Bucket Feature Flags MCP Server

What is Bucket Feature Flags MCP Server?

The Bucket Feature Flags MCP Server is a Model Context Protocol server that enables AI development tools to understand and interact with your Bucket feature flags. It creates a bridge between feature management workflows and AI-powered tools like Cursor IDE and Claude Desktop.

How to use Bucket Feature Flags MCP Server?

Run npx bucket mcp from your Bucket-initialized project directory. Optional --port (default 8050) and --app-id flags are supported. This starts an SSE server at http://localhost:8050/sse. For clients that don't support SSE, configure a STDIO proxy using supergateway.

Key features of Bucket Feature Flags MCP Server

• Starts a local SSE server for MCP integration.
• Supports both SSE and STDIO transport modes.
• Works with Cursor IDE and Claude Desktop.
• Enables AI to manage feature flags and access.
• Integrates directly with your Bucket project configuration.

Use cases of Bucket Feature Flags MCP Server

• AI-powered feature flag creation and management.
• Querying feature flag states during AI-assisted development.
• Granting or revoking feature access via natural language prompts.
• Generating TypeScript types for features automatically.

FAQ from Bucket Feature Flags MCP Server

What is the Model Context Protocol (MCP)?

MCP is an open protocol that provides a standardized way to connect AI models to data sources and tools. It enables your development environment to understand feature flags, their states, and their relationships within your codebase.

How do I set up the MCP server?

Run npx bucket mcp from a directory that has been initialized with bucket init. Then configure your client to connect to the server using either the SSE URL or STDIO proxy.

Which AI clients are supported?

The server works with Cursor IDE (using SSE configuration) and Claude Desktop (using STDIO proxy configuration). Other clients compatible with the Model Context Protocol may also work.

What are the runtime dependencies?

You need Node.js and the @bucketco/cli npm package installed. The server runs locally on your machine and starts an SSE server on a configurable port.

Are there any known limits or issues?

MCP is in very early development and changes are frequent. If something doesn't work, check the Model Context Protocol website and open an issue on the Bucket JavaScript SDK GitHub repository.