khalidsaidi/ragmap
MapRag is a RAG-focused subregistry and MCP server designed to discover and route to retrieval-capable MCP servers. It acts as a specialized directory within the Model Context Protocol (MCP) ecosystem, allowing AI assistants and agents to dynamically locate the most suitable retrieval server based on structured constraints. By specifying criteria such as required data sources, latency requirements, or capability filters like citation support or local-only execution, MapRag intelligently routes retrieval requests to the optimal server. This ensures efficient and context-aware offloading of retrieval tasks, reducing the overhead of managing multiple MCP servers manually. The server exposes an MCP interface itself, making it directly usable by any MCP-compatible client, such as Claude Desktop or other AI applications. It provides explainable ranking, giving transparency into why a particular server was selected. The system continuously monitors server reachability and updates its registry to reflect availability and capabilities. A browse UI is included for manual exploration and configuration, and developers can filter servers by various properties. Built as a monorepo with a modular structure, MapRag includes packages for core functionality, apps for web-based interfaces, and scripts for ingestion and maintenance. It leverages Firebase for hosting and data persistence, and is designed to run in diverse environments, including cloud, local, and edge deployments. The project is open source, encouraging community contributions and customization, and it integrates with existing MCP discovery mechanisms like Glama for expanded server coverage. MapRag solves the fragmentation problem in the growing MCP landscape by providing a centralized, intelligent routing layer for retrieval-augmented generation tasks. It simplifies the developer experience, enhances the reliability of AI systems that depend on real-time information retrieval, and offers a scalable approach to managing
- Category
- Developer Tools
- Pricing
- Open Source
Problem solved
As the MCP ecosystem expands, developers and AI agents face the challenge of discovering and selecting the most appropriate retrieval-capable MCP server for a given query. Without a unified discovery and routing mechanism, agents must either hardcode server choices or manually manage connections, leading to inefficiency and reduced flexibility. MapRag solves this by providing a dynamic registry that indexes retrieval servers, evaluates them against structured constraints, and routes requests with explainable ranking, enabling seamless and optimal retrieval integration for AI-powered applications.
Use cases
- An AI coding assistant dynamically retrieves documentation from a vendor-specific MCP server by specifying the vendor as a constraint in MapRag, ensuring up-to-date API references.
- A research agent queries MapRag to find an MCP server that indexes academic papers, filters by availability of citation support, and routes the request to fetch relevant studies.
- A privacy-focused chatbot routes user queries to a local, on-device RAG server for personal data retrieval, using MapRag's 'localOnly' constraint to avoid cloud services.
- An enterprise AI system enforces policy-based routing by configuring MapRag to only select MCP servers that meet security and compliance requirements.
- During an outage, MapRag automatically bypasses unreachable servers and reroutes requests to healthy alternatives with similar capabilities, maintaining service continuity.
- Developers use the browse UI to explore available retrieval servers, inspect their capabilities, and test constraints before integrating them into an agent.
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
Pros
- Explainable ranking provides transparency into server selection.
- Structured constraints allow fine-grained routing based on data sources, latency, and capabilities.
- Continuous reachability checks keep the registry accurate and reliable.
- Cross-platform support, including local and cloud deployments.
- Open source and extensible, encouraging community-driven enhancements.
Cons
- Relatively early stage with limited community adoption (as of May 2026).
- Dependency on the MCP ecosystem; limited utility outside MCP-compatible agents.
- May require initial configuration effort to define meaningful constraints and integrate with custom servers.
Official website: https://github.com/khalidsaidi/ragmap
Canonical LambdaBase page: https://www.lambdabase.com/tools/khalidsaidi-ragmap