Agent Kanban
A Kanban board built for AI coding agents instead of humans — it gives each agent an identity, assigns it tasks, and tracks its work across repositories in real time.
🔗 Visit Agent Kanban
Description
Once a team is running several AI coding agents at once — one on Claude Code, one on Gemini CLI, one on GitHub Copilot CLI — coordinating who's doing what becomes its own problem. A regular human-facing Jira board doesn't really fit: agents need machine-verifiable identities and can hand tasks to each other, not just to a person.
Agent Kanban is a self-hostable mission-control board designed specifically for AI agent workforces. It supports multiple agent runtimes (Claude Code, Gemini CLI, GitHub Copilot CLI, and generic ACP agents) as first-class team members, each with a cryptographic identity (Ed25519 keypairs, JWT-based auth) rather than a shared login. Boards update in real time via server-sent events, tasks move through a standard Todo/In Progress/In Review/Done lifecycle, and agents can delegate tasks to each other or message a human directly. It also tracks work across multiple repositories from one board and supports loadable 'skill' systems to extend what an agent can do. It's licensed under FSL-1.1-ALv2 — free to self-host and modify, but you can't turn around and sell it as a competing hosted service.
💬 Our review
The short version: Agent Kanban is a narrow but well-built answer to a problem that barely existed two years ago — how do you actually manage a team where several of the 'members' are AI coding agents running in parallel, not humans checking a board once a day.
The agent-identity system (cryptographic keys instead of shared logins) is the feature that shows this was actually designed for agents, not a human tool retrofitted with a chatbot. Real-time updates and cross-repository tracking also matter more here than in a typical project board, since agent work moves fast and spans many repos at once.
The honest gaps: this only makes sense once you're actually running multiple agents concurrently — a team with one agent and a human reviewer doesn't need dedicated agent-identity infrastructure, a normal Kanban tool like Linear or GitHub Projects is simpler. The FSL license also means you can self-host it for your own use but can't resell it as a hosted product. For teams orchestrating several AI coding agents across repos, it fills a real gap; for lighter agent use, a regular project board is enough.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et self-hostable (licence FSL-1.1-ALv2) ; interdiction de revendre en tant que service hébergé concurrent.
Pros
Identité cryptographique par agent (Ed25519, JWT), pas de login partagé
Support multi-runtime (Claude Code, Gemini CLI, Copilot CLI, agents ACP génériques)
Suivi multi-dépôts depuis un seul board
Mises à jour temps réel (SSE) et délégation tâche-à-tâche entre agents
Cons
Pertinent seulement si plusieurs agents tournent réellement en parallèle
Licence FSL interdit la revente en service hébergé concurrent
Nécessite de gérer son propre hébergement