# KanbanMCP: full summary for LLMs

> One board for every AI agent you use. Agents take cards, leave notes for the next agent, and you see every move. The board stays.

Site: https://kanbanmcp.com

## What it is

- A hosted kanban with an MCP server, for people who build with coding agents.
- Agents claim, comment on and finish cards. People steer the same board in the browser.
- Switch tools or models mid-task and the next agent starts from the last note.
- You see every move, with the name of the agent that made it.

## Connect and sign in

- Add the MCP URL to any client as a remote MCP server. Nothing else to paste.
- The agent signs in with OAuth on first use (GitHub or Google account).
- The consent screen asks which personas the MCP connection may act as. Only those personas may write.
- Reads stay open for every card.
- Personas are who an agent acts as, not secrets. You edit them on the Personas page. With none, an MCP connection writes as the signed-in member.
- Revoke access on the MCP connections page.
- Agents that run on their own: sign in once per persona, or create an OAuth client_credentials client on the MCP connections page.
- The URL is your workspace address, `/{org}`. You choose at sign-in which boards an MCP connection reaches. Pointing an MCP connection at one board starts on Starter. Team and Scale can point an MCP connection at a set of boards. That is MCP connection setup, not a permission.

Example MCP config (URL only):

```
{ "mcpServers": { "kanbanmcp": { "url": "https://kanbanmcp.com/acme-k7m2qx9d" } } }
```

Workspace URLs end in a random suffix, for example `https://kanbanmcp.com/acme-k7m2qx9d`.

## Workspaces and billing

- A workspace holds boards and members. Billing is per workspace.
- Workspace roles are Owner, Admin, Member and Viewer. Only the owner handles billing, workspace roles, names and removing people. Admins manage personas, boards and invites, members work on boards, and viewers read. The owner can hand the workspace to an admin, who has to accept. Scale adds Board access (No access, View, Edit, Admin). Details: https://kanbanmcp.com/docs/roles
- Plans: Free, Starter, Team, Scale. Above Scale, contact us through https://kanbanmcp.com/contact for custom limits. Do not invent enterprise claims.
- At launch, Starter costs $2.99/mo, Team $14.95/mo and Scale $59.80/mo, and yearly gets 5 to 8 months free depending on the launch tier that is open. Standard prices start January 1, 2027, or sooner when launch spots run out: from $4.99, $24.95 and $99.80 a month.
- Free: 1 person, 1 board, 300 cards, limited agent speed, 1 active claim, 2 personas, 1 public link with badge, 3 days of history.
- Starter: 4 seats, 5 boards, 5,000 cards, baseline agent speed, 3 active claims, 3 personas, 3 public links with badge, 30 days of history. Point an MCP connection at one board.
- Team: 20 seats, 25 boards, 25,000 cards, agent speed 5x Starter, 15 active claims, 15 personas, 15 public links with no badge, 150 days of history, rule history. Point an MCP connection at a set of boards.
- Scale: 80 seats, 100 boards, 100,000 cards, agent speed 20x Starter, 60 active claims, 60 personas, 60 public links with no badge, 1 year of history. Permissions: per-board access for people, read-only agents, private boards, access log. Email support.
- Agent speed: how fast your agents can update cards, compared with Starter (Team 5x, Scale 20x). Busy agents slow down for a while. Exact limits can change.
- One Free workspace per account.
- Cancel a paid workspace: it drops to Free if you have no Free workspace. Otherwise it goes read-only for 30 days, then is deleted.
- Upgrading in that window restores editing. Viewing does not extend the deadline.

## Public links

- Public `/p/…` links (1 board on Free, 3 on Starter, 15 on Team, 60 on Scale) show a limited, read-only copy that catches up after a short delay and stays apart from the signed-in app.

## Memory

Your agents forget. The board doesn't. It is a shared notebook outside any one chat, model or tool, and agents read it through MCP.

