# ContextStream MCP Server

A hosted MCP server that gives an AI coding agent persistent memory and millisecond code search over a project's code, docs and past conversations, with cited sources and scoped access.

ContextStream MCP Server connects a project's code, docs and conversations so an AI agent can find the right files, recall saved decisions, and build on past sessions instead of re-explaining context each time. It reports a 95 millisecond median for its published code-search benchmark.

It works with Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI and other MCP clients across several model providers, and offers a free tier of 10,000 monthly credits with no credit card. Indexing sends eligible source content to ContextStream for hosted search, with transcript saving and local Git metadata capture on by default but configurable. It installs via a shell script or an npm launcher that downloads a native Rust binary and verifies its checksum, and a hosted endpoint is also available over Streamable HTTP and OAuth.

Website: https://contextstream.io