Harbor MCP
Provides MCP tools to manage a local-first Kanban board, including projects, issues, dependencies, and labels. Enables AI clients to interact with the board via a stdio MCP server connected to a NestJS API.
README
<p align="center"> <img src="docs/assets/harbor-icon.svg" alt="Harbor" width="128" height="128" /> </p>
<h1 align="center">Harbor</h1>
<p align="center"> Local-first Kanban board with an MCP server for any compatible client.<br /> Run the full stack on your machine — UI, API, Postgres, and agent tools. </p>
<p align="center"> <a href="#quick-start">Quick start</a> · <a href="#what-you-can-do">Features</a> · <a href="#mcp">MCP</a> · <a href="#development">Development</a> </p>
What is Harbor?
Harbor is a single-operator Kanban that stays on localhost. Use the web UI to manage projects and issues, or drive the same board from any MCP client (Claude, Cursor, and others) through a stdio MCP server that talks to the Nest API.
MCP client ──stdio──► Harbor MCP ──HTTP Bearer──► Nest API ──► Postgres
Browser ─────────────────────────► Harbor UI ──/api──► Nest API
| Package | Path | Role |
|---|---|---|
@kanban/api |
apps/api |
NestJS + Prisma + OpenAPI |
@kanban/web |
apps/web |
Harbor UI (Vite + React) |
@kanban/mcp |
apps/mcp |
Stdio MCP server |
What you can do
- Projects & boards — multiple projects, Columns and List layouts, column colors, reorder, done columns
- Issues — create, edit, move, archive/restore, delete; priority, type, due date, effort (hours / LOC)
- Dependencies — issue links (
blocks,relates_to,duplicates) with cycle checks; blocker summaries via MCP - Labels, comments, activity — label attach/detach, threaded comments, activity history
- Attachments — multipart file upload in the issue drawer (UI/API; max 20 MB)
- Live updates — board refreshes over SSE when the API or an MCP agent changes data
- Global search —
Ctrl/Cmd+Kacross projects - MCP tools — projects, columns, issues, links, blockers, epics, labels, comments, activity
- Auth — first-run admin signup, cookie sessions for the UI, Bearer MCP tokens from Harbor
Quick start
Requirements: Bun 1.3+, Docker
git clone git@github.com:Eidantz/Harbor.git
cd Harbor
bun install
bun run setup:env # copies .env.example → .env and links apps/api/.env
bun run docker:up # builds & starts db + api + web
| Service | URL (defaults) |
|---|---|
| Harbor UI | http://localhost:3010 |
| API health | http://localhost:3001/health |
| OpenAPI | http://localhost:3001/api/docs |
Ports come from .env (WEB_PORT, API_PORT, POSTGRES_PORT). Defaults are 3010 / 3001 / 5433 so Harbor can sit next to stacks that already use 3000 / 5432 (e.g. Langfuse).
On a fresh database, open the UI and create the admin account. After that, the same screen is sign-in. Seed data includes sample project KAN.
Stop the stack with bun run docker:down.
MCP
Harbor exposes a stdio MCP server any compatible client can launch (Claude Desktop / Claude Code, Cursor, and others).
- Start Harbor (
bun run docker:up) and complete admin signup. - In the UI: MCP tokens → Create token (copy it once).
- Build the server:
bun run build:mcp
- Point your client at the server with the same command + env:
{
"mcpServers": {
"harbor": {
"command": "node",
"args": ["/absolute/path/to/Harbor/apps/mcp/dist/index.js"],
"env": {
"KANBAN_API_URL": "http://127.0.0.1:3001",
"KANBAN_API_TOKEN": "<paste token from Harbor>"
}
}
}
}
| Client | Where to put the config |
|---|---|
| Claude Desktop | Claude → Settings → Developer → Edit Config (claude_desktop_config.json) |
| Claude Code | Project or user .mcp.json (same mcpServers shape) |
| Cursor | Copy docs/mcp.example.json → .cursor/mcp.json |
Use an absolute path to apps/mcp/dist/index.js when the client does not start in the repo root. Reload MCP after changing config.
Full tool catalog: apps/mcp/README.md.
Development
Keep Postgres in Docker; run API and web on the host:
bun run setup:env
docker compose up -d db
bun run db:migrate:deploy
bun run db:seed
bun run dev # API :3001 + web :3000
Useful scripts: bun run dev:api, bun run dev:web, bun run db:studio, bun run smoke.
Auth & privacy
| Client | Mechanism |
|---|---|
| Harbor UI | First-run signup, then login → HTTP-only session cookie |
| MCP / scripts | Authorization: Bearer <token> from MCP tokens |
- Compose publishes ports on
127.0.0.1only. - Real secrets live in
.envand local MCP config files (e.g..cursor/) — gitignored. Use.env.exampleanddocs/mcp.example.jsonas templates. - Rotate
SESSION_SECRETand revoke MCP tokens for anything beyond casual local use.
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.