Mission Control

Mission Control is an open-source, self-hosted dashboard for AI agent orchestration built with Next.js 16 and TypeScript. It enables teams to manage AI agent fleets, dispatch tasks, track token costs, and coordinate multi-agent workflows from a single interface — powered by SQLite with no Redis or Postgres required. The platform ships with 32 panels covering tasks, agents, skills, logs, memory, security, cron scheduling, alerts, webhooks, and pipelines, all updated in real time via WebSocket and SSE.

Key Features:

• Agent Management — Monitor agent status, configure models, view heartbeats, and manage the full agent lifecycle from registration to retirement, including local agent discovery from common directories.
• Task Board — Kanban board with six columns, drag-and-drop, priority levels, threaded comments, inline sub-agent spawning, and multi-project support with per-project ticket prefixes.
• Memory Knowledge Graph — Explore agent knowledge through a Memory Browser, filesystem-backed memory tree, and interactive relationship graph for sessions, memory chunks, and linked knowledge files.
• Skills Hub — Browse, install, and security-scan agent skills from ClawdHub and skills.sh registries, with built-in checks for prompt injection, credential leaks, and dangerous shell commands.
• Cost Tracking — Token usage dashboard with per-model breakdowns, trend charts, and session-level cost analysis powered by Recharts.
• Security Audit & Agent Trust — Real-time posture scoring (0–100), secret detection, MCP tool call auditing, injection attempt tracking, and per-agent trust scores with configurable hook profiles.
• Agent Eval Framework — Four-layer evaluation covering output evals, trace evals, component evals, and drift detection against a 4-week rolling baseline.
• Natural Language Scheduling — Create recurring tasks with natural language expressions like "every morning at 9am"; the built-in parser converts them to cron and spawns dated child tasks automatically.
• Claude Code Integration — Read-only bridge surfaces Claude Code team tasks, sessions, token usage, and configs directly on the dashboard.
• Framework Adapters — Built-in adapters for OpenClaw, CrewAI, LangGraph, AutoGen, Claude SDK, and a generic fallback normalize registration, heartbeats, and task reporting to a common interface.
• Role-Based Access Control — Viewer, operator, and admin roles with session and API key auth, plus Google Sign-In with admin approval workflow.
• Zero External Dependencies — SQLite-backed stack runs with a single pnpm start; Docker Compose and hardened production profiles are also provided.

To get started, clone the repository and run bash install.sh --local or docker compose up. Open http://localhost:3000/setup to create your admin account, then register your first agent via the UI or REST API. Full documentation, an OpenAPI 3.1 spec with 101 endpoints, and an interactive Scalar UI at /api-docs are included.