OpenFeature MCP Server

The OpenFeature Model Context Protocol (MCP) Server enables AI coding assistants to interact with OpenFeature through a standardized protocol. It provides SDK installation guidance and feature flag evaluation capabilities directly within your AI-powered development environment.

The OpenFeature MCP Server is a local tool that connects AI coding assistants (like Cursor, Claude Code, VS Code, and Windsurf) to OpenFeature functionality. It acts as a bridge between your AI assistant and OpenFeature capabilities, enabling intelligent code generation and migration, SDK installation guidance, and feature flag evaluation.

This server is published to the MCP Registry under dev.openfeature/mcp.

⚠️ AI Agent Behavior: AI agents are non-deterministic and may not complete tasks correctly. Always manually review their changes before committing. If you encounter issues, please open an issue with details about your AI agent (e.g., Claude Code + Sonnet 4.5, Cursor + gpt-5-codex) with the commands you used and the behavior you saw.

Quick Start

NPX Install

The easiest way to use the OpenFeature MCP Server is through NPX, which requires no installation:

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

NPM Global Install

You can install the MCP server globally:

npm install -g @openfeature/mcp

Then configure your AI assistant to use the global installation:

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

AI Assistant Configuration

Cursor

📦 Install in Cursor

To open Cursor and automatically add the OpenFeature MCP, click the install button above.

Alternatively, navigate to Cursor Settings -> Tools & MCP -> New MCP Server and add to ~/.cursor/mcp_settings.json:

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

VS Code

📦 Install in VS Code

To open VS Code and automatically add the OpenFeature MCP, click the install button above.

Alternatively, add to .continue/config.json:

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

Claude Code

Add the server via the Claude Code CLI:

claude mcp add --transport stdio openfeature npx -y @openfeature/mcp

Then manage the connection with /mcp in the CLI.

Windsurf

In the Manage MCP servers raw config, add:

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

Codex CLI

Edit ~/.codex/config.toml:

[mcp_servers.openfeature]
command = "npx"
args = ["-y", "@openfeature/mcp"]

Restart Codex CLI after saving.

Gemini CLI

Edit ~/.gemini/settings.json:

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

Restart Gemini CLI after saving.

Claude Desktop

Edit your Claude Desktop config at:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the following configuration:

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

Restart Claude Desktop after saving.

Available Tools

The OpenFeature MCP Server provides two main tools accessible to AI assistants:

SDK Installation Guide: install_openfeature_sdk

Fetches installation instructions for OpenFeature SDKs in various languages and frameworks. Optionally includes provider-specific setup documentation.

SDK Tool Parameters

Parameter	Type	Required	Description
technology	string	Yes	Target language/framework (see supported list below)
providers	string[]	No	Provider identifiers to include installation instructions

Supported Technologies

The technologies list is built from the available prompts/*.md, updated automatically using scripts/build-prompts.js

Technology	SDK
android	Android Kotlin SDK
dotnet	.NET SDK
go	Go SDK
ios	iOS Swift SDK
java	Java SDK
javascript	JavaScript Web SDK
nestjs	NestJS SDK
nodejs	Node.js SDK
php	PHP SDK
python	Python SDK
react	React SDK
ruby	Ruby SDK

Supported Providers

The provider list is automatically sourced from the OpenFeature ecosystem (open-feature/openfeature.dev repo).

See scripts/build-providers.js for details on how the provider list is maintained.

OFREP Flag Evaluation: ofrep_flag_eval

Evaluate feature flags using the OpenFeature Remote Evaluation Protocol (OFREP). Supports both single flag and bulk evaluation.

OFREP Tool Parameters

Parameter	Type	Required	Description
base_url	string	No	Base URL of your OFREP-compatible flag service
flag_key	string	No	Flag key for single evaluation (omit for bulk)
context	object	No	Evaluation context (e.g., { targetingKey: "user-123" })
etag	string	No	ETag for bulk evaluation caching
auth	object	No	Authentication configuration
auth.bearer_token	string	No	Bearer token for authorization
auth.api_key	string	No	API key for authorization

OFREP Configuration

To use OFREP flag evaluation features, configure authentication and endpoint details. The server checks configuration in this priority order:

1. Environment Variables

   • OPENFEATURE_OFREP_BASE_URL or OFREP_BASE_URL
   • OPENFEATURE_OFREP_BEARER_TOKEN or OFREP_BEARER_TOKEN
   • OPENFEATURE_OFREP_API_KEY or OFREP_API_KEY

2. Configuration File: ~/.openfeature-mcp.json

Example ~/.openfeature-mcp.json:

{
  "OFREP": {
    "baseUrl": "https://flags.example.com",
    "bearerToken": "<your-token>",
    "apiKey": "<your-api-key>"
  }
}

You can override the config file path using the OPENFEATURE_MCP_CONFIG_PATH environment variable.

Note: All logs are written to stderr. The MCP protocol messages use stdout.

MCP Usage Examples

SDK Installation Example

"install the OpenFeature SDK for Node.js with the flagd provider"

The AI will use the MCP to fetch relevant installation instructions and attempt to install the OpenFeature SDK with the correct provider.

Flag Evaluation Example

When interacting with your AI assistant:

"Can you check the value of the 'new-checkout-flow' feature flag for 'user-123'?"

The AI will use the MCP to evaluate the flag using OFREP and provide you with the result, along with additional metadata like variant and reason.