Not another Agent builder — the governance layer for your OpenClaw Agents. Real-time monitoring, cost tracking, human-in-the-loop approvals, and immutable audit trails.
As your OpenClaw Agents grow in number and complexity, operators face unprecedented challenges
Agents run 24/7, but you only discover incidents after damage occurs. No real-time observability.
Token costs grow exponentially with usage. Without budget circuit breakers and anomaly detection, costs spiral unchecked.
Agents autonomously call external tools and APIs. A single unauthorized operation can cause irreversible damage.
Agent configs, model parameters, and permission rules scattered across files and interfaces. No unified view or version control.
From visibility to actionability, from cost control to compliance — all in one place
Real-time high-risk operation monitoring with rich impact assessment cards. One-click approve or deny. Default fail-closed with askFallback=deny.
Real-time feed of all Agent activities. Filter by Agent, type, or time. Daily summaries keep you in the loop.
Multi-dimensional cost attribution, trend analysis, budget circuit breakers, and anomaly detection. Model tiered routing for instant savings.
Three-state health monitoring (healthy / degraded / offline) with 30-second offline alerts. Emergency kill switch with manual recovery confirmation.
Zero-gap config visibility across 6 tabs — Overview, Files, Tools, Skills, Channels, Cron. Diff, versioning, and drift detection.
Append-only audit log with full-text search and trace_id linking. Every operation traceable to a person.
Expo/React Native mobile app for approvals, monitoring, budgets, and kill switch workflows from anywhere.
Full CRUD for scheduled jobs with consecutive failure alerts and execution history. Visual scheduling configuration.
Connect your OpenClaw instance in 5 minutes. Gain full visibility instantly.
Securely connect your OpenClaw instance via WebSocket. All Agents auto-discovered.
Real-time activity feed, cost analytics, and health monitoring — ready instantly.
Configure approval rules, set budget circuit breakers, define alert policies. From reactive to proactive.
ClawButler is a fully LLM-driven product. You can configure everything manually on the Web — but the recommended way is: let your AI read the docs, and it handles the rest.
Structured project docs at /llms.txt, optimized for LLM comprehension. One read, full understanding — your AI knows every API, every CLI command, every config option.
Every governance capability is a scriptable CLI command. Connect agents, set budget breakers, deploy templates, configure approval rules — your AI operates it all, autonomously.
The Web becomes your real-time monitoring cockpit. Watch what AI configured, review audit trails, catch anomaly alerts. Intervene only when you choose to.
What we shipped — and what it enables you to do
Agent auto-discovery, HITL approvals, live activity feed, cost tracking with budget breakers, audit trail, kill switch
Config Safety V2 (diff + drift + rollback), Templates, Runbooks, CLI (60+ commands), Federation, One-click install
Trusted Change Control, supply chain scanning, per-agent cost accuracy, idempotency, edge agent V2
Workflow orchestration + topology DAG, session monitor, approval bridge, zero-spinner performance, 28-page user manual