OpenClaw v2026.2.17: Claude Sonnet 4.6 Support, 1M Context, Slack/Telegram Enhancements
OpenClaw releases major update with Claude Sonnet 4.6 and 1M context window support, Slack native streaming, Telegram inline button styles, iOS Share Extension, and critical security fixes (OC-09) among 100+ changes.
OpenClaw released version 2026.2.17 on February 17, implementing over 100 changes including Claude Sonnet 4.6 support, 1M token context window beta, major Slack/Telegram improvements, iOS feature enhancements, and critical security fixes.
Claude Sonnet 4.6 and 1M Context Support
Sonnet 4.6 Integration
The release adds support for Claude Sonnet 4.6 via the anthropic/claude-sonnet-4-6 alias, with forward-compatible fallback when upstream catalogs still only expose Sonnet 4.5.
1M Token Context Window (Beta)
Opus/Sonnet models now support opt-in 1M token context window. Setting params.context1m: true in model parameters maps to the anthropic-beta: context-1m-2025-08-07 header.
This enables processing entire codebases, lengthy contracts, and dozens of research papers in a single request, significantly improving long-context reasoning performance.
Messaging Platform Enhancements
Slack Native Streaming
Implemented native single-message text streaming for Slack using chat.startStream/appendStream/stopStream. Reply threading aligns with replyToMode configuration, streaming is enabled by default, and falls back to normal delivery on failure.
Telegram Inline Button Styles
Added inline button style support (primary/success/danger) across message tool schema, Telegram action parsing, send pipeline, and runtime prompt guidance.
User message reactions are now surfaced as system events, with notification scope configurable via channels.telegram.reactionNotifications.
Subagent Functionality Expansion
/subagents spawn Command
Added /subagents spawn for deterministic subagent activation from chat commands, improving control over subagent initialization.
Enhanced Context Management
Subagents now preemptively guard accumulated tool-result context before model calls by truncating oversized outputs and compacting oldest tool-result messages to prevent context-window overflow crashes.
Explicit recovery guidance was added for [compacted: tool output removed to free context] and [truncated: output exceeded context limit] markers, instructing re-reading with smaller chunks instead of full-file cat.
Comprehensive iOS Improvements
Share Extension
Added iOS share extension that forwards shared URL/text/image content directly to gateway agent.request, with delivery-route fallback and optional receipt acknowledgements.
Talk Mode Enhancements
- Background Listening: Toggle keeps Talk Mode active while backgrounded (off by default for battery safety)
- Voice Directive Hint: Toggle for Talk Mode prompts to disable ElevenLabs voice-switching instructions, saving tokens when not needed
- Hardened Barge-in: Disables interrupt-on-speech when output route is built-in speaker/receiver, reducing false interruptions from local TTS bleed-through
Onboarding Improvements
Added QR-first onboarding wizard with setup-code deep link support, pairing/auth issue guidance, and device-pair QR generation improvements for Telegram/Web/TUI fallback flows.
Cron/Automation Feature Expansion
Webhook-Based Delivery
Separated per-job webhook delivery (delivery.mode = "webhook") from announce delivery. Enforces valid HTTP(S) webhook URLs with temporary legacy notify + cron.webhook fallback for stored jobs.
Schedule Stagger Control
Added deterministic default stagger for recurring top-of-hour cron schedules (including 6-field seconds cron). Auto-migrates existing jobs to persisted schedule.staggerMs and adds openclaw cron add/edit --stagger plus --exact overrides for per-job timing control.
Usage Telemetry
Logs per-run model/provider usage telemetry in cron run logs/webhooks, with a local usage report script for aggregating token usage by job.
Security Fixes
OC-09 Credential Theft Path
Fixed OC-09 credential-theft path via environment-variable injection.
Config $include Confinement
Confined $include resolution to the top-level config directory, hardened traversal/symlink checks with cross-platform-safe path containment, and added doctor hints for invalid escaped include paths.
Session Transcript File Permissions
New session transcript JSONL files are created with user-only (0o600) permissions. Extended openclaw security audit --fix to remediate existing transcript file permissions.
