Claude Context Freemium

What it does

Claude Context is a Model Context Protocol server that gives coding agents semantic search over an entire codebase, so they can find relevant code by meaning ("functions that handle user authentication") instead of grepping or loading whole directories into context turn after turn.

What is inside

• Hybrid search combining BM25 keyword matching with dense vector search for natural-language queries
• AST-based code chunking for structure-aware indexing across many languages
• Incremental indexing via Merkle trees, so only changed files get re-indexed
• Reports roughly 40% lower token usage than loading full directories for context
• A companion VSCode extension (Semantic Code Search) and a standalone @zilliz/claude-context-core package for direct integration into other tools

Pricing

The MCP server itself is free and MIT-licensed. Running it costs whatever your chosen embedding provider (OpenAI, VoyageAI, Ollama or Gemini) charges per embedding call, plus a vector store: either self-hosted Milvus, which is free and open source, or Zilliz Cloud's managed service, which offers a free tier before usage-based billing kicks in. There is no separate fee for Claude Context beyond those dependencies.

Works with

Claude Code (official integration), Claude Desktop, Cursor, Windsurf, Cline, Codex CLI, Gemini CLI, and VSCode via its MCP extension.

How to install or connect

For Claude Code:

claude mcp add claude-context \
  -e OPENAI_API_KEY=sk-your-openai-api-key \
  -e MILVUS_ADDRESS=your-zilliz-cloud-public-endpoint \
  -e MILVUS_TOKEN=your-zilliz-cloud-api-key \
  -- npx @zilliz/claude-context-mcp@latest

Requires Node.js 20+ and a vector store — either self-hosted Milvus or Zilliz Cloud.

Maintenance and safety

MIT licensed, built and maintained by Zilliz (the team behind Milvus), with 12.6k+ stars, 930+ forks and 217+ commits on master. Core indexing runs locally; code embeddings are sent to whichever embedding provider and vector store you configure, so self-hosting Milvus keeps everything on your own infrastructure if that matters for your codebase's confidentiality requirements.

Who should use it

Developers working in large or unfamiliar codebases who want their coding agent to retrieve relevant code by meaning rather than repeated file exploration — the token savings and Merkle-tree incremental re-indexing matter most on big, actively changing repos. Teams with strict data-residency needs should lean toward self-hosted Milvus and a local embedding model rather than a cloud embedding API, since code snippets are sent to whichever provider you configure.