomain root (#2630)
- feat: move docs to canonical subdomain
- update coderabbit
News
- mradermacher/Medina-Qwen3.5-27B-OpenClaw-Uncensored-i1-GGUF
- OpenClaw on DigitalOcean — Deploy your own personal AI assistant with OpenClaw on a DigitalOcean Droplet® server, a powerful open-source platform that runs entirely on your infrastructure. OpenClaw connects to the messaging...
- Claude Code refuses requests or charges extra if your commits mention "OpenClaw"
- OpenClaw – Moltbot Renamed Again
- OpenClaw’s memory is unreliable, and you don’t know when it will break
- How I connected OpenClaw to a Chinese frontier model and cut agent costs by 90% — I've been running OpenClaw for a few months. Love the agent workflow. Not so much the monthly...
- I think I found the first real reason to build AI agent workflows in OpenClaw — A practical case for using OpenClaw to automate receipt-to-ledger workflows: bounded inputs and outputs, clear human review gates, and a better split
- Building a Cross-Framework AI Agent Network Using XMPP (OpenClaw Hermes) — 🧩 The Problem: AI Agents Are Still Isolated Most AI agent frameworks today are powerful —...
- I thought the cheap model would save my OpenClaw bill, then I watched $100 disappear in 2 days — A practical DEV post for OpenClaw users: why using one cheap model everywhere often increases total cost, how to route by task risk instead, and why c
- Building a consent-preserving auth.md skill for OpenClaw agents — Signup forms were designed for humans with browsers. AI agents get a 401 and have nowhere to...
- How to Build a Multi Channel AI Appointment Agent Using AgenDuet, Bedrock Nova Sonic and OpenClaw — Ever had a conversation with a voice AI that suddenly froze up mid sentence? You ask it to change your appointment, and there is this…Continue reading on Medium »
- No GPU, No API, No Problem — Everyone says you need a $3,000 GPU to run AI locally. I had a 4GB laptop and a problem the cloud couldn’t solve. Here’s what actually…Continue reading on Level Up Coding »
- (PLR) OpenClaw AI Expertise Review: 2026 Pros & Cons — A smart, ready-to-sell PLR package that turns AI skills into revenue.Continue reading on Medium »
- OpenClaw and Virtualisation Infrastructure — Talking to Harvester HCI Like a Human — I have a two-node Harvester cluster and It runs VMs, manages storage and handles whatever workloads I throw at it.Continue reading on Medium »
- Is OpenClaw the Secret Weapon Against AI Risks? (What Silicon Valley Isn’t Telling You) — Subject: “We have a problem.”Continue reading on Medium »
- **How to Run OpenClaw på Raspberry Pi E35 Hardware i 2024 — Hemmeligheden Bag Lynhurtig… — Det lyder som science fiction, ikke? Men i 2024 er det ikke længere en drøm. Det er en realitet, som hundredvis af udviklere og hobbyister…Continue reading on Medium »
- How I deployed my openclaw assistant (and what I use it for) — Firstly.. what is openclaw?Continue reading on Medium »
- Giving Your AI Assistant a Voice and Eyes in Microsoft Teams — Giving Your AI Assistant a Voice — and Eyes — in Microsoft TeamsContinue reading on Medium »
- Kimi — AI assistant supports complex math, coding, image understanding, and reasoning with up to 128k token processing, short and long chain-of-thought tasks, visual and text input, accurate benchmarks,...
Don't miss what's next. Subscribe to Openclaw Newsletter: