Keep an AI coding agent organized across multiple sessions on a complex project without losing context.
Track task dependencies so an agent knows which work is safe to start and which tasks are blocked.
Store discoveries and insights that automatically surface in future agent sessions.
Manage multi-step coding projects where an agent needs to remember progress and decisions.
Requires Dolt installation and understanding of SQL schema setup for the version-controlled database.
Beads is a memory and task-tracking system designed specifically for AI coding agents, tools like Claude Code, Codex, or similar automated assistants that write code on your behalf. The core problem it solves is that these agents tend to lose track of where they are on long or complex tasks, especially when work spans many steps or multiple agent sessions. Instead of leaving the agent to maintain a messy markdown to-do list in its memory, Beads provides a structured "issue tracker" built on a version-controlled database called Dolt (which works like Git, but for databases with SQL tables). Each task gets an ID, can have dependencies on other tasks, and tracks status as work progresses. The agent uses simple command-line commands to claim a task, mark it done, store insights for later, and check what's ready to work on next. Tasks can be linked, one task blocks another, so the agent always knows what it can safely start. A "memory" feature lets the agent record discoveries that get surfaced automatically in future sessions. You would use Beads when you're running an AI agent on a multi-step coding project and want it to stay organized across sessions without losing context. It works on macOS, Linux, Windows, and FreeBSD, and can be installed via a shell script or package manager. It is written in Go.
Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.