What it does

Claude Context is an MCP server that embeds your whole codebase into a vector database (Zilliz Cloud/Milvus), then exposes semantic search to any MCP-compatible agent. When Claude Code or Cursor needs context, it queries the index instead of loading entire directories into the prompt. The project ships as two npm packages—@zilliz/claude-context-core and @zilliz/claude-context-mcp—plus a VS Code extension.

The interesting bit

The “cost-effective” pitch is the real hook. The README explicitly frames this as a money-saver: dumping millions of lines into Claude’s context window every request is expensive, so pre-filtering via embeddings keeps token bills down. It’s a pragmatic admission that vector search is mostly plumbing here, not magic.

Key highlights

• Supports a laundry list of MCP clients: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, Gemini CLI, OpenAI Codex CLI, Qwen Code, Cline, Roo Code, Augment, Cherry Studio, and Void
• Requires Node.js 20+ and two external API keys: OpenAI for embeddings, Zilliz Cloud for the vector database
• Also offers a VS Code extension (zilliz.semanticcodesearch) for non-MCP use
• Sister project memsearch provides persistent markdown-based memory across sessions

Caveats

• Hard dependency on Zilliz Cloud; no self-hosted Milvus option is mentioned in the README
• OpenAI embedding costs are separate from whatever you’re already paying for the LLM
• The README is 90% configuration boilerplate for different editors, which suggests setup friction varies wildly by client

Verdict

Worth a look if you’re burning through context-window budget on a large monorepo and already use one of the supported agents. Skip it if you want local-only infrastructure or a single-click install.