Taiga MCP Server
About
MCP server connecting AI assistants to Taiga - projects, epics, user stories, tasks, issues, sprints, wiki, and more.
Details
- Author
- madebyclowd
- Categories
- Developer Tools
Jump to
Setup
Install Taiga MCP Server in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/madebyclowd/taiga-mcp-server
Follow the installation instructions in the repository README, then restart your MCP client.
Connect Claude, Cursor, Windsurf, or any other MCP-compatible AI assistant to yourTaigaproject management tool— read and update epics, user stories, tasks, issues, sprints, comments, and wiki pages just by chatting, in plain English. No coding required to use it.
In plain words:MCP (Model Context Protocol) is a standard that lets an AI assistant safely use outside tools — like reading your files or browsing the web. This package is that "tool" for Taiga: it's the bridge that lets your AI assistant see and change things in your Taiga account when you ask it to.
Package on npm:@madebyclowd/taiga-mcp-server
Not an official Taiga product.This is an independent, open-source integration.
- What can I actually do with this?
- Before you start
- Install
- Set it up with Claude Desktop, Claude Code, Cursor, or Windsurf
- Running it as a shared HTTP server instead
- All settings (environment variables)
- What's covered, and how it handles problems
- Frequently asked questions
- Contributing / running the tests
- Security
- License
Once it's connected, you can ask your AI assistant things like:
- "Show me all open bugs in the Website project."
- "Create a task called 'Fix login button' under the current sprint."
- "Move user story #482 to the QA column."
- "Summarize what changed on epic #12 this week."
- "Search the wiki for our deployment steps."
You don't type any code or API calls yourself — you just ask, in normal sentences, and the assistant does the rest.
Behind the scenes, the assistant does this by calling this server, which talks to Taiga's real API using your own Taiga login. It only does what your Taiga account is already allowed to do — it can't see or change anything you don't already have permission for in Taiga itself.
- Projects, epics, user stories, tasks, issues, sprints (milestones), wiki pages, comments (including edit/delete), attachments (including download), project members, voting/watching, ref lookup (#436→ the right item), batch create, and search.If it's in Taiga, there's a good chance this covers it.
- Assign people by email or full name instead of hunting down a numeric user id — the server figures out who you mean from the project's member list.
- Every response tells the AI assistant whether more results are waiting (paging) and lets it ask for a shorter or fuller version of the data (averbositysetting:minimal/standard/full). In plain terms: asking "list my backlog" won't dump a wall of raw data into your conversation and eat up the assistant's memory of what you've been talking about — it gets just enough to answer you.
- Cheap filter-id lookups (user_story_filters_data,task_filters_data,issue_filters_data) instead of paging through a full list just to find a valid status or tag id.
- Each tool is clearly labeled as read-only, safe-to-repeat, or something that changes data — so your AI assistant (and you) can tell at a glance what's harmless to run and what deserves a second look before confirming.
- A fallback "raw request" tool for the rare Taiga API endpoint that doesn't have a dedicated tool yet (webhooks, import/export, and similar admin-level features).
- A Taiga account— either onTaiga Cloudor a self-hosted Taiga instance.
- Node.jsversion 22.11 or newer, or version 24 or newer.
- An AI assistant that supports MCP— Claude Desktop, Claude Code, Cursor, Windsurf, or any other MCP-compatible client.
You donotneed to install or run anything Taiga-related yourself — this connects to your existing Taiga account over the internet (or your self-hosted instance's URL).
You don't need to install this by hand. Your AI assistant will download and run it automatically the first time it's used, vianpx. If you'd rather install it yourself:
# Run it once, without installing anything permanently npx -y @madebyclowd/taiga-mcp-server
Or install it globally, using whichever package manager you already use — you only need one of these:
pnpm add -g @madebyclowd/taiga-mcp-server # or npm install -g @madebyclowd/taiga-mcp-server
Set it up with Claude Desktop, Claude Code, Cursor, or Windsurf
All of these tools use the same kind of config file, usually a JSON file with anmcpServerssection. Add this to yours:
Option A: using a Taiga access token (recommended)
This is the safer option — your password is never stored anywhere.
- Log into Taiga in your browser.
- Get your auth token (in Taiga Cloud, this is available via your account settings or the API; for self-hosted Taiga, ask your admin how tokens are issued).
- Add this to your config:
{ "mcpServers": { "taiga": { "command": "npx", "args": ["-y", "@madebyclowd/taiga-mcp-server"], "env": { "TAIGA_BASE_URL": "https://api.taiga.io", "TAIGA_TOKEN": "your-taiga-auth-token" } } } }
Option B: using your username and password
{ "mcpServers": { "taiga": { "command": "npx", "args": ["-y", "@madebyclowd/taiga-mcp-server"], "env": { "TAIGA_BASE_URL": "https://api.taiga.io", "TAIGA_USERNAME": "your-username", "TAIGA_PASSWORD": "your-password" } } } }
Using self-hosted Taiga?Just changeTAIGA_BASE_URLto your own instance's URL, for examplehttps://taiga.mycompany.com.
After saving the config, restart your AI assistant (or reload its MCP connections). It should now show Taiga tools as available.
