builderz-labs/mission-control

What it solves

Mission Control provides a centralized, self-hosted control plane for managing multiple AI agents and runtimes. It solves the problem of fragmented visibility when using various agent frameworks, making it difficult to track who owns a task, what was executed, where failures occurred, and how much budget is being spent.

How it works

It acts as a governance layer that sits above agent runtimes rather than replacing their internal reasoning. It uses a local SQLite database to store state and provides a dashboard for dispatching tasks, inspecting runs, and reviewing results. Agents connect via a REST API, CLI, or MCP server to register themselves, claim tasks from a queue, and report heartbeats and results.

Who it’s for

Operators who manage a fleet of AI agents across different runtimes (such as CrewAI, LangGraph, AutoGen, or Claude Code) and need a single place to monitor operations, enforce quality gates, and track costs.

Highlights

• Task Governance: A task board with a pipeline from inbox to completion, including an "Aegis" quality gate for manual approval.
• Multi-Runtime Support: Adapters for various frameworks including OpenClaw, CrewAI, LangGraph, and AutoGen.
• Operational Visibility: Real-time activity streams, cost tracking, token usage views, and cron-based task scheduling.
• Knowledge Management: A memory browser and relationship graph to inspect agent memory and linked session knowledge.
• Extensible Interfaces: Accessible via Web UI, CLI, MCP server, and an OpenAPI-described REST API.