kaiten-mcp (ViktorOgnev)
kaiten-mcp
An MCP server for Kaiten with 246 tools across 27 modules: cards, Service Desk, automations, audit and charts
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Tools cover write operations across nearly the whole Kaiten API, including roles, groups and automations
- A shared HTTP server with OAuth grants access to multiple users of one instance, so the config needs careful review
Install
Manual install
cp .env.example .envFill in KAITEN_SUBDOMAIN and KAITEN_TOKEN for local stdio, or MCP_PUBLIC_URL and MCP_OAUTH_ISSUER_URL for a shared HTTP server.
This is third-party code. Review the repository files before installing.
What it does
The server covers nearly the whole Kaiten API: 246 tools across 27 modules, from cards, comments and checklists to Service Desk (47 tools, the largest module), webhooks, automations and workflows, roles and groups, audit and analytics, and charts. It runs over stdio for local Claude Code or Claude Desktop connections, or over HTTP for a Docker deployment with OAuth authorization for multiple users of one shared instance. The API address is built from a company subdomain and base domain, or set directly as a full URL. It ships five skills: kaiten-audit, kaiten-deploy, kaiten-heavy-data, kaiten-metrics and managing-kaiten, with a detailed tool reference and standard workflows.
Who it is for. For Kaiten teams who need the broadest possible API coverage, including Service Desk, roles and automations, not just basic cards.
Good fit when
- You need access to the full surface of the Kaiten API, including Service Desk, roles, groups and webhooks
- You need a shared HTTP server with OAuth for multiple users at one company
- You want ready skills for audit, metrics and deploying the server itself
Not a fit when
- You only need a small toolset for cards: 246 tools can be excessive for an agent's context
- You are not ready to run Docker and configure OAuth for a shared server
- You need a CLI, not the MCP protocol: the same author has a separate kaiten-cli
Example request
Build a weekly Service Desk ticket summary and show a chart by request typeLimitations
It needs Docker or Python 3.11+ and a Kaiten account with an API token. A production shared HTTP server needs MCP_HTTP_AUTH_MODE=oauth and a public URL; the local shared bearer token mode is marked legacy. The large tool count (246) can eat a lot of an agent's context if the client loads them all at once.
How to disable. Stop the container with docker compose down for the HTTP variant, or remove the server from your stdio client config and delete the .env file with the token.
MCP
- Transport
- stdio, http
- Authentication
- OAuth
| Environment variables | |
|---|---|
| KAITEN_SUBDOMAIN required | Your company's kaiten.ru subdomain |
| KAITEN_TOKEN required, secret | Kaiten user API token for stdio or the legacy shared HTTP mode |
| MCP_HTTP_AUTH_MODE | HTTP transport auth mode: oauth, shared or none |
| MCP_PUBLIC_URL | Public MCP endpoint URL for a production shared server |
Security check
- Tools cover write operations across nearly the whole Kaiten API, including roles, groups and automations
- A shared HTTP server with OAuth grants access to multiple users of one instance, so the config needs careful review
README in short
The Russian README gives a table of all 27 modules with tool counts, explains the two transports (stdio and HTTP), and details environment variables for local and shared HTTP deployment, including oauth, shared and none auth modes. A separate deploy folder holds a Caddyfile, bootstrap and deploy scripts, and .env examples for production and a ChatGPT tunnel. skills/managing-kaiten includes dedicated TOOLS_REFERENCE.md and WORKFLOWS.md files.
FAQ
How many tools does the server have?
246 tools across 27 modules, from cards to Service Desk, automations and charts, per the README's table.
Can I run a shared server for multiple users?
Yes, HTTP transport with MCP_HTTP_AUTH_MODE=oauth is built for a production shared deployment with a public URL.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI