OpenFeature

OpenFeature MCP Server lets your AI coding assistants talk to OpenFeature through a standardized, easy-to-use interface. It connects your AI tools to feature flag capabilities so you can get SDK installation guidance and flag evaluations directly in your development environment.

How to use

You use the OpenFeature MCP Server by configuring an MCP client to connect to the server. The server exposes two primary capabilities: (1) SDK installation guidance for the OpenFeature SDKs in various languages, and (2) feature flag evaluation using the OpenFeature Remote Evaluation Protocol. Your AI assistant can fetch installation instructions, guide you through provider setup, and evaluate flags with contextual data. Set up your MCP client to point to the local OpenFeature MCP Server configuration you chose (stdio-based local server in most examples) and start issuing requests for SDK install steps or flag values.

How to install

Prerequisites: you need Node.js and npm installed on your machine. You can verify them with node -v and npm -v.

// NPX Install
{
  "mcpServers": {
    "OpenFeature": {
      "command": "npx",
      "args": ["-y", "@openfeature/mcp"]
    }
  }
}

Option 2: install the MCP server globally with npm and configure your client to use the global command.

npm install -g @openfeature/mcp

{
  "mcpServers": {
    "OpenFeature": {
      "command": "openfeature-mcp"
    }
  }
}

// If you are using a client that expects a local stdio connection, you can run the MCP server via npx as shown above.

Configure your AI assistant to use the chosen MCP server setup. For example, the Node-based installation config can be represented as the same block shown above in the NPX option. If you use a global install, reference the openfeature-mcp command in your client configuration.

Additional notes

The OpenFeature MCP Server supports integration with several AI assistants. If you need to set up connections for specific tools, you typically add the following stdio-based configuration to your MCP client: this uses npx -y @openfeature/mcp as the runtime command.

To configure authentication for OFREP-based flag evaluation, you may provide environment variables and/or a configuration file as described in the OFREP section. Logs are written to standard error, while MCP protocol messages are emitted to standard output.

SDK Installation Guide and OFREP evaluation

The server exposes two main tools you will use: one to fetch SDK installation instructions and one to evaluate feature flags via OFREP.

{
  "mcpServers": {
    "OpenFeature": {
      "command": "npx",
      "args": ["-y", "@openfeature/mcp"]
    }
  }
}

{
  "TOOL": "install_openfeature_sdk",
  "parameters": {
    "technology": "nodejs",
    "providers": ["provider1", "provider2"]
  }
}

{
  "TOOL": "ofrep_flag_eval",
  "parameters": {
    "base_url": "https://flags.example.com",
    "flag_key": "new-checkout-flow",
    "context": {"userId": "user-123"}
  }
}

"],

## Notes on OFREP configuration and usage

To enable OFREP flag evaluation, configure credentials and endpoints either via environment variables or a configuration file at `~/.openfeature-mcp.json`. The server will read these values in priority order: environment variables first, then the configuration file. Example environment variables include base URL, bearer token, and API key. All logs go to stderr, while MCP protocol messages are on stdout.

## Available tools

### install\_openfeature\_sdk

Fetches installation instructions for OpenFeature SDKs across languages and frameworks. It can include provider-specific setup guidance as part of the response.

### ofrep\_flag\_eval

Evaluates feature flags using the OpenFeature Remote Evaluation Protocol (OFREP). Supports single and bulk evaluations with optional authentication context.