Context Engine

Context Engine is a retrieval stack for AI coding assistants, delivered as an MCP server. It indexes code files, dependencies, documentation, Git history and issues into a vector index, embeddings, a symbol graph and a memory store, then exposes them to the assistant through tools for semantic search, symbol navigation, cited answers and persistent memory, so developers stop explaining their codebase to the AI in every session. It is built for developers and teams using assistants such as Claude, Cursor, Windsurf, Copilot, Kiro or any MCP compatible client. The repo search tool finds code by meaning rather than keywords, the symbol graph tool traces callers, callees and definitions, cross repository search follows API boundaries from frontend to backend, and memory tools store and recall context across sessions. A context answer tool retrieves relevant code spans and generates an explanation with references to the implementation. The vendor cites search latency under 100 milliseconds, 16 IDE integrations and 32 languages. A self hosted mode, called Singular, runs the full stack on your own infrastructure so code never leaves the network. Prices are not published. The documentation lists Individual, Team, Business, Organization and Enterprise tiers defined by limits, from 1 member, 3 collections and 100 MB of storage on Individual up to unlimited members, collections and storage with SSO and audit export on Enterprise, but gives no amounts. Bring your own key support lets users route query expansion and answer generation to their own model provider in exchange for higher usage limits, though embedding and storage costs still apply.