ClawButler Documentation
The OpenClaw-first governance platform for agent fleets and local runtimes — monitor, approve, and control with full audit trails.
What is ClawButler?#
Organizations now run a mix of OpenClaw gateways and local coding runtimes. Costs grow without visibility, high-risk operations happen outside review, and local Hermes / Claude Code / Codex sessions remain operationally fragmented.
ClawButler is the governance layer across that mix. It stays OpenClaw-first, while also pairing runtime hosts that can discover Hermes, Claude Code, Codex, and OpenClaw instances, ingest selected runtimes as connectors, and project attach-mode sessions into the same control plane.
Built on open standards — MCP (Model Context Protocol, by Anthropic) for connecting agents to tools and data, A2A (Agent2Agent Protocol, by Google) for agent-to-agent collaboration — ClawButler is fully open-source. OpenClaw connectors remain the fullest supported path today; runtime-host integrations expand visibility without claiming full provider-native parity.
Core Concepts#
Agent Registry
A unified identity layer where OpenClaw connectors and runtime-host-discovered instances are registered, discovered, and managed. It supports enable or disable controls, structured configuration tabs, and health tracking from one control plane.
MCP (Model Context Protocol)
An open standard by Anthropic for connecting AI applications to external data sources and tools. ClawButler uses MCP to auto-discover Agents and their capabilities — tools, files, databases — with unified authentication and access control.
A2A (Agent2Agent Protocol)
An open protocol by Google for agent-to-agent communication and interoperability. A2A complements MCP: while MCP connects Agents to tools, A2A enables Agents to delegate tasks and share context with each other — with full authorization and audit trails.
HITL Approvals
Human-in-the-loop workflows intercept high-risk operations before execution. Configurable rules with risk-level assessment, cost estimation, and batch processing — approve once or allow-always for trusted patterns.
Audit Trail
Append-only event log with monotonic sequence IDs captures every agent action. Trace linking connects related operations across agents, and PII masking ensures compliance while maintaining accountability.
How It Works#
Connect
Connect an OpenClaw gateway or pair a Runtime Host. ClawButler can discover OpenClaw plus Hermes / Claude Code / Codex instances and register only the ones you choose to govern.
Monitor
Track every connector and projected session through live SSE streaming, cost analytics, and health monitoring. Attach-mode session sync gives local-runtime investigations a shared timeline.
Govern
Set approval rules, budgets, and monitoring. OpenClaw remains the fullest control path; runtime-host integrations currently focus on discovery, connector intake, and attach-mode session sync.
Tech Stack#
| Layer | Technology |
|---|---|
| Web | Next.js 15 (App Router) + React 19 + Tailwind CSS 4 + Auth.js v5 |
| API | FastAPI + Python 3.12 + SQLAlchemy 2 (async) + Alembic + structlog |
| Database | PostgreSQL 17 + Valkey 8 (Redis-compatible) |
| CI | GitHub Actions (lint / typecheck / test / build) |
| Deploy | Docker Compose + Caddy reverse proxy (auto HTTPS) |
Quick Start#
Cloud
Sign in to Cloud, pair a Runtime Host, discover local runtimes, and start governing them without deploying extra control-plane infrastructure.
Start FreeSelf-Host
Deploy on your own infrastructure, then connect OpenClaw gateways or pair Edge Agents / Runtime Hosts for discovery and governance.
Self-Host GuideReady to govern OpenClaw and local runtimes?
Start free, pair a runtime host in minutes, or self-host for full data privacy and operator control.