Tool Functionality Improvements
read Tool Auto-Paging
The read tool now auto-pages across chunks when no explicit limit is provided, scaling per-call output budget from model contextWindow so larger contexts can read more before context guards activate.
Web Search/Fetch URL Allowlists
Added URL allowlists for web_search and web_fetch to address security and compliance requirements.
Browser Launch Customization
Added extraArgs config for custom Chrome launch arguments, enabling advanced browser control including headless mode and sandbox disabling.
Discord/Mattermost/Feishu Improvements
Discord
- Exposed native
/execcommand options (host/security/ask/node) for autocomplete and structured inputs - Added support for reusable interactive components with
components.reusable=true - Implemented per-button
allowedUsersallowlist to restrict button clicks
Mattermost
Added emoji reaction actions plus reaction event notifications, including explicit boolean remove flag to avoid accidental removals.
Feishu
Added Bitable create-app/create-field tools for automation workflows.
Memory Search Enhancement
Added FTS (full-text search) fallback plus query expansion for memory search, improving semantic search accuracy and providing alternative search paths for low-relevance hits.
Provider Support Expansion
Z.AI
Enabled tool_stream by default for real-time tool call streaming, with opt-out via params.tool_stream: false.
OpenAI Codex CLI
Added OpenAI Codex CLI auth provider integration.
Ollama/Qwen
Handles Qwen 3 reasoning field format in Ollama responses.
Docker/Infrastructure Improvements
Added optional OPENCLAW_INSTALL_BROWSER build arg to preinstall Chromium + Xvfb in the Docker image, avoiding runtime Playwright installs.
Key Bug Fixes (Selected)
Reply Threading Consistency
Maintains reply context sticky across streamed/split chunks and preserves replyToId on all chunk sends across shared and channel-specific delivery paths (including iMessage, BlueBubbles, Telegram, Discord, and Matrix), ensuring follow-up bubbles stay attached to the same referenced message.
Telegram Streaming Stability
- Keeps draft-stream preview replies attached to the user message for
replyToMode: "all"in groups and DMs - Prevents streaming final replies from being overwritten by later final/error payloads
- Debounces first draft-stream preview update with 30-char threshold
Cron Job Loop Fix
Prevents spin loops when cron jobs complete within the scheduled second by advancing the next run and enforcing a minimum refire gap.
Gateway Auto-Update Improvements
Prevents restart crash loops after failed self-updates by restarting only on successful updates, stopping early on failed install/build steps, and running openclaw doctor --fix during updates to sanitize config.
Migration and Compatibility
This release contains no breaking changes, but note the following:
- Session Transcript Files: New files are created with
0o600permissions - Cron Schedules: Existing jobs are automatically migrated to stagger settings
- Discord Commands: Numeric IDs are converted to string IDs (doctor auto-repairs)
Contributors
This release includes 100+ PRs from 60+ contributors. Key contributors:
- @tyler6204 - Subagent, BlueBubbles, Reply Threading improvements
- @mbelinky - iOS (Share Extension, Talk Mode, Onboarding)
- @obviyus - Telegram (streaming, button styles, threading)
- @thewilloftheshadow - Discord improvements
- @natedenh - Slack streaming
- @aether-ai-agent - Security fixes (OC-09, Config $include)
Availability
OpenClaw v2026.2.17 is available from the official GitHub releases page:
- GitHub: https://github.com/openclaw/openclaw/releases/tag/v2026.2.17
- Update Command:
openclaw update
Recommendations for Existing Users:
- Back up configuration before updating
- Run
openclaw doctor --fixto validate configuration - Update promptly due to included security fixes (OC-09)
Assessment
This release demonstrates OpenClaw’s maturity with a comprehensive update. Claude Sonnet 4.6 support and 1M context window directly improve long-context reasoning performance. Slack/Telegram enhancements facilitate enterprise adoption, while iOS improvements significantly enhance mobile user experience.
The inclusion of security fixes (OC-09) is particularly important, warranting prompt updates for existing users. Despite containing 100+ changes, the release avoids breaking changes, demonstrating careful attention to backward compatibility.
Source: OpenClaw GitHub Releases
関連記事
OpenClaw v2026.2.17リリース:Claude Sonnet 4.6対応、1Mコンテキスト、Slack/Telegram強化
OpenClawが大規模アップデートをリリース。Claude Sonnet 4.6と1Mコンテキストウィンドウに対応、Slackネイティブストリーミング、Telegramインラインボタンスタイル、iOS Share Extension、セキュリティ修正(OC-09)を含む100件以上の変更を実装。
OpenClaw v2026.2.15リリース - Discord Components v2対応、ネストされたサブエージェント、大規模セキュリティ強化
OpenClawの最新版v2026.2.15がリリース。Discord Components v2による対話型UI、ネストされたサブエージェント機能、SHA-256への移行を含む30件以上のセキュリティ修正を実施。
Claude Sonnet 4.6リリース、Opus級の性能をSonnet価格で実現
AnthropicがClaude Sonnet 4.6を発表。コンピューター使用能力が人間レベルに到達し、1Mトークンコンテキストウィンドウをベータ版で提供。ユーザーの70%が前バージョンより高評価、59%がOpus 4.5を上回る評価。
人気記事
ChatGPT(OpenAI)とClaude(Anthropic)の機能比較 2026年版。コーディング・長文解析・コスト・API料金の違いを検証
ChatGPT(GPT-4o/o3)とClaude(Sonnet 4.6/Opus 4.5)を2026年時点の最新情報で比較する。コーディング能力、長文処理、日本語品質、API料金、無料プランの違いをSWE-benchなどのベンチマーク結果とともに解説する。
【2026年2月20日 所感】「AIがコードを書く」は仮説から現実になった——しかし私たちはその意味をまだ消化できていない
2026年2月20日に観測したコーディングエージェント関連ニュースの総括と所感。Anthropicの自律性研究、cmux、MJ Rathbunのエージェント事故、HN「外骨格 vs チーム」論争、Stripe Minions週1000件PR、Taalas 17k tokens/sec——朝から夜までの流れを通じて見えてきた「AIがコードを書く時代」の実相を考察する。
868のスキルをnpx 1コマンドで——「Antigravity Awesome Skills」が主要AIコーディングエージェントの共通スキル基盤になりつつある
Claude Code・Gemini CLI・Codex CLI・Cursor・GitHub Copilotなど主要AIコーディングアシスタントを横断する868以上のスキルライブラリ「Antigravity Awesome Skills」(v5.4.0)を詳細分析。Anthropic・Vercel・OpenAI・Supabase・Microsoftの公式スキルを統合した設計思想、ロール別バンドル・ワークフロー機能、SKILL.mdによる相互運用性のアーキテクチャを解説する。
最新記事
AIエージェント間通信の標準化競争が始まる——AquaとAgent Semantic Protocolが同日登場
2026年2月23日、Hacker Newsに2つのAIエージェント通信プロジェクトが同日掲載された。Go製CLI「Aqua」とセマンティックルーティングを実装する「Agent Semantic Protocol」は、MCPが解決できないP2P・非同期通信の課題に取り組む。
Claude Sonnet 4.6、無料・Proプランのデフォルトモデルに——社内テストでOpus 4.5を59%の確率で上回る
Anthropicは2026年2月17日にリリースしたClaude Sonnet 4.6を、claude.aiの無料・Proプランのデフォルトモデルに設定した。価格はSonnet 4.5と同額の$3/$15 per 1Mトークン。社内評価ではコーディングエージェント用途でOpus 4.5を上回る結果が出ている。
GoogleがOpenClaw経由のGemini利用ユーザーのアカウントを永久停止——月額$250請求継続のまま
2026年2月23日、Hacker Newsで140pt/107コメントを集めたレポートによると、GoogleはOpenClaw(サードパーティクライアント)経由でGeminiを使用していたGoogle AI Pro/Ultraユーザーを予告なしに永久停止した。技術的・経済的背景を整理する。