Kanboard MCP Server
Provides GitHub OAuth login for the remote HTTP transport, letting users sign in with their GitHub account and restricting access to a configurable allowlist of GitHub logins.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Kanboard MCP Server show me all my overdue tasks"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Kanboard MCP Server
An MCP server that lets Claude create, update and track tasks on your self-hosted Kanboard, from Claude Code, Claude Desktop, claude.ai or the Claude mobile apps.
It talks to Kanboard only through its JSON-RPC API, so it works with any Kanboard instance you can reach over HTTP (localhost, LAN or a public URL) and with any Kanboard database backend.
Tools
Tool | What it does |
| Projects with IDs and board links |
| Columns of a project with their open tasks |
| Tasks in a project, filtered by |
| Kanboard filter syntax ( |
| Overdue tasks, for you or for one project |
| Details, tags, subtasks, comments and attachment list |
| Read an attachment: images come back as images, text files as text |
| Title, description, priority, column, assignee, due date, tags |
| Move a task to another column (by name) |
| Complete or reopen |
| Add a progress note (Markdown) |
| Checklist items and their status |
| Assignable users |
| Which Kanboard user and API mode the server is using |
Read-only tools are annotated as such, so Claude clients can auto-approve them and ask before writes.
Related MCP server: kanban-mcp
1. Get a Kanboard API token
You can use either kind of token:
Personal access token (recommended): Kanboard → My profile → API → generate a token. Set
KANBOARD_USERNAMEto your login name. Actions run as you and follow your project permissions.Application token: Kanboard → Settings → API. Set
KANBOARD_USERNAME=jsonrpc. It has full access to every project. SetKANBOARD_DEFAULT_USER_IDso comments have an author.
The two are easy to mix up: the application token used with your own username fails with
401 Unauthorized.
Also set Kanboard's Settings → Application settings → Application URL to the address you open
Kanboard at; otherwise the task links Claude shows you point to http://localhost/....
2. Use it from Claude Code on the same machine (stdio)
No clone needed. uv fetches and runs the server:
claude mcp add --scope user kanboard \
--env KANBOARD_URL=http://localhost:8080 \
--env KANBOARD_USERNAME=your-username \
--env KANBOARD_API_TOKEN=your-token \
-- uvx --from git+https://github.com/b-ak/kanboard-mcp-server kanboard-mcp--scope user makes the server available in every project; without it, it is registered for the
current folder only.
Claude Desktop's claude_desktop_config.json takes the same command:
{
"mcpServers": {
"kanboard": {
"command": "uvx",
"args": ["--from", "git+https://github.com/b-ak/kanboard-mcp-server", "kanboard-mcp"],
"env": {
"KANBOARD_URL": "http://localhost:8080",
"KANBOARD_USERNAME": "your-username",
"KANBOARD_API_TOKEN": "your-token"
}
}
}
}3. Use it from claude.ai and the Claude mobile apps (remote HTTP)
claude.ai and the mobile apps connect from Anthropic's cloud, so the server must:
run where it can reach Kanboard (e.g. on the same Raspberry Pi),
be reachable at a public HTTPS URL (Cloudflare Tunnel, Tailscale Funnel or a reverse proxy),
require login. GitHub OAuth is built in and limited to the GitHub accounts you list.
Claude (web / mobile / Code) ──HTTPS──▶ tunnel/proxy ──▶ kanboard-mcp :8000 ──▶ Kanboarddocs/deployment.md walks through the whole setup step by step: Kanboard token, GitHub OAuth App, Docker Compose, Cloudflare Tunnel, connecting Claude, verification and troubleshooting.
Once it's running at https://kanboard-mcp.example.com:
claude.ai / mobile: Settings → Connectors → Add custom connector, URL
https://kanboard-mcp.example.com/mcp, then sign in with GitHub. Connectors you add on claude.ai are also available in the Claude mobile apps.Claude Code:
claude mcp add --scope user --transport http kanboard https://kanboard-mcp.example.com/mcp, then run/mcpto sign in.
4. Tell Claude when to use it
Connecting gives Claude the tools, not the habits. Put your rules (which project, when to create,
move, comment or close tasks, what to ask first) in your project's CLAUDE.md or a claude.ai
Project's instructions. Start from docs/claude-md-template.md.
Configuration
All settings are environment variables. See .env.example. The CLI also accepts
--env-file, --transport, --host and --port.
Variable | Default | Description |
| required | Kanboard base URL; |
|
| Your username (personal token) or |
| required | API token |
| Comment author when using the application token | |
|
| Set |
|
| Request timeout in seconds |
|
|
|
|
| HTTP bind address (Docker image uses |
|
|
|
| Shared secret for | |
| Public URL of this server, for | |
| GitHub OAuth app credentials | |
| Comma-separated GitHub logins allowed to connect | |
| derived | Optional key for tokens the server issues |
|
| Where OAuth state is stored ( |
The server refuses to serve HTTP on a non-loopback address without authentication, unless you set
MCP_ALLOW_UNAUTHENTICATED=true because a proxy in front of it already handles auth.
