Hey HN - we built Morphik MCP to solve a common problem: finding specific information across scattered technical docs. We've experimented with GraphRAG, ColPali, contextual embeddings, and more. MCP emerged as the solution that unifies these approaches.

Features: - Multimodal search across text, diagrams, and videos - Natural language knowledge base management - Fully open-source with responsive support

What sets MCP apart is its ability to return images (including diagrams) directly to the MCP client. Users have applied it to search over data ranging from blood tests to patents, and we use this daily with Cursor and Claude.

Give it a spin, and let us know what you think.

Link to our repo: https://github.com/morphik-org/morphik-core