Context Lens
Index local project and reference files into a LanceDB-backed semantic knowledge base so agents answer from your corpus instead of hallucinating.
Overview
Context Lens is a MCP server for the Build phase that provides a local semantic search knowledge base over 25+ file types using serverless LanceDB.
What is this MCP server?
- Serverless LanceDB semantic index—advertised as zero setup
- 100% local processing for privacy-sensitive repos
- Supports 25+ file types for ingestion into the knowledge base
- PyPI context-lens v0.1.8 with stdio MCP transport
- Pairs naturally with Bookmark Lens for URLs vs files split in Cornel Croi’s toolkit
What problem does it solve?
Agents invent details when your specs and repo docs are too large to fit in context windows every session.
Who is it for?
Privacy-conscious indie devs building agent workflows who want LanceDB RAG on disk without a cloud vector DB bill.
Skip if: Builders who only need a single file in chat or require enterprise multi-tenant hosted search.
What do I get? / Deliverables
Your agent queries a local embedding index so answers cite indexed project files across builds and follow-up iterations.
- Semantic search MCP tools over indexed local files
- Grounded agent answers tied to 25+ supported file types
Recommended MCP Servers
Journey fit
Context retrieval is wired while building agent features and integrations; it also supports review and iteration when codebases and specs grow. agent-tooling is the right shelf for a stdio MCP that turns 25+ file types into embeddings-backed search for coding agents.
How it compares
Local file RAG MCP server, not a hosted documentation crawler or generic web search tool.
Common Questions / FAQ
Who is Context Lens for?
Solo builders and agent users who need semantic search over local project files with LanceDB and stdio MCP.
When should I use Context Lens?
Use it while building and extending agent tooling whenever grounded retrieval from specs, code, and docs beats pasting files manually.
How do I add Context Lens to my agent?
Install PyPI package context-lens (v0.1.8), add the stdio MCP server definition to your client, and configure indexed paths per the context-lens GitHub README.