← All updates

DAILY RELEASE TRACKER

OpenClaw

Personal AI assistant platform, gateway, skills, memory, chat integrations.

GitHub releasesLast checked: Nov 01, 2026Refreshes daily
v2026.10.10-beta.5Latest

openclaw 2026.10.10-beta.5

Explanation

OpenClaw v2026.10.10-beta.5 focuses on security, authentication, or permission flow changes, model selection, reasoning, or inference behavior, developer workflow and CLI productivity. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • Cron: missed slots with a stale stored next run time are no longer dropped, and startup catch-up runs are admitted after a restart. (#179928; related #179927) Thanks @andrewbowlds, @RomneyDa, and @shakkernerd.
  • iOS: connect manually when the host includes `https://` or `wss://`, and prevent history recall during composition or selection. (#169875, #143056; related #144678) Thanks @masatohoshino, @RomneyDa, @bjesuiter, and @vincentkoc.
  • Complete contribution record
  • PR #172691 Thanks @joshavant.
Read original update →
v2026.9.36

openclaw 2026.9.36

Explanation

OpenClaw v2026.9.36 focuses on security, authentication, or permission flow changes, model selection, reasoning, or inference behavior, developer workflow and CLI productivity. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • Microsoft Teams allowlists stay restrictive: name-keyed `teams` or `groupAllowFrom` entries that do not resolve now grant nothing instead of opening the bot to every team or group. (#184989) Thanks @eleqtrizit, @RomneyDa, and @shakkernerd.
  • Replies keep their text: comparison prose after emoji and other astral characters is no longer stripped from outbound replies. (#187415) Thanks @RomneyDa.
  • Certificate trust and Docker images: the eMudhra trust remediation is applied, and Docker release images carry source/version-bound evidence with more reliable architecture builds. (#186639) Thanks @RomneyDa.
  • Reply delivery: keep comparison prose after astral characters in the outbound sanitizer. (#187415) Thanks @RomneyDa.
Read original update →
v2026.10.9

openclaw 2026.10.9

Explanation

OpenClaw v2026.10.9 focuses on security, authentication, or permission flow changes, developer workflow and CLI productivity, memory, context handling, or conversation state. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • Apply `tools.fs.workspaceOnly` to Gateway-built tools, so delegate changes still require approval and media tools stay inside the workspace. (#185094) Thanks @eleqtrizit, @RomneyDa, and @shakkernerd.
  • Keep live credentials out of MCP CLI inspection and mask Google OAuth client secrets in logs. (#180321, #180308, #183193) Thanks @brandon-julio-t, @wr-web, @LeizhenZhang, and @RomneyDa.
  • Recover intact agent databases and stale full-text indexes after SQLite tokenizer changes, including update snapshots. (#176234, #176220, #176833, #181206) Thanks @ndakota79 and @RomneyDa.
  • Repair publication-complete updates after a remount while retaining the original launcher and sealed recovery helper. (#186557, #186527) Thanks @ericcaiwx-star, @sennyhoney, and @RomneyDa.
Read original update →
v2026.10.8

openclaw 2026.10.8

Explanation

OpenClaw v2026.10.8 focuses on security, authentication, or permission flow changes, model selection, reasoning, or inference behavior, developer workflow and CLI productivity. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • Fresh OpenAI setup: first-run model metadata no longer restricts the model picker to one model. (#175945) Thanks @matthematics1137 and @jb-dev1.
  • Channel accounts: default Slack, Google Chat, X, Nextcloud Talk and Zalo accounts keep starting and keep their secrets after a named account is added. (#176267, #180450, #182608) Thanks @555aaditya, @progerjkd, and @masatohoshino.
  • Slack: an isolated `invalid_auth` is retried before a socket-mode channel is disabled, and silent relay network drops reconnect. (#181056, #182368) Thanks @mushuiyu886 and @marns.
  • Remote access and history: Tailscale Serve and Funnel recover after tailscaled restarts, and cron history stays readable after provider changes. (#180081, #181311) Thanks @enki, @saariuslystoned, and @xtimor.
Read original update →
v2026.10.10-beta.4

