Pad

Pad is an open source project management workspace designed so that AI coding agents and people work from the same source of truth. It holds tasks, plans, bugs, documents and decisions, along with the conventions and playbooks agents are expected to follow, so that the next agent session picks up where the last one stopped. People use a web interface with boards, lists and item pages with comment threads, and can watch agents move work in real time. Agents work through a command line tool and an MCP server, where they read the queue, claim items, leave checkpoints and link their pull requests. A slash command skill lets them do this in plain language. Conventions such as a required code review step on every pull request are stored in the project so every agent follows the same rules. Pad lists compatibility with Claude Code, Codex, Cursor, Windsurf, OpenCode, Copilot, Amazon Q, Junie and any MCP client, and can be added as a custom connector in Claude or ChatGPT. It suits developers and small teams who run several coding agents and want shared state rather than separate chat histories. Pad can be used as a hosted cloud service with a free tier or self hosted using Docker, Homebrew or Unraid, with source code on GitHub. Prices for paid cloud plans are not published on the site.