Skip to content

Loom Documentation

"From a single thread of an idea, we weave complete software."

Loom is an autonomous AI agent orchestration platform that coordinates multiple specialized AI agents to build, maintain, and operate software projects.

What Can Loom Do?

  • Orchestrate AI agents with specialized roles (PM, Engineer, QA, DevOps, Designer, Code Reviewer)
  • Manage distributed work through beads (work items) with dependencies and priority
  • Autonomously fix bugs from detection through investigation, fix, verification, and PR creation
  • Execute reliable workflows with human-in-the-loop approval gates
  • Bootstrap entire projects from a Product Requirements Document (PRD)
  • Monitor everything with Prometheus metrics, Jaeger tracing, and Loki logging

Quick Navigation

I want to... Go to...
Get Loom running in 10 minutes Quick Start
Learn how to use the web UI User Guide
Configure and deploy Loom Administrator Guide
Understand the architecture or contribute Developer Guide
Look up API endpoints or config options Reference

Architecture at a Glance

flowchart LR
    UI[Web UI] --> API[Control Plane API]
    API --> D[Dispatcher]
    D --> NATS[NATS JetStream]
    NATS --> A1[Agent: Coder]
    NATS --> A2[Agent: Reviewer]
    NATS --> A3[Agent: QA]
    API --> PG[(PostgreSQL)]
    API --> CS[Connectors Service]
    CS --> P[Prometheus]
    CS --> J[Jaeger]
    CS --> G[Grafana]

Project Status

Loom is self-maintaining — it uses its own agent pipeline to fix bugs, review code, and ship improvements. All five microservices architecture phases are complete.