Semantic code search, with a hosted trade-off

Claude Context adds semantic code search to Claude Code through MCP. It also sends your code to two other services, so read the setup before you paste a key.

The repo is zilliztech/claude-context, MIT licensed, about 12.6k stars. The idea: index your codebase into a vector database, then let the agent search it by meaning instead of reading folder after folder. The README says this is cheaper on large codebases than loading whole directories into context.

The quick start needs three things: an OpenAI API key for embeddings, a Zilliz Cloud account for the vector database, and Node 20 or later. You register it with claude mcp add claude-context, pass your keys as environment variables, and point it at npx @zilliz/claude-context-mcp@latest.

Two trade-offs to weigh. With that setup your code text goes to an embedding provider and a hosted vector database, so check that your employer or client allows it. And a vector index finds code that looks similar to the question, which is different from knowing who calls what.

Last week I covered codebase-memory-mcp, a local graph index that makes the opposite trade. Which one would your security team accept?