mcp-redis

The Redis MCP Server exposes a natural language interface for querying and manipulating Redis data via the MCP protocol. It supports a broad set of Redis data types (strings, hashes, lists, sets, sorted sets, streams) and provides tools to perform common data operations in a conversational way. You can issue natural language requests like storing configuration in a string, indexing or caching vector data, managing sessions with expirations, or performing vector search via the built-in query engine. In addition, there are auxiliary tools for documentation lookup and server management, allowing you to search Redis docs, manage vector indexes, and retrieve database information directly through MCP-enabled clients. The server is designed to work with standard MCP clients over stdio transport, enabling easy integration with AI agents and orchestration frameworks.

To use it, run the server (via uvx from PyPI) and connect your MCP client. The included tools map to Redis operations such as string manipulation, hash operations (including embedding storage), lists for queues, sets for unique value tracking, and streams for event logging. You can also leverage the docs search tool to ask questions about Redis best practices or specific commands, and the query engine to manage and search vector indexes. When querying, frame requests in natural language and rely on the MCP layer to translate them into the appropriate Redis commands and return structured results.

Prerequisites:

• Python 3.14+ installed on your system
• Internet access to install PyPI packages

1. Create and activate a Python environment (optional but recommended)

• On macOS/Linux:
  • python3 -m venv venv
  • source venv/bin/activate
• On Windows:
  • python -m venv venv
  • venv\Scripts\activate

2. Install the Redis MCP Server from PyPI

pip install redis-mcp-server

3. Start the MCP server using uvx (the recommended workflow from PyPI)

uv run redis-mcp-redis

Note: The exact command may vary depending on your setup. The README example demonstrates using uvx with a Redis MCP server image; in practice, you would start the installed package and ensure your Redis instance is reachable at the configured URL.

4. Connect an MCP client

• Use any MCP-compatible client to connect via stdio and start issuing natural language requests to Redis through the MCP server.

Optional (Docker):

• If you prefer Docker, you can build or pull a container image that bundles the Redis MCP Server and a Redis instance, then run it with appropriate environment variables for URL, authentication, and ACLs.

Tips and common issues:

• Ensure Redis is reachable at the configured URL (default redis://localhost:6379/0). If you use authentication, update the URL or provide credentials via environment/config.
• The server supports a wide range of Redis data types; when asking for complex operations (like indexing or embeddings), ensure your data is stored in a compatible format (e.g., hashes with embedding fields or vectors in a suitable structure).
• If you enable EntraID/AAD authentication, configure the necessary Azure credentials in environment variables or the Redis ACLs to avoid access issues.
• The MCP stdio transport is the default; if you plan to use HTTP-based or streaming transports in the future, monitor the project for updates on transport support.
• For debugging, check the server logs for command translations and Redis responses to fine-tune your natural language prompts.