Development
uv sync
uv run pytest
uv run ruff check . && uv run ruff format --check .
uv run kanboard-mcp --env-file .env # stdio
uv run kanboard-mcp --env-file .env --transport http # http on 127.0.0.1:8000
npx @modelcontextprotocol/inspector # optional: browse the toolsLicense
Apache-2.0
Available Tools
17 toolsadd_commentB
Add a comment (Markdown supported) to a task, e.g. a progress log entry.
Inputs: content (required), task_id (required). Returns: result
close_taskA
Close (complete) a task. It can be reopened with reopen_task.
Inputs: task_id (required). Returns: result
create_subtaskC
Add a subtask (checklist item) to a task.
Inputs: title (required), task_id (required), user_id, time_estimated. Returns: result
create_taskC
Create a task. Defaults to the project's first column.
Inputs: title (required), project_id (required), tags, column, +4 more. Returns: object
get_boardARead-only
Show a project's board: each column with its open tasks, in order.
Inputs: project_id (required). Returns: object
get_meARead-only
Show which Kanboard user/API mode this server is connected as.
Inputs: no named parameters. Returns: object
get_overdue_tasksARead-only
List open tasks whose due date has passed.
Inputs: project_id. Returns: result
get_taskBRead-only
Get everything about a task: details, tags, subtasks, comments and attachments.
Use this to summarise or review a task's progress.
Inputs: task_id (required), max_comments. Returns: object
get_task_fileARead-only
Read a task attachment: images are returned as images, text files as text.
Inputs: file_id (required).
list_projectsARead-only
List Kanboard projects (ID, name, board URL). Start here to find project IDs.
Inputs: include_inactive. Returns: result
list_project_usersARead-only
List users that can be assigned tasks in a project, as {user_id: name}.
Inputs: project_id (required). Returns: object
list_tasksBRead-only
List tasks in a project, optionally filtered by status and column.
Inputs: project_id (required), column, status. Returns: result
move_taskA
Move a task to another column of its board (e.g. 'In progress', 'Done').
Inputs: column (required), task_id (required), position. Returns: object
reopen_taskB
Reopen a closed task.
Inputs: task_id (required). Returns: result
search_tasksARead-only
Search tasks with Kanboard's filter syntax across one or all projects.
Inputs: query (required), project_id. Returns: result
update_subtaskC
Change a subtask's status or title.
Inputs: task_id (required), subtask_id (required), title, status. Returns: result
update_taskB
Update task fields. Only the fields you pass are changed.
Inputs: task_id (required), tags, title, due_date, +3 more. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 17 tools
Most tools target a clear resource+action pair (get_task, move_task, close_task, add_comment), and the create/update split is well signposted. The only real overlap is list_tasks vs search_tasks vs get_board vs get_overdue_tasks, which all return task collections, but the descriptions differentiate them by filter mechanism (status/column vs Kanboard filter syntax vs board layout vs due date).
Every name is snake_case verb_noun: get_*, list_*, search_*, create_*, update_*, move_*, close_*, reopen_*, add_*. Verb choice is semantically consistent (get for single-item fetch, list for collections, create/update for mutations), with only trivial variance like add_comment vs create_subtask reflecting the domain.
17 tools is slightly on the heavy side but each covers a distinct slice of the Kanboard domain (identity, projects, board, tasks, subtasks, comments, files, users) with little redundancy. Nothing looks like filler; it sits just above the comfortable band rather than being bloated.
The task lifecycle is well covered (create, update, move, close, reopen, comment, subtask, read attachment), but project management is read-only — no create_project, update_project, or archive, and there is no delete_task, delete_subtask, or delete_comment. File handling is also read-only (get_task_file with no upload), so an agent hitting those operations dead-ends.
Maintenance
Related MCP Connectors
Simple kanban for small teams. Read and update your Moxboard boards and tasks from Claude.
Remote MCP for Kanban AI boards—manage projects, tasks, and comments from AI tools.
AI-native Kanban board — connect Claude to claim, work and move your tasks over MCP.
Task & board management for AI agents + humans. Kanban, comments, digests via MCP.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI-powered project management by connecting Claude to Planka kanban boards via 75+ MCP tools for creating, updating, and managing projects, boards, lists, and cards.1GPL 3.0
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage Planka kanban boards, including creating, updating, and organizing tasks, lists, and cards via MCP.65MIT
- FlicenseNot gradedqualityBmaintenanceMCP server that enables AI agents to manage tasks, boards, lists, and comments on a Kanban-style project management system through JSON-RPC methods.2-
- AlicenseAqualityBmaintenanceEnables Claude Code to control a Planka 2.x board via 15 MCP tools, with writes re-read and verified so reported success matches the board, plus an optional workflow that turns the board into durable agent memory.1556 npm4MIT