Your agents' plan, in plain sight.
Give your agents real work: features, bugs, releases, research. Planban is the live board where they scope it, plan it, build it, and keep it current as they go, so you always see where every project is up to and what needs you.
Every Item, tracked as it happens.
An Item is a feature, a bug, a release, a piece of research. Your agents scope it, plan it, build it, and keep its card current while they do. You see where it is and what needs you, the moment it changes.
- AgentScopedPendingSpec written: purpose, scope, acceptance.
From a note or an issue.
- AgentPlannedUp NextPlan: 3 phases, owner gate after phase 1.
Only when the work needs one.
- AgentBuildingIn ProgressPhase 2 of 3. Runs behind a flag.
The blue edge means an agent is on it now.
- YouNeeds youIn ProgressOwner: try it on a real repo, then ship or revise.
One next action says what to decide.
- YouDoneCompleteShipped in v1.2. Accepted by you.
Only you mark it Complete.
Planban does not write the code or make the plan. Codex and Claude do. Planban is the always-accurate picture of that work, living beside you and your agents, kept true in real time.
Prompt to install,to open.
- 1Ask your agent to set it up
One prompt installs Planban and opens the interactive tutorial, in Codex or Claude.
- 2Open the board from any thread
Type /planban in Codex, or /planban:pb in Claude Code, to bring the board to any thread.
- 3Keep the work structured
Use Items for independent outcomes. Use Groups when several Items contribute to one larger outcome.
Install Planban from piercekearns/planban. Follow the Install With Codex details in the public GitHub README exactly, verify the plugin and MCP tools work, open the interactive tutorial in the Codex in-app browser, then ask whether I want to set up Planban for a local project.
One board for the whole project.
Every outcome is an Item with a Status, a summary, a next action, and its own Spec and Plan. Related Items become a Group whose progress you can read at a glance. Nothing lives only in a chat transcript.
Talk to your agents. The board stays true.
Your agents do the work. Planban keeps the picture true. Agents are trusted editors of the plan; you own intent, priority, and acceptance.
Bring your plans where you work.
Start from repo docs, issues, Notion pages, Linear tickets, Jira work, copied notes, or a plain-language update. Ask your agent to turn that context into draft Planban Items and Groups. Review them before they become your working plan. No native integration is required.
Local and agent-readable.
Structured commands, a local API, MCP tools, and readable files support the same planning model. A drag on the board and a tool call from an agent run the same operation and write the same history.
CLI, API, and MCP
The CLI gives agents stable local commands for setup, Status changes, moving Items into Groups, and documents. The API and MCP expose the same structured actions, so the UI is not the only way to manage the board.
Local state
Repo discovery stays in `.planban/`; live board state stays on your device, readable and writable by you and local agents.
Portable references
Copy Reference puts one line on the clipboard: the Item's title and its planban: id. Paste it into any Codex or Claude thread and the agent resolves the board, Status, Spec, Plan, and next action instead of rediscovering the project.