MCP-RAG

What is MCP-RAG?

MCP-RAG is an AI-powered chat system that integrates multiple MCP servers (Filesystem, Slack, Brave-Search) to enable the Claude AI to retrieve and reference information from local documents, Slack conversations, and real-time web search based on user queries. It is built for users who want a unified assistant that can seamlessly access diverse data sources.

How to use MCP-RAG?

Install uv (see README for install commands), clone the repository, create a .env file with ANTHROPIC_API_KEY, SLACK_BOT_TOKEN, SLACK_TEAM_ID, and BRAVE_API_KEY, then set up a virtual environment and install dependencies with uv venv and uv sync. Run the client using uv run client.py path/to/dir/you/want/to/use.

Key features of MCP-RAG

• Integrates Filesystem, Slack, and Brave Search MCP servers.
• Claude automatically selects the right data source per query.
• No explicit instructions needed from the user.
• Access local folders, Slack history, and live web results.
• Simple setup with environment variables and one command to run.

Use cases of MCP-RAG

• Search and retrieve information from specified local folders.
• Access and reference Slack conversations and channels.
• Perform real-time web searches for current information.
• Use a single chat interface for documents, team chat, and internet queries.

FAQ from MCP-RAG

What are the prerequisites to run MCP-RAG?

You need to install uv and have Python available. API keys for Anthropic, Slack (bot token and team ID), and Brave Search are required.

How do I set up the required API keys?

Create a .env file and add ANTHROPIC_API_KEY, SLACK_BOT_TOKEN, SLACK_TEAM_ID, and BRAVE_API_KEY with your credentials.

Which local files can be accessed?

MCP-RAG can access any folder specified in the path/to/dir/you/want/to/use argument when running the client.

Does MCP-RAG store any data?

The README does not specify data storage. The system accesses files, Slack, and web services temporarily during query processing.

What transport or authentication does MCP-RAG use?

No transport or authentication details beyond the API tokens are provided in the README.