NiFiMCP

NiFi MCP (Model Context Protocol)

This repository contains the NiFi MCP project. It can be used in two ways, either:

• As a plain MCP Server that tools like Cursor AI, Claude Code, Antigravity etc. can connect to via stdio
• Or, with a built-in (contained in this repo) chat bot that you can connect to from your browser.

Either approach allows you to interact with Apache NiFi instances using natural language. There are currently around 30 tools available, covering building, updating, running, debugging, and documenting nifi flows. This has been tested with Nifi versions 1.23 and 1.28, but should work on other versions like 2.x assuming the Nifi REST Api stays consistent.

The built-in chat bot requires you to supply your own LLM API keys, better models give better results and selecting higher token context can help too. I have however found that using Cursor AI in Auto mode gives the best performance and results. It turned out to be a much better interface than I could build with my own chat bot. So I only recommend the built-in chat bot if you have no other options.

Example Conversations (Using built-in chat bot)

• Build A New Flow From A Spec - See how the system creates a complete NiFi flow from requirements
• Debug An Existing Flow - Watch the system diagnose and fix issues in an existing flow

Recent Updates

Version 2.0 - Major UI Overhaul

• Added direct MCP server support for IDE's like Cursor AI (best option)
• Enhanced the built in chatbot
  • Migrated from Streamlit to FastAPI/JavaScript client for improved performance and user experience
  • Enhanced real-time WebSocket communication for better workflow status updates
  • Improved message handling and duplicate message prevention
  • Modern, responsive UI with better accessibility and mobile support
  • Added a new workflow "Flow Documentation" to help generate better documentation about a flow

For the latest updates and release notes, see the GitHub Releases page.

Setup Instructions

Note: These instructions should also be followed after pulling a new version as there may be new package requirements.

Common Setup Steps (All Versions)

1. Clone the Repository:

   git clone [email protected]:ms82119/NiFiMCP.git
   cd NiFiMCP
   

2. Set Up a Virtual Environment: It's recommended to use a virtual environment to manage dependencies. You can create one using venv:

   python3 -m venv .venv
   

3. Activate the Virtual Environment:

   • On macOS and Linux:

     source .venv/bin/activate
     

   • On Windows:

     .venv\Scripts\activate
     

4. Install Dependencies: Use uv to install dependencies based on pyproject.toml and uv.lock:

   uv sync
   

5. Update the config.yaml file with your Nifi server details and LLM API keys Use the config.example.yaml as your guide for format and structure Note that the LLM API keys are only needed if you are using the built-in client

Using the NiFi MCP server (stdio, any MCP client)

You can use the NiFi MCP tools from any MCP-capable client over stdio, without running the REST server or chat UI. Configure your host to run the server from the repository root with at least one NiFi server in config.yaml.

Portable entrypoint (after uv sync):

{
  "command": "uv",
  "args": ["run", "python", "-m", "nifi_mcp_server.stdio_server"],
  "env": {}
}

Optional: set NIFI_SERVER_ID in env to an id from config.yaml; if omitted, the first server in that file is used.

• Full setup (working directory, python without uv, JSON examples): MCP stdio setup
• Agent / LLM behavior (any client): MCP Agent Guide
• Cursor only (.cursor/mcp.json, reload): Cursor MCP Setup

Using the Built-In Chat Bot

You need two separate terminal processes (two different Python apps). The browser is the only “client”—open the chat URL once; do not start a second copy of api.main on the same port.

Process	App	Port (example)	Role
1	nifi_mcp_server.server:app	8000 or 8001	MCP / NiFi backend (REST + tools the UI calls)
2	api.main:app	3000 or 3001	Serves the chat web UI and talks to process 1

MCP_SERVER_PORT must match the port process 1 listens on (it tells process 2 where to find the MCP server).

1. MCP backend (run first):

uvicorn nifi_mcp_server.server:app --reload --port 8000

If port 8000 is in use, pick another port and set MCP_SERVER_PORT to the same value:

MCP_SERVER_PORT=8001 uvicorn nifi_mcp_server.server:app --reload --port 8001

2. Chat UI web server (second terminal):

uvicorn api.main:app --reload --port 3000

If the MCP server uses 8001, point the UI at it:

MCP_SERVER_PORT=8001 uvicorn api.main:app --reload --port 3000

You may use another UI port (e.g. --port 3001) if 3000 is busy; only one api.main process may use that port.

3. Open in a browser:
http://localhost:3000 (or whatever port you chose for api.main, e.g. http://localhost:3001).

Usage Tips

For detailed usage information, tips, and UI features, see the Usage Guide.

Running Automated Tests

For comprehensive testing information and examples, see the Testing Guide.

License

This project is licensed under the Apache License 2.0.