OpenClaw v2026.2.15 Released: Discord Components v2 and Nested Sub-Agents
OpenClaw releases v2026.2.15 one day after v2026.2.14, adding Discord interactive components, nested sub-agents with configurable depth, plugin hook enhancements, and continued security hardening.
OpenClaw has released version 2026.2.15 just one day after v2026.2.14, demonstrating the project’s rapid development pace. The release introduces Discord Components v2 for rich interactive prompts, nested sub-agent capabilities, plugin hook enhancements, and continues the security hardening effort from the previous release.
Major New Features
Discord Components v2: Rich Interactive UI Discord users can now interact with agents through native Discord UI components including buttons, selects, modals, and attachment-backed file blocks. This enables more sophisticated interaction patterns beyond plain text, with refined exec approval UX and embed passthrough support. Thanks to @thewilloftheshadow for implementing this extensive feature.
Nested Sub-Agents (Sub-Sub-Agents)
OpenClaw now supports sub-agents spawning their own children with configurable depth. Operators can set agents.defaults.subagents.maxSpawnDepth: 2 to allow nested sub-agents. The implementation includes:
maxChildrenPerAgentlimit (default 5) to prevent runaway spawning- Depth-aware tool policy
- Proper announce chain routing across multiple levels
This feature enables more complex hierarchical agent workflows. Thanks to @tyler6204.
Plugin Hook Enhancements
Extensions can now observe prompt/input context and model output usage details through exposed llm_input and llm_output hook payloads. This provides deeper visibility into agent operations for monitoring and analytics purposes. Thanks to @SecondThread.
Per-Channel Ack Reaction Overrides Slack, Discord, and Telegram now support per-channel acknowledgment reaction overrides at account/channel level to accommodate platform-specific emoji formats. Thanks to @zerone0x.
Cron Webhook Enhancements
New notify toggle for finished-run webhook delivery and dedicated webhook auth token support (cron.webhookToken) for outbound cron webhook posts. Thanks to @advaitpaliwal.
Security Improvements
Sandbox Configuration Hardening
- Replaced deprecated SHA-1 with SHA-256 for sandbox configuration hashing, ensuring deterministic sandbox cache identity
- Blocked dangerous sandbox Docker configs (bind mounts, host networking, unconfined seccomp/apparmor) to prevent container escape via config injection
- Preserved array order in config hashing for correct container recreation triggers
Logging and Secret Protection
- Redacted Telegram bot tokens from error messages and stack traces to prevent accidental secret leakage
- Redacted sensitive session/path details from status responses for non-admin clients (full details remain available to
operator.admin)
Input Sanitization
- Hardened chat.send inbound message handling by rejecting null bytes, stripping unsafe control characters, and normalizing Unicode to NFC
- Sanitized workspace paths before embedding into LLM prompts to prevent instruction injection via malicious directory names
- Prevented stored XSS via assistant name/avatar in Control UI
Skills and Download Security
- Restricted download installer targetDir to per-skill tools directory to prevent arbitrary file writes
- Capped downloaded response body size before HTML parsing to prevent memory exhaustion
Group Chat and Context Improvements
Persistent Group Chat Awareness The system now injects group chat context (name, participants, reply guidance) into the system prompt on every turn, not just the first. This prevents the model from losing awareness of which group it’s in and incorrectly using the message tool to send to the same group. Thanks to @tyler6204.