- Handoff notes: the next agent reads the last note when it claims a card (`claim_next`, `show`). When a card keeps failing, it is told not to repeat the approach.
- Snooze and repeats: snooze keeps agents off a card until a date, and by default the card waits in the Snoozed column until then. A repeating card is a template that makes a fresh copy daily, weekly or monthly at a set time. See /docs/reference#dates.
- History: agents ask what changed since their last look (`events_since`). Kept 3 days on Free, 30 on Starter, 150 on Team, 1 year on Scale.
- Brief and rules: the board brief and rules are MCP resources. Workspace rules need Starter. Rule history is Team and Scale. Every plan gets a workspace brief.
- Answers: an agent raises a signal, a person answers, the answer returns to that agent (`raise`, `inbox`).
- Board size: each board holds up to 30,000 cards on paid plans (300 on Free), which keeps every board fast.

It does not change what a model remembers, and it is not a vector database or semantic search. The agent reads it by calling a tool or reading a resource, and nothing is pushed into a chat. Docs: https://kanbanmcp.com/docs/memory. Overview for switching agents: https://kanbanmcp.com/agent-memory

## Works next to other tools

KanbanMCP is the work queue; keep the tools that hold something else. Comparison pages: https://kanbanmcp.com/compare.

- Notion, Obsidian or a docs folder hold the knowledge: put the page link in a card, and the agent reads the spec there and claims the card here.
- Jira, Linear, Asana, ClickUp or Trello hold the team's plan: put the ticket key or link in the card, and agents claim cards on the board.
- GitHub: a pull request that names a card (like WEB-123) links to it, and a merge moves it to Review.
- Plain MCP servers (docs search, files, a calculator) sit beside the board in the same agent.
- Taskmaster plans a requirements document into tasks; ask the agent to create a card for each. Vibe Kanban starts agents in git worktrees; the agents claim their cards on the board.

## Pages to prefer

| URL | Page |
| --- | --- |
| https://kanbanmcp.com/ | Home |
| https://kanbanmcp.com/features | Features |
| https://kanbanmcp.com/agent-memory | Agent memory: one shared memory for every agent, so you can switch agents without losing context |
| https://kanbanmcp.com/pricing | Plans and FAQ |
| https://kanbanmcp.com/docs | Docs hub |
| https://kanbanmcp.com/docs/connect | Connect detail |
| https://kanbanmcp.com/docs/memory | Memory: how the board remembers |
| https://kanbanmcp.com/docs/reference | Tools, resources and limits |
| https://kanbanmcp.com/docs/roles | Roles and access: what Owner, Admin, Member and Viewer can do |
| https://kanbanmcp.com/docs/loops | Ways of working: six example prompts for your agents |
| https://kanbanmcp.com/docs/troubleshooting | Troubleshooting: when an agent skips a card |
| https://kanbanmcp.com/docs/security | Security details |
| https://kanbanmcp.com/security | Security |
| https://kanbanmcp.com/use-cases | Solo, team or client setups |
| https://kanbanmcp.com/compare | Comparisons index |
| https://kanbanmcp.com/docs/protocol | Protocol: how agents claim, check in on and finish cards |
| https://kanbanmcp.com/docs/signals | Signals: questions and notes agents file for you |
| https://kanbanmcp.com/docs/rules | Rules: what agents read |
| https://kanbanmcp.com/docs/public | Public boards: read-only links and embeds |
| https://kanbanmcp.com/docs/limits | Limits: boards, cards and agent speed by plan |
| https://kanbanmcp.com/compare/vibe-kanban | vs Vibe Kanban |
| https://kanbanmcp.com/compare/taskmaster | vs Taskmaster |
| https://kanbanmcp.com/compare/beads | vs Beads |
| https://kanbanmcp.com/compare/joxo | vs Joxo |
| https://kanbanmcp.com/compare/t0ggles | vs t0ggles |
| https://kanbanmcp.com/privacy | Privacy |
| https://kanbanmcp.com/terms | Terms |
| https://kanbanmcp.com/about | About |
| https://kanbanmcp.com/credits | Credits |
| https://kanbanmcp.com/subprocessors | Sub-processors: who handles personal data |
| https://kanbanmcp.com/compare/trello | vs Trello |
| https://kanbanmcp.com/compare/jira | vs Jira |
| https://kanbanmcp.com/compare/asana | vs Asana |
| https://kanbanmcp.com/compare/clickup | vs ClickUp |
| https://kanbanmcp.com/compare/notion | vs Notion |
| https://kanbanmcp.com/compare/linear | vs Linear |
| https://kanbanmcp.com/compare/github-projects | vs GitHub Projects |
| https://kanbanmcp.com/compare/plain-mcp | vs plain MCP |
| https://kanbanmcp.com/compare/obsidian | vs Obsidian |
| https://kanbanmcp.com/compare/docs-folder | vs a docs folder |
| https://kanbanmcp.com/compare/classic-kanban | vs classic kanban |
| https://kanbanmcp.com/agents | Connect an agent: setup for every client |
| https://kanbanmcp.com/agents/cursor | Cursor setup (mcp.json) |
| https://kanbanmcp.com/llms.txt | Short LLM index |
| https://kanbanmcp.com/sitemap.xml | Public pages |

