The latest AI news and updates of OpenClaw — Open-source self-hosted AI assistant that runs across messaging platforms with plugin skill support. Covering OpenClaw's latest product updates and launches from the past 90 days.
OpenClawOct 24OpenClaw published a conversation detailing the security architecture of OpenClaw Enterprise, an open-source platform for persistent AI agents. The system isolates the control plane from untrusted agent workloads to enforce security policies like RBAC and audit logs. Red Hat and NVIDIA are contributing to the project, which is now pre-1.0 and open for enterprise pilots.
OpenClawOct 20OpenClaw integrated the MiniMax M3.1 Flash model into its self-hosted AI assistant. Installations running OpenClaw v2026.10.1 or newer gain access to the model without requiring a client update.
OpenClawOct 19OpenClaw released v2026.10.5, introducing a searchable sidebar with subagent panels, Skill Workshop 2.0 with version history and Undo, and public conversation links that allow anonymous access. This update includes 2,250 pull requests from 637 contributors, focusing on agent management and session sharing.
OpenClawOct 16OpenClaw launched out-of-the-box support for the new NVIDIA RTX Spark laptops. The integration features a guided setup that eliminates terminal requirements, offering a choice between local models or cloud providers. This collaboration with NVIDIA and Microsoft brings the AI assistant to the new Windows hardware lineup.
OpenClawOct 11OpenClaw released v2026.10.1, introducing installed-skill discovery, MCP app support beside chat, and full backups to external storage or Cloudflare R2. The update also streamlines Mac setup and includes fixes from 2,425 pull requests by 402 contributors.
OpenClawOct 3OpenClaw released v2026.9.8, adding OpenAI's GPT-6.1 Sol as a selectable model. The update also fixes delegated agent replies, reduces memory usage in larger Codex setups, and resolves Windows startup and update failures. This release includes 43 pull requests from 8 contributors.
OpenClawOct 3OpenClaw integrated Tencent’s AI-Infra-Guard (AIG) into ClawScan, the CLI tool powering security reviews on ClawHub. Every skill and plugin uploaded to ClawHub now undergoes AIG review across nine risk categories, including instruction hijacking and remote payload execution. ClawScan evaluates AIG findings alongside NVIDIA’s SkillSpector, with an AI judge performing the final assessment to classify malicious content.
OpenClawSep 30OpenClaw released v2026.9.7, delivering snappier performance under load and smoother long conversations. The update introduces support for OpenAI’s Agents API and Sign in with ChatGPT (Beta), alongside improved Apple device chat and restart recovery. It also hardens the update process with new verified backups and rollback protection for users upgrading from v2026.9.5.
OpenClawSep 29The OpenClaw Foundation announced OpenClaw Enterprise, an open-source control plane for managing persistent AI agents in sensitive environments. Developed with Red Hat, NVIDIA, and OpenAI, the platform provides governance, auditability, and security boundaries for agentic workloads. It runs on internal infrastructure and remains free for any organization to use, with internal pilots currently underway at Red Hat and OpenAI.
OpenClawSep 26Microsoft introduced Autopilot, an always-on personal agent built on the OpenClaw runtime. The collaboration includes significant upstream contributions from Microsoft engineers, such as policy conformance checks, native Windows support, and reliability improvements. Autopilot is currently rolling out to private preview customers.
OpenClawSep 24OpenClaw removed approximately 400,000 lines of code from its test suite while maintaining existing code coverage. The team utilized its new test-audit skill, which automates the identification and removal of redundant or low-value tests generated by AI models. The skill employs an authoring gate and audit sweeps to prune test surfaces against defined value bars.
OpenClawSep 24OpenClaw 2026.9.6 adds supported routes for Claude Opus 5.5, GPT-6 Sol and Luna, and Grok 4.7. The release introduces restart recovery for unfinished work, 30-day Usage reporting, a public GitHub reader, and remote workspace tools for Files, Memory, and Skills. It also adds live meeting notes and optional decision models, including TypeSafe Jev and local ONNX choices.
OpenClawSep 22OpenClaw added support for decision models in its core and plugins, enabling agents to perform fast, typed judgments like scoring or classification. The api.runtime.decisions.evaluate interface and decision_evaluate core tool provide a shared path for configuring these models alongside conversational ones. This opt-in feature is available in current development checkouts, with provider packages pending a supporting release.
OpenClawSep 19OpenClaw released version 2026.9.5, introducing atomic updates, plugin hot reloading, conversation sharing, expanded GPT Live, shared browser pages, conversation archiving, and specialist-agent setup. This release incorporates 4,179 pull requests from 502 contributors. The update focuses on platform reliability and multi-user collaboration surfaces.
OpenClawSep 11OpenClaw v2026.9.4 introduces unified plugin and skill discovery, a steerable chat workflow for learning skills from past conversations, and support for OpenAI's GPT Image 2.5. The update also adds expanded cloud-worker management and enables terminal-based agent Q&A. This release includes contributions from 293 people.
OpenClawSep 8OpenClaw v2026.9.3 introduces clean update recovery and faster session reconnection to minimize work interruptions. The update adds live browser automation streaming, revocable public chat links, and searchable meeting transcripts. Repository-backed work now runs directly on cloud nodes, and the Mac app gains native browser tabs that persist across conversation switches.
OpenClawSep 6OpenClaw v2026.9.2 adds support for OpenAI's GPT-6 Astra and Meta's Muse Spark 1.3, enabling text and image processing for eligible accounts. The update improves reliability by allowing interrupted tasks to resume after a restart and accelerating long-chat performance. Additionally, shared multi-agent Gateways now default to broader cross-agent conversation access, requiring users to review visibility settings before upgrading.
OpenClawSep 3OpenClaw released v2026.9.1, adding native Mermaid diagram rendering across chat surfaces and a streamlined setup path that reuses existing AI access. The update improves update safety with readiness checks and reduces memory overhead for long conversations. This release includes 1,186 pull requests and contributions from 281 people.
OpenClawSep 2OpenClaw released v2026.8.2, introducing a Linux desktop companion and a Home button that docks your personal agent beside active work. The update adds four new Control UI themes and enables background session creation. It also includes a broad set of reliability fixes for voice, updates, and recovery processes across the platform.
OpenClawAug 31OpenClaw released version 2.0, its largest update to date, featuring a rebuilt browser interface and new shared cloud sessions for multiplayer collaboration. Installation now auto-detects existing API keys and local models to accelerate setup. This release shifts the platform from a single-user agent into a collaborative environment for teams and families to manage tasks together.
OpenClawAug 12OpenClaw added support for Grok 4.6, a model developed by xAI. The platform now includes Grok 4.6 as an available model for autonomous agent workflows.
OpenClawJul 30OpenClaw launches monthly extended-stable releases, providing long-lived support with backported security and reliability fixes. The platform also debuts a public maturity scorecard, which inventories features by readiness to track suitability for critical production workloads. These additions establish a foundation for enterprise-grade stability in autonomous agent deployments.
Frequently asked questions