Stay up to date with OpenClaw releases, trending topics, and ecosystem updates.
Showing 9 of 83 articles
A new warning in OpenClaw's gateway sandboxing reference unmasks a deployment gotcha that's been silently breaking containerized installs: when the Gateway runs inside Docker but bridges the host socket for sandbox orchestration, the workspace path has to be identical on both sides — or the heartbeat dies with a permission error.
OpenClaw's documentation now ships a first-party Hostinger guide covering both the 1-Click managed plan and a Docker-driven VPS install. Pre-purchased AI credits at checkout, WhatsApp and Telegram setup steps, and an hPanel-driven update flow round out the new reference — and quietly redraw who can self-host an AI assistant.
A new section in OpenClaw's automation reference exposes a Vixie-cron quirk that's been sabotaging scheduled agents for years. When day-of-month and day-of-week are both non-wildcards, the scheduler fires on either — not both. Here's the trap, the workaround, and why the warning took so long.
Five shared scenarios. Two summary files. One pass/fail gate. OpenClaw's new first-wave parity pack runs GPT-5.4 and Opus 4.6 through the same agentic tests and produces a machine-readable verdict — ending months of anecdotal model-behavior debates.
OpenClaw now supports pluggable agent harnesses, letting external runtimes like Codex execute agent turns natively. The first extension ships with full app-server integration, thread management, and a new plugin SDK surface.
A new opt-in execution contract detects when AI agents respond with plans instead of tool calls, retries with steering instructions, and fails closed if the agent still won't act. Currently targeting GPT-5 models.
OpenClaw's latest update replaces GLM-4.7 Flash with Google's Gemma 4 as the recommended local model for Ollama users, while also bumping cloud model defaults to newer versions of GLM and MiniMax.
Mistral Small 4 now supports adjustable reasoning in OpenClaw, letting users dial thinking intensity up or down through the platform's existing thinking-level controls.
OpenClaw now ships Arcee AI as a bundled provider plugin, giving users instant access to the Trinity Large Thinking, Trinity Large Preview, and Trinity Mini 26B models — with both direct API and OpenRouter authentication paths.
Get started in under 5 minutes with DeployClaw.