Kanban board for planning work and running coding agents like Claude Code and Codex in isolated workspaces, with diff review and PR creation.
- Stars
- 28,009
- Forks
- 1,895
- Open issues
- 1,507
- Last commit
- yesterday
What is it?
Beads (the `bd` CLI) is an issue tracker built for AI coding agents. Instead of keeping plans in loose markdown files, it stores tasks as a dependency-aware graph in a version-controlled SQL database. Agents can ask for work that has no open blockers, claim a task atomically, and close it so that dependent tasks become ready, which helps them keep track of long-running work without losing context.
The storage layer is Dolt, a SQL database with Git-style branching and merging. Tasks get hash-based IDs such as `bd-a1b2` to avoid collisions when several agents or branches create issues in parallel. Hierarchical IDs support epics, sub-tasks and graph links like `relates-to`, `duplicates` and `supersedes`. Cross-machine sync runs through `bd dolt push` and `bd dolt pull` against a git remote, and the tool can run embedded or against an external Dolt server.
`bd init` sets up a project and can install integrations for Claude Code, Codex, Cursor and other agents. The CLI also supports a stealth mode that works without git, and an experimental HTTP backend for connecting to a shared `bd serve` instance.
Key features
- 01Stores tasks as a dependency graph with `bd ready` listing unblocked work.
- 02Atomic task claiming with `bd update --claim` for multi-agent workflows.
- 03Hash-based IDs and Dolt branching reduce merge conflicts across branches.
- 04Hierarchical IDs for epics and sub-tasks, plus graph link types.
- 05Semantic compaction summarizes old closed tasks to save context window.
- 06Git-free mode and agent integrations for Claude Code, Codex and Cursor.
Use cases
- Giving a coding agent persistent task memory across long, multi-step projects.
- Coordinating several agents working on the same repository on different branches.
- Tracking planning issues separately when contributing to open-source projects.
- Running isolated task tracking in CI or in a non-git version control setup.
Quick start
- 1
Install the bd CLI system-wide
Install with the official instructions in the README - 2
Go to your project directory
cd your-project - 3
Initialize beads in the project
bd init
Things to know
- Requires a Dolt-backed database, adding a dependency your team must understand.
- Several HTTP backend behaviors diverge from local mode and are documented as experimental.
- Install script and binary verification steps are needed for secure manual installs.
Topics
FAQ
- What is beads?
- A distributed, Dolt-backed graph issue tracker that gives coding agents persistent, dependency-aware task memory via the bd CLI. Beads (the `bd` CLI) is an issue tracker built for AI coding agents. Instead of keeping plans in loose markdown files, it stores tasks as a dependency-aware graph in a version-controlled SQL database. Agents can ask for work that has no open blockers, claim a task atomically, and close it so that dependent tasks become ready, which helps them keep track of long-running work without losing context.
- Is beads free to use? What license does it have?
- beads is open source under the MIT license. Check the license terms before using it in commercial products.
- Is beads still maintained?
- Actively maintained. The last commit to beads was on Oct 19, 2026.
- What can I use beads for?
- Giving a coding agent persistent task memory across long, multi-step projects. Coordinating several agents working on the same repository on different branches. Tracking planning issues separately when contributing to open-source projects. Running isolated task tracking in CI or in a non-git version control setup.
- How do I install beads?
- Install the bd CLI system-wide Go to your project directory: cd your-project Initialize beads in the project: bd init
- Where can I get beads?
- The source code and installation instructions are on GitHub: https://github.com/gastownhall/beads
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Desktop, mobile, web and CLI app for running and orchestrating multiple coding agents such as Claude Code, Codex and OpenCode in parallel worktrees.
A local browser-based review surface that lets you annotate agent plans, markdown, HTML and code diffs, then send feedback back to coding agents.
A macOS app for browsing, editing, and organizing AI coding agent skills and agents across Claude Code, Cursor, Codex, Windsurf, and Amp.
Persistent memory for AI coding agents that captures sessions automatically and injects relevant context into later sessions across Claude Code, Cursor, Codex and other MCP clients.
A collection of 67 specialized skill packs that turn Claude Code into a pair programmer for full-stack development.