Unleash MCP Server

What is Unleash MCP Server about?

Enables LLM‑driven applications to interact with an Unleash feature‑flag backend. Through MCP calls, AI agents can query flag status, expose flag metadata, create new flags, update existing ones, and retrieve a list of projects.

How to use Unleash MCP Server?

1. Install dependencies

   npm i
   

2. Build the TypeScript source

   npm run build
   

3. Run the server (default HTTP transport on port 3001)

   npm start
   

4. Configure an MCP client (e.g., Claude, Cursor) with the following JSON snippet:

   {
     "mcpServers": {
       "unleash": {
         "command": "npx",
         "args": ["-y", "unleash-mcp"],
         "env": {
           "UNLEASH_URL": "YOUR_UNLEASH_END_POINT",
           "UNLEASH_API_TOKEN": "YOUR_UNLEASH_API_TOKEN",
           "MCP_TRANSPORT": "stdio",
           "MCP_HTTP_PORT": 3001
         }
       }
     }
   }
   

5. Inspect the server with the MCP inspector utility if needed.

Key features of Unleash MCP Server

• Flag status query – Retrieve the enabled/disabled state of a single flag.
• Flag metadata exposure – Provide flag description, rollout strategy, and other properties to LLMs.
• Create & update flags – AI agents can programmatically add new flags or modify existing ones.
• Project listing – List all Unleash projects, useful for dynamic discovery.
• Multiple transports – Supports both HTTP/SSE and STDIO transports.
• Typed TypeScript codebase – Strong typing, JSDoc documentation, and clean architecture.

Use cases of Unleash MCP Server

• Feature‑flag driven feature rollout controlled by AI assistants.
• Dynamic UI/UX adaptation where an LLM decides which UI components to enable based on flag status.
• Automated testing pipelines that toggle flags before running tests.
• Continuous deployment governance allowing AI tools to gate deployments via flag checks.
• Multi‑tenant SaaS environments where each tenant’s feature set is managed through LLM queries.

FAQ from the Unleash MCP Server

Q: Which Node.js version is required? A: Node.js v18 or higher.

Q: Do I need a TypeScript compiler installed globally? A: No. The project’s npm run build script uses the local dev dependencies.

Q: Can I use the server over HTTP instead of STDIO? A: Yes. Set MCP_TRANSPORT to http (default) and optionally adjust MCP_HTTP_PORT.

Q: How are environment variables passed to the server? A: Through the env object in the MCP server configuration (e.g., UNLEASH_URL, UNLEASH_API_TOKEN).

Q: Is there a way to test the server locally? A: Run npm test after building; the test suite covers flag retrieval and project listing.

Unleash MCP Server

A Model Context Protocol (MCP) server implementation that integrates with Unleash Feature Toggle system.

Overview

This project provides a bridge between LLM applications and Unleash feature flag system using the Model Context Protocol (MCP). It allows AI applications to:

1. Check feature flag status from Unleash
2. Expose feature flag information to LLMs
3. Create feature flag
4. Update feature flag
5. List all projects

Table of Contents

• Requirements
• Installation
• Architecture
• Configuration
• Development
• Contributing
• License

Requirements

• Node.js (v18 or higher)
• TypeScript (v5.0 or higher)
• Access to an Unleash server instance

Installation

# Install dependencies
npm i

Architecture

The Unleash MCP Server acts as a bridge between LLM applications and the Unleash feature flag system:

+----------------+      +-------------------+      +----------------+
|                |      |                   |      |                |
|  LLM App       | <--> |  Unleash MCP      | <--> |  Unleash API   |
|  (MCP Client)  |      |  Server           |      |  Server        |
|                |      |                   |      |                |
+----------------+      +-------------------+      +----------------+

Development

Project Structure

unleash-mcp-server/
├── src/
│   ├── index.ts              # Main entry point
│   ├── server.ts             # Server implementation
│   ├── config.ts             # Configuration handling
│   ├── transport/            # MCP transport implementations
│   │   ├── http.ts           # HTTP/SSE transport
│   │   └── stdio.ts          # STDIO transport
│   ├── unleash/              # Unleash API client implementations
│   │   ├── unleash-client.ts # Main Unleash client
│   │   ├── get-feature-flag.ts
│   │   └── get-all-projects.ts
│   ├── resources/            # MCP resource implementations
│   │   ├── flags.ts          # Feature flag resources
│   │   └── projects.ts       # Project resources
│   ├── tools/                # MCP tool implementations
│   │   ├── get-flag.ts       # Get feature flag tool
│   │   └── get-projects.ts   # Get projects tool
│   └── prompts/              # MCP prompt implementations
│       ├── flag-check.ts     # Check single flag
│       └── batch-flag-check.ts # Check multiple flags
├── tests/                    # Tests
└── package.json              # Project configuration

Coding Standards

• Naming Conventions:

  • Files: Use kebab-case.ts (e.g., feature-flag.ts)
  • Classes: Use PascalCase (e.g., UnleashClient)
  • Functions/Methods: Use camelCase (e.g., getFlagStatus)
  • Interfaces/Types: Use PascalCase (e.g., FeatureFlagConfig)

• Imports:

  • Always include .js extension when importing local files
  • Follow import ordering: Node.js built-ins → External dependencies → Local imports
  • Use named exports over default exports

• Documentation:

  • Use JSDoc comments for public functions, classes, and interfaces
  • Document complex logic with inline comments

Building

# Compile TypeScript
npm run build

# Run the server
npm start

Testing

# Run tests
npm test

Inspecting

# MCP stdio inspect
npm run build
npx @modelcontextprotocol/inspector node dist/index.js

# MCP sse inspect
npm start
npx @modelcontextprotocol/inspector

How to use

For Claude or Cursor config:

{
  "mcpServers": {
    "unleash": {
      "command": "npx",
      "args": [
        "-y",
        "unleash-mcp"
      ],
      "env": {
        "UNLEASH_URL": "YOUR_UNLEASH_END_POINT",
        "UNLEASH_API_TOKEN": "YOUR_UNLEASH_API_TOKEN",
        "MCP_TRANSPORT": "stdio",
        "MCP_HTTP_PORT": 3001
      }
    }
  }
}

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support the Project

If you find this project helpful, consider buying me a coffee!

Scan the QR code above or click here to support the development of this project.

License

This project is licensed under the MIT License - see the LICENSE file for details.