OpenClaw-Observer-for-Claude-CoWork

See your entire OpenClaw deployment at a glance — agent health, channel status, token burn, session activity, and error trends in one interactive dashboard. No dependencies, no setup, just ask Claude and get a visual report in seconds.

Running a multi-agent OpenClaw deployment means juggling 10+ agents, multiple messaging channels, cron jobs, and token budgets — with no single place to see what's happening. You end up running openclaw status and openclaw health repeatedly, parsing JSON in your head, and missing problems until they escalate. OpenClaw Observer turns three CLI commands into a visual command center.

Generate an interactive HTML dashboard — Collects data from openclaw health, openclaw status --deep, and openclaw sessions --all-agents, then renders a self-contained dark-themed dashboard with zero external dependencies.

Monitor channel health at a glance — Color-coded cards for every configured channel (Telegram, Discord, iMessage, Slack, WhatsApp). Green = running, yellow = configured but stopped, red = error with details.

Track token usage across all agents — Bar charts showing which agents are consuming the most tokens, helping you spot cost concerns before they spiral. Includes model distribution breakdown (Claude, GPT, Qwen, etc.).

Surface session issues automatically — Filterable session table highlighting aborted runs, stale sessions, and context window limits. Agent dropdown lets you drill into specific agents.

Interpret results with actionable flags — After generating the dashboard, Claude proactively flags red/yellow/green patterns: channels configured but not running, agents with disabled heartbeats, high token consumers, and aborted sessions.

git clone https://github.com/zurbrick/OpenClaw-Observer-for-Claude-CoWork.git ~/.claude/skills/openclaw-observer

The ski