How Squad runs coordinated AI agents inside your repository

Squad is an open source project built on GitHub Copilot that initializes a preconfigured multi-agent AI team directly inside a repository with two commands. It uses a coordinator agent to route tasks to specialists (backend, frontend, tester, docs) that run in parallel with independent context windows. Key architectural patterns include a 'drop-box' pattern using a versioned decisions.md file for shared memory, context replication across agents rather than splitting, and explicit versioned memory stored in .squad/ folder files. The reviewer protocol prevents the original agent from revising its own rejected work, forcing genuine independent review. The goal is to make agentic development accessible without heavy orchestration infrastructure or deep prompt engineering expertise.