openclaw 2026.10.10-beta.4

Explanation

OpenClaw v2026.10.10-beta.4 focuses on security, authentication, or permission flow changes, model selection, reasoning, or inference behavior, developer workflow and CLI productivity. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • Gateway security: keep an environment-referenced `hooks.token` out of agent host-exec and CLI child environments at startup and across config reloads, while preserving explicit per-command overrides. (#179199) Thanks @VACInc and @RomneyDa.
  • Model refresh: let explicit catalog refreshes use their existing extended discovery budget through HTTP acquisition while ordinary preparation remains bounded. (#177313) Thanks @jlapenna and @RomneyDa.
  • Runtime security: backport patched embedded runtimes to the 2026.10.10 line. (#181074) Thanks @RomneyDa and @joshavant.
  • Claude CLI: stop prompt-cache prefix churn after a child completion. (#172026) Thanks @Marvinthebored and @VACInc.
Read original update →
v2026.9.35

openclaw 2026.9.35

Explanation

OpenClaw v2026.9.35 focuses on security, authentication, or permission flow changes, model selection, reasoning, or inference behavior, developer workflow and CLI productivity. This note is useful as a direction signal for teams tracking AI agent tooling and production workflows.

Official highlights

  • SQLite 3.54 data safety: Doctor and update snapshots now repair stale full-text indexes left by the SQLite tokenizer change instead of quarantining agent databases. Fixes #176220. (#176234, #181206)
  • Kimi K3.1 Pro: `moonshot/kimi-k3.1-pro`, `kimi/k3.1-pro`, and their Highspeed variants get the K3 thinking, effort, and video controls; K3.1 rejects `off` locally instead of returning a provider 400. (#181923) Thanks @obviyus.
  • Gateway responsiveness: bound regex and tag scans for Microsoft Teams HTML, tool-call-shaped replies, and long streamed markup. (#183438, #183252, #174833, #175472, #180765) Thanks @LeizhenZhang and @JasonBuildAI.
  • Plugins: replace dangling dependency junctions instead of refusing plugin migration, copy native captures across filesystems, and tolerate ctime-only stat drift in native companion admission. (#179549, #179445, #179949) Thanks @coaiMax.
Read original update →
v2026.9.34

openclaw 2026.9.34

Explanation

This extended-stable release enhances workspace safety by validating patch envelopes to prevent file deletion errors and restores critical `before_tool_call` hooks for agent execution. Operators benefit from improved managed upgrade convergence and gateway stability, ensuring reliable plugin reconciliation and robust model startup under varying ownership conditions.

Official highlights

  • Safer agent file edits: `apply_patch` now validates an ordered envelope before mutating the workspace, so a later invalid hunk cannot delete files changed by an earlier operation. (#160409) Thanks @masatohoshino, @obviyus, and @RomneyDa.
  • More reliable managed upgrades: app-owned updates now converge official plugins alongside core and retain repair diagnostics. Fixes #174120. (#174257) Thanks @kazuyuki-eguchi and @RomneyDa.
  • Memory Wiki data safety: OKF concept re-imports now preserve human-authored notes. Fixes #175700. (#175707) Thanks @YangManBOBO and @RomneyDa.
  • Plugin hooks restored: `before_tool_call` hooks now run under `openclaw agent exec`. Fixes #158824. (#159333) Thanks @joseairosa and @RomneyDa.
Read original update →
v2026.10.7

openclaw 2026.10.7

Explanation

This update enhances system resilience by introducing replay-safe session recovery and safer update activation mechanisms, ensuring that interrupted operations and maintenance events do not corrupt state or block workflows. Additionally, it resolves critical stability issues such as Gateway CPU wedging during log rollover and improves macOS onboarding, providing operators with more robust infrastructure and developers with reliable session continuity.

Official highlights

  • Google models: add Gemini 4 Argon support. (#171852) Thanks @RomneyDa.
  • macOS onboarding: allow fresh profiles to start the Gateway and complete automatic setup. (#173246) Thanks @RomneyDa.
  • Gateway logging: prevent a deleted or stale temporary log directory from wedging the Gateway at full CPU during date rollover. (#160433) Fixes #130807 Thanks @masatohoshino, @WadeOwenWatts, and @RomneyDa.
  • Control UI queue state: clear slash-steer chips after the queued command is consumed. (#180024) Thanks @RomneyDa.
Read original update →