Memory and Context Enhancements
- Made
buildFtsQueryUnicode-aware so non-ASCII queries (including CJK) produce keyword tokens instead of falling back to vector-only search - Resolved
memory/YYYY-MM-DD.mdplaceholders with timezone-aware runtime dates - Applied configured model contextWindow overrides after provider discovery
Channel-Specific Fixes
Telegram:
- Omitted
message_thread_idfor DM sends/draft previews to prevent “message thread not found” errors - Replaced inbound placeholder with successful preflight voice transcript in message body context
- Implemented retry logic for inbound media getFile calls (3 attempts with backoff)
- Finalized streaming preview replies in place instead of sending duplicate final messages
- Disabled block streaming when
channels.telegram.streamModeis off
Discord:
- Preserved channel session continuity when runtime payloads omit
message.channelId - Deduped native skill commands by skill name in multi-agent setups
- Ensured role allowlist matching uses raw role IDs
LINE:
- Failed closed on webhook startup when channel token or secret is missing
- Treated LINE accounts as configured only when both credentials are present
Agent and Subagent Improvements
Timeout Handling
- Returns explicit timeout error reply when embedded runs timeout before producing payloads
- Prevents silent dropped turns during slow cache-refresh transitions
Browser Control
- When browser control service is unavailable, returns explicit non-retry guidance so models don’t loop on repeated browser tool calls until timeout
Subagent Stability
- Used child-run-based deterministic announce idempotency keys to prevent duplicate announce retries
- Preserved
agents.defaults.model.fallbackswhen subagent sessions carry model overrides
OpenAI Integration
- Forced
store=truefor direct OpenAI Responses/Codex runs to preserve multi-turn server-side conversation state
Web UI and TUI Improvements
Web UI:
- Hid BOOTSTRAP.md in Agents Files list after onboarding completion
- Fixed NO_REPLY placeholder display when messaging tool sends succeeded
Memory/QMD:
- Scoped managed collection names per agent to prevent cross-agent collection clobbering
- Precreated glob-backed collection directories before registration
Community Contributions
This release includes contributions from over 30 community members, with notable contributions from:
- @thewilloftheshadow (Discord Components v2)
- @tyler6204 (Nested sub-agents, group chat context)
- @SecondThread (Plugin hooks)
- @aether-ai-agent (Multiple security improvements)
- @fr33d3m0n (Security hardening)
Rapid Release Cadence
The release of v2026.2.15 just one day after v2026.2.14 demonstrates OpenClaw’s commitment to continuous improvement and responsiveness to community needs. The project maintains a balance between introducing new features (Discord Components v2, nested sub-agents) and ensuring platform stability through ongoing security hardening and bug fixes.
Source: OpenClaw GitHub Releases
Related Articles
OpenClaw v2026.2.14 Released: Major Security Hardening and 100+ Bug Fixes
OpenClaw releases v2026.2.14 with extensive security improvements, TUI stability enhancements, memory system optimizations, and 100+ bug fixes across channels, agents, and tools.
Claude Code's 28 Official Plugins Revealed - Undocumented Feature Extensions
Reddit user discovers 28 official Claude Code plugins, most undocumented. Includes TypeScript LSP, security scanning, context7 documentation search, and Playwright automation.
Kimi Integrates OpenClaw Natively - 5,000+ Community Skills and 40GB Cloud Storage in Browser
Moonshot AI's Kimi.com now supports OpenClaw natively in browser tabs, offering 24/7 uptime, ClawHub access with 5,000+ skills, 40GB cloud storage, and pro-grade search capabilities.
Popular Articles
868 Agentic Skills, One Command: Antigravity Awesome Skills Becomes the Cross-Tool Skill Standard
Antigravity Awesome Skills (v5.4.0) delivers 868+ battle-tested skills for Claude Code, Gemini CLI, Codex CLI, Cursor, GitHub Copilot, and five other AI coding assistants via a single npx command. With official skills from Anthropic, Vercel, OpenAI, Supabase, and Microsoft consolidated under one MIT-licensed repository, it's emerging as the portable skill layer for the fragmented AI coding agent landscape.
How Claude Sonnet 4.6 Agent Teams Achieve 4x Productivity: Practical Insights from Anthropic's Own Research
Two Anthropic studies—a survey of 132 internal engineers and an analysis of 1M+ real-world agent interactions—reveal the precise delegation strategies and autonomy patterns that enable high-performing teams to multiply output with Claude Sonnet 4.6 agent teams.
What Actually Makes OpenClaw Special: The Full Story from VibeTunnel to 200k+ GitHub Stars
The three-stage VibeTunnel→Clawdbot→OpenClaw evolution, Pi runtime philosophy, why HEARTBEAT is the real differentiator from Claude Code, and the ClawHub supply chain attack (12% of skills were malicious). An unvarnished look at the most used and most misunderstood OSS agent.
Latest Articles
Two AI Agent Communication Projects Hit Hacker News Simultaneously, Targeting MCP's Blind Spots
Aqua and Agent Semantic Protocol appeared on Hacker News on the same day, both tackling the same unsolved problem: how AI agents communicate directly without a central broker, across network boundaries, and asynchronously.
Claude Sonnet 4.6 Becomes the Default for Free and Pro Users — Outperforms Opus 4.5 on Coding Agent Benchmarks
Anthropic has made Claude Sonnet 4.6 the default model for claude.ai's Free and Pro plans. Released February 17, 2026, it matches Sonnet 4.5 pricing at $3/$15 per million tokens while internal Claude Code evaluations show it beating the previous frontier model, Opus 4.5, 59% of the time on agentic coding tasks.
Google Permanently Bans AI Pro Users for Accessing Gemini via OpenClaw, Continues Charging $250/Month
A Hacker News post garnering 140 points and 107 comments details how Google terminated Google AI Pro and Ultra accounts without warning after users accessed Gemini through OpenClaw, a third-party client. The incident surfaces deeper issues around prompt caching, subscription economics, and how AI providers enforce terms of service.