OpenClaw AI Framework v2026.2.17 Adds Anthropic Claude Support

OpenClaw AI Framework released version 2026.2.17 on February 17, 2026. It brings support for Anthropic’s Claude Sonnet 4.6 model and a 1-million-token context window in beta. Security fixes address past vulnerabilities too.

This update helps agents handle long conversations better. Developers steipete pushed over 100 changes. It fits iOS, Slack, and more platforms seamlessly.

BEST WINTER DEALS

Editor's Choice

Private Internet Access

Access content across the globe at the highest speed rate.

70% of our readers choose Private Internet Access

70% of our readers choose ExpressVPN

ExpressVPN

Browse the web from multiple devices with industry-standard security protocols.

Nord VPN

Faster dedicated servers for specific actions (currently at summer discounts)

The framework faces security scrutiny from earlier flaws like CVE-2026-25253. That allowed remote code execution via tokens and WebSockets. New patches harden against credential theft and more.

Key New Features

Sonnet 4.6 works with fallbacks if catalogs lag. Subagents spawn via /subagents command. iOS share extension forwards text, URLs, images directly.

Slack gains message streaming and draft previews. Telegram adds styled buttons and reactions. Discord improves /exec autocomplete and button controls.

Cron jobs get webhooks and usage logs. Web tools use URL allowlists. Memory search boosts with FTS fallback and query expansion.

Voice calls preload greetings. Docker builds optional Chrome. Z.AI tools stream by default.

Security and Risks

Past audits found hundreds of issues, including unrestricted shell access. ClawHub marketplace had malicious plugins at 10.8% rate. They stole wallets and Keychain data via social tricks.

Version 2026.2.17 patches critical paths like config traversal. Update fast to avoid infostealers targeting OpenClaw files.

Users gain enterprise-ready tools now. But secure setups matter. Limit agent access and review skills.

FAQ