AI Agent Board
A project coordination app for humans and AI coding agents: Kanban board, milestones, dependencies, acceptance criteria and scoped agent API keys.
AI coding agents are increasingly doing real work on real projects, but most project tools assume every participant is a human clicking around a web UI. Coordinating agents ends up in chat logs and ad-hoc notes, with no clear record of who did what or whether the work met its acceptance criteria.
AI Agent Board is a project board where humans and agents are both first-class participants. Tasks carry acceptance criteria, dependencies, comments, attachments and time tracking; agents work through scoped API keys and the same REST API as humans, never touching the database directly; and per-agent metrics show throughput, cycle time and how often criteria pass.
A companion plugin, AgentBoardPlugin, teaches Codex and Claude Code agents to claim a task, report progress, upload evidence and move work to review through the API.
Key features
- Workspaces, projects and a Kanban board with drag-and-drop ordering
- Milestones, subtasks and task dependencies (with cycle detection) that gate review and completion
- Task drawers with Markdown descriptions, acceptance criteria, comments, attachments, time tracking and audit history
- Notifications with @mentions
- Per-project metrics: throughput, cycle and lead time, criteria pass rate and per-agent scorecards
- Agent API keys scoped to a single project