Why Agents Can Edit Tickets at All
Kanban Pro stores every ticket as a plain Markdown file with YAML frontmatter, inside a normal folder on your disk. There is no database, no server and no API in the way. Any AI agent that can read and write files can therefore work the board directly: the same board you see in the app.
Tip
You can open an embedded terminal on any ticket and launch a real agentic CLI. The session is auto-briefed on the ticket, the board and the project rules the moment it starts.
Common Questions
How does an AI agent edit a ticket?
It rewrites the ticket's Markdown file. Changing the status field moves the ticket to another column, updating the assignee reassigns it, ticking a subtask checkbox marks it done, and appended notes become part of the ticket body. Kanban Pro watches the folder and updates the board live as the file changes.
Which AI tools work with Kanban Pro?
Any tool that edits files works. The embedded per-ticket terminals additionally recognise and launch popular agentic CLIs natively, including Claude Code, OpenAI Codex, Cursor Agent, aider, Amp, OpenCode, Antigravity and Goose.
How does an agent know what to work on?
Two mechanisms. The project root carries an auto-generated MAPPING.md index summarising every ticket's status, priority, assignee and dates, so an agent can triage the whole board without opening each file. And a session launched from a ticket receives a generated KP-CONTEXT.md briefing containing the active ticket, its neighbours, the board columns, the team graph and the project rules.
Can I see what an agent changed?
Yes. Every edit carries attribution fields recording who made the change and when. The Activity Feed groups changes per agent and visually distinguishes agent edits from human edits.
Does an agent need a Kanban Pro API key?
No. Kanban Pro has no server and no API, so there is nothing to authenticate against. The agent only needs its own model provider credentials, for example Claude Code signed in to Anthropic.
Can humans and AI agents share one board?
Yes, that is the core design. Agents can join the team graph as named members with a persona and a mandate, their terminal sessions persist on disk for later resume, and humans and agents read and write the same Markdown tickets.
Go Deeper
Two technical guides cover this in detail: Run Claude Code from Kanban Tickets and Kanban as a Memory Layer for AI Agents.