Running it as a shared HTTP server instead
The setup above (called "stdio") runs a private copy of the server just for you, on your own computer. If you instead want to runone shared serverthat a team can connect to over the network, use the HTTP mode.
The important difference: in HTTP mode, nobody's Taiga login is stored on the server. Each person connecting sends theirownTaiga token with each request, so everyone keeps their own Taiga permissions and audit trail — nobody has to share one account.
# Start the HTTP server — same package, a different command it exposes npx -y --package=@madebyclowd/taiga-mcp-server -- taiga-mcp-server-http
Each client connects by sending their Taiga token in a standard HTTP header:
POST /mcp HTTP/1.1 Host: your-server:3000 Content-Type: application/json Authorization: Bearer <your-taiga-auth-token>
If you're exposing this server on the internet (not just on your own machine or local network), read theHTTP_ALLOWED_ORIGINSnote in the settings table below — the default is permissive on purpose for easy local/team use, and you should lock it down for a public deployment.
In HTTP mode, the server itself doesn't hold any Taiga login — each connecting client sends its own token, so none of the settings below are about credentials.
What's covered, and how it handles problems
Plain description of what happens behind the scenes, so you know what to expect:
- It uses your own Taiga permissions.If you can't see or edit something in Taiga's own web app, this server can't either — it's making the same API calls your browser would.
- Your password/token is never written to disk or logged.It's only kept in memory while the server is running.
- If two people edit the same item at the same time, Taiga would normally reject the second save. This server automatically re-fetches the latest version and retries the save once instead of just failing.
- If Taiga temporarily rate-limits requests, the server waits and retries automatically instead of giving up right away.
- Errors are returned in a structured way(which field was wrong, and why) instead of a generic "something went wrong" — this helps the AI assistant actually understand and fix the problem instead of guessing.
- Tested against the real Taiga API, not just simulated responses — the automated test suite includes tests that run against an actual Taiga project, in addition to the usual offline tests.
- Released withnpm provenance— a signed record proving the published package was built from this exact GitHub repository, not modified or uploaded from somewhere else.
Every list-shaped tool (*_list,comment_list,attachment_list,epic_related_user_stories,search) now returns{ items, pagination }instead of a bare array. If you have code parsing one of these responses directly as an array, read.itemsinstead.paginationis{ count, current_page, has_next }. This is the point at which the tool surface's response shapes are considered stable — it's been tested carefully, including live against the real Taiga API, but hasn't had real-world use by others yet, so if you hit a problem, pleasereport it.
What is MCP / Model Context Protocol?It's a standard that lets AI assistants (like Claude) connect to external tools and data sources in a consistent way, instead of every integration being custom-built. This package implements that standard for Taiga.
How do I connect Claude (or Cursor, Windsurf) to Taiga?Add this package to your assistant's MCP config with your Taiga token — seeSet it upabove. It works the same way for any MCP-compatible client, not just Claude.
Is this made by the Taiga team?No. This is an independent, open-source project, not affiliated with or endorsed by Taiga/Kaleidos.
Do I need to run my own server?No, for normal personal use. Your AI assistant starts and stops this server automatically in the background. You only need to run it yourself as a standalone server if you're setting it up for a team (seeHTTP mode).
Does this work with self-hosted Taiga, or only Taiga Cloud?Both. Just pointTAIGA_BASE_URLat your own instance.
Is it safe to give this my Taiga password?You don't have to — a Taiga access token (Option A above) is the recommended way, and your password never gets involved at all. If you do use username/password, it's only kept in memory, never written to disk or logged.
Can it do things I'm not allowed to do in Taiga?No. It uses your own Taiga account and is limited to the exact same permissions you already have.
# Install dependencies pnpm install # Build pnpm build # Lint, format check, type-check pnpm lint pnpm format pnpm typecheck # Run the offline test suite pnpm test # Check dependencies for known security issues pnpm audit # Run the live test suite against a real Taiga project (see docs/testing.md) pnpm test:integration
Seedocs/testing.mdfor how to set up the live-test project.
SeeSECURITY.mdfor how to privately report a security problem.
This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.
Create crafted UI components inspired by the best 21st.dev design engineers.
Bring agent evaluations, observability, and synthetic test set generation directly into your IDE for free with Galileo's new MCP server
An MCP server to help AI assistants to answer questions and generate AccelByte Extend SDK code more effectively .
MCP server for AI Diagram Maker — generate beautiful software engineering diagrams directly inside Cursor, Claude Desktop, Claude Code, or any MCP-compatible AI agent
ALAPI MCP Tools,Call hundreds of API interfaces via MCP
AI-powered SVG animation generator that transforms static files into animated SVG components using the Allyson platform
MCP server that gives AI assistants on-demand access to 1,500+ amCharts docs, ~300 code examples, and 1000+ class API references.
APIMatic MCP Server is used to validate OpenAPI specifications using APIMatic. The server processes OpenAPI files and returns validation summaries by leveraging APIMatic’s API.
One shared context layer for AI agents and humans — live API specs, DB schemas, and versioned contracts across repos so every agent and teammate works from the same source of truth.
Build and deploy full-stack Next.js apps with 98 tools for React, AWS, and MongoDB
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
