Overview
Claude Context is an MCP server from Zilliz, the company behind the Milvus vector database. It indexes your codebase for semantic search, so a coding agent can find relevant code across millions of lines without several rounds of searching. Despite the name, it works with any MCP client.
Key capabilities
- Semantic code search: retrieve relevant code from the whole repository.
- MCP server: works with Claude Code, Codex CLI, Gemini CLI and others.
- Cost control: avoids loading entire directories into the model for every request.
Tradeoffs
- Requires a vector database and an embedding API key, so there are extra services to set up.
- Your code is embedded via OpenAI and stored in the vector database you choose; check that fits your policies.
- The index must stay in sync with your code.
Ideal for
- Very large codebases
- Teams already using Milvus or Zilliz
- Agents that waste tokens exploring files
Not ideal for
- Small repos
- Code that can't be sent to a third-party embedding API
Getting started
- Create a Milvus database; Zilliz Cloud has a free tier and gives you an endpoint and API key.
- Get an OpenAI API key for the embedding model.
- Add the server to Claude Code:
claude mcp add claude-context -e OPENAI_API_KEY=... -e MILVUS_ADDRESS=... -e MILVUS_TOKEN=... -- npx @zilliz/claude-context-mcp@latest(Node.js 20+). - Ask the agent to index your codebase, then search it in plain language. The docs include configs for Codex CLI and Gemini CLI.
Technical notes
- Install:
npx @zilliz/claude-context-mcp@latestvia your MCP client config - Requires: Node.js 20+, Milvus/Zilliz, OpenAI API key
- License: MIT
- Source: zilliztech/claude-context
- GitHub stars: 12,688 (as of 2026-11-12)