Last updated: 2026-04-03 UTC · Maintained by: Koda & Nexus
Team chat integration — all agents connected as individual bots.
Accounts: @koda (main), @catalyst (business), @nexus (technocrat), @atlas (allrounder), @leika-builder, @leika-creator
Config: channels.mattermost — baseUrl: http://prod.lmphq.net:8065
Chatmode: oncall — agents respond when @mentioned or DM'd
Features: Reactions ✅ · Slash commands ✅ · Config writes ✅ · Streaming ✅
// Enable in openclaw.json
"plugins": { "entries": { "mattermost": { "enabled": true } } }
"channels": { "mattermost": { "enabled": true, "baseUrl": "http://your-mm:8065", "accounts": { ... } } }
OpenTelemetry export — ships gateway metrics and logs to Loki/Grafana.
Endpoint: http://10.1.2.130:3100/otlp · Protocol: http/protobuf
Signals: Metrics ✅ · Logs ✅ · Traces ❌ (disabled to reduce volume)
Dashboards: Gateway Logs · Token & Inference · Claude & Gemini Tokens
"diagnostics": { "otel": { "enabled": true, "endpoint": "http://10.1.2.130:3100/otlp", "protocol": "http/protobuf", "serviceName": "openclaw-gateway" } }
Agent Communication Protocol runtime — enables spawning coding agents (Claude Code, Codex, etc.) as isolated sessions.
Status: Plugin loaded and operational.
Usage: Agents use sessions_spawn(runtime="acp") to delegate coding tasks.
"acp": { "enabled": true, "dispatch": { "enabled": true } }
OpenClaw internal service plugin — handles core platform features.
Handles OAuth token management for Qwen models via Alibaba Cloud portal.
Phase 2 of LMP Memory Architecture — automatically extracts facts, decisions, preferences, corrections, and entities from agent conversations after each turn.
Hook: message_sent (fires independently of sessionKey)
Extraction: 5 categories via Nano-30B. Confidence floor: 0.75 general, 0.95 sensitive.
Timeout: 30s per extraction cycle
Output: auto-capture-{agentId}.md in technocrat workspace memory dir
Status: Operational. Cross-session recall verified. Quality tuning (P2.5) pending — confidence scores cluster at 0.92, skip-list needed for ephemeral noise.
Matrix/Element chat integration — 4 accounts configured (@koda, @markus, @nexus, @atlas on matrix.lmphq.net).
Issue: ESOCKETTIMEDOUT on sync — likely stale access tokens or encryption key setup. Blocks Mattermost when enabled.
Fix needed: Refresh Matrix access tokens, verify E2EE key backup, test with encryption: false first.
"plugins": { "entries": { "matrix": { "enabled": true } } } // After fixing tokens
OAuth helper for Gemini CLI. Not needed since we use API key auth for Gemini search.
| Plugin | Purpose | Setup |
|---|---|---|
| discord | Discord bot integration | channels.discord + bot token |
| telegram | Telegram bot | channels.telegram + BotFather token |
| signal | Signal messenger via linked device | channels.signal + signal-cli |
| WhatsApp via linked device | channels.whatsapp + QR pairing | |
| slack | Slack workspace bot | channels.slack + Slack app token |
| device-pair | Mobile companion app pairing (iOS/Android) | plugins.entries.device-pair |
| memory-core | Enhanced memory with vector store | Already using via memorySearch config |
| memory-lancedb | LanceDB vector memory backend | Alternative to default memory store |
| thread-ownership | Per-thread agent ownership in group channels | plugins.entries.thread-ownership |
| talk-voice | Voice conversations | Needs TTS + STT setup |
| voice-call | Phone-style voice calls | Needs telephony provider |
| ollama | Ollama auto-discovery & model pull | Already configured manually |
| vllm | vLLM endpoint auto-discovery | Already configured manually |
| open-prose | Long-form writing assistant | plugins.entries.open-prose |
| llm-task | Background LLM task runner | plugins.entries.llm-task |
| copilot-proxy | GitHub Copilot proxy | plugins.entries.copilot-proxy |
| diffs | Code diff handling | plugins.entries.diffs |
| Also available: bluebubbles · feishu · googlechat · imessage · irc · line · msteams · nextcloud-talk · nostr · sglang · synology-chat · tlon · twitch · zalo · zalouser · minimax-portal-auth | ||
1. Add to config:
// In ~/.openclaw/openclaw.json → plugins.entries
"plugins": {
"entries": {
"plugin-name": { "enabled": true }
}
}
2. Configure the channel (for messaging plugins):
// Add channel config with accounts + tokens
"channels": {
"discord": { "enabled": true, "accounts": { "bot": { "botToken": "...", "agent": "main" } } }
}
3. Restart gateway:
systemctl restart claw # Or from chat: /restart
Or use the CLI:
openclaw doctor --fix # Interactive setup wizard
OpenClaw · Host: openclaw (10.1.2.70) · 6 agents · 6 active plugins · 178 skills