Engram - MCP Server for AI Dev Agents

Engram provides a live, queryable knowledge graph of a codebase that AI agents can use to retrieve context without a warm‑up phase. It indexes repository files, monitors changes, and offers a set of MCP tools that supply error traces, pull‑request diffs, test coverage, and impact analysis in milliseconds, reducing token consumption by roughly sixty percent. Persistent episode memory stores past interactions, allowing agents to recall relevant solutions across sessions and transform informal prompts into structured instructions.

The platform targets developers who integrate AI agents into their development workflow, especially those needing rapid, context‑aware code fixing, refactoring, or feature implementation. By exposing a dashboard and language‑server protocol integration, it lets agents and users query the graph, view recent activity, and orchestrate tasks such as error‑to‑fix pipelines and multi‑repo navigation.

Distinctive aspects include a diffusion‑style “live snapshot” of mixed‑language projects, incremental re‑indexing that keeps the graph up to date, and a tiered toolbox that combines memory, orchestration, and prompt intelligence. The system is designed to start locally and scale to team environments while maintaining low latency (e.g., loading eight files in 167 ms).