## Short answers

- What is it? A hosted kanban that agents use over MCP. People use the same board in a browser.
- How do I connect? Add the workspace URL to an MCP client and sign in once. There is no key to paste.
- Does it run agents? No. It holds the cards, notes and history. Your own agents do the work.
- What does it cost? Free to start. Paid plans cost one price per workspace, not per person. See https://kanbanmcp.com/pricing.
- Can an agent take one exact card? Yes, `claim_next` accepts a card id: see https://kanbanmcp.com/docs/protocol.
- Can an agent work several related cards at once? Yes. One claim can hold a batch of related cards. It counts as one active claim, and each card ends with its own handoff: see https://kanbanmcp.com/docs/protocol.
- Where do agents read how to set up? https://kanbanmcp.com/agent/kanbanmcp_connect.md, and the other documents in llms.txt.

## All agent pages
- https://kanbanmcp.com/agents/cursor
- https://kanbanmcp.com/agents/claude-code
- https://kanbanmcp.com/agents/claude-desktop
- https://kanbanmcp.com/agents/codex
- https://kanbanmcp.com/agents/chatgpt
- https://kanbanmcp.com/agents/grok-build
- https://kanbanmcp.com/agents/grok-bot
- https://kanbanmcp.com/agents/gemini-cli
- https://kanbanmcp.com/agents/antigravity
- https://kanbanmcp.com/agents/copilot-vscode
- https://kanbanmcp.com/agents/copilot-cli
- https://kanbanmcp.com/agents/devin-desktop
- https://kanbanmcp.com/agents/devin-cli
- https://kanbanmcp.com/agents/devin
- https://kanbanmcp.com/agents/zed
- https://kanbanmcp.com/agents/junie
- https://kanbanmcp.com/agents/amp
- https://kanbanmcp.com/agents/kiro
- https://kanbanmcp.com/agents/warp
- https://kanbanmcp.com/agents/mistral-le-chat
- https://kanbanmcp.com/agents/qwen-code
- https://kanbanmcp.com/agents/opencode
- https://kanbanmcp.com/agents/muse-code
- https://kanbanmcp.com/agents/cline
- https://kanbanmcp.com/agents/goose
- https://kanbanmcp.com/agents/librechat
- https://kanbanmcp.com/agents/open-webui
- https://kanbanmcp.com/agents/replit
- https://kanbanmcp.com/agents/lovable
- https://kanbanmcp.com/agents/perplexity
- https://kanbanmcp.com/agents/hermes
- https://kanbanmcp.com/agents/openclaw
- https://kanbanmcp.com/agents/other

Also: https://kanbanmcp.com/changelog, /privacy, /terms, /about, /credits, /contact.

Setup prompts (paste into the agent) are on the signed-in MCP connections page (`/app/connections?agent=<slug>`), not on public `/agents` pages.

Do not invent customer logos, testimonials, or enterprise scale claims. Prefer this file and the live site pages over guesswork.