Skip to main content
MCP server CX Codex

engrava — Claude Code MCP server: what it does, how to install

An MCP server that gives AI agents a persistent SQLite memory database. It enables hybrid retrieval, MindQL queries, and a thought graph over stdio.

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add --env ENGRAVA_DB_PATH=<ENGRAVA_DB_PATH> --env ENGRAVA_MCP_CONFIG=<ENGRAVA_MCP_CONFIG> --env ENGRAVA_MCP_READ_ONLY=<ENGRAVA_MCP_READ_ONLY> --transport stdio engrava -- uvx engrava-mcp
Open on GitHub

What engrava does

The server exposes persistent agent memory backed by SQLite over stdio. A minimal start needs only ENGRAVA_DB_PATH pointing to a database file. In this setup, semantic vector search remains inactive because no embedding model is configured, though keyword search and graph queries function normally.

Activating vector retrieval requires ENGRAVA_MCP_CONFIG pointing to an engrava.yaml file. The server also supports ENGRAVA_MCP_READ_ONLY to disallow write tools. Even with write tools disabled, the process still requires filesystem write access at startup to create or upgrade the SQLite schema.

What it helps with

  • persist agent thoughts across sessions in SQLite
  • query past context using MindQL and text search
  • link memory nodes inside a thought graph
  • run the memory store in read-only mode to prevent edits

Example prompts for the agent

  • Save the current architectural decisions to the Engrava memory store
  • Query memory using MindQL to find earlier database migration notes
  • Inspect graph connections for the AuthController node

How to install engrava

Platform CLI

claude mcp add --env ENGRAVA_DB_PATH=<ENGRAVA_DB_PATH> --env ENGRAVA_MCP_CONFIG=<ENGRAVA_MCP_CONFIG> --env ENGRAVA_MCP_READ_ONLY=<ENGRAVA_MCP_READ_ONLY> --transport stdio engrava -- uvx engrava-mcp

Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from engrava-mcp <command>

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about engrava

How to install engrava in Claude Code?

Run claude mcp add engrava -e ENGRAVA_DB_PATH=<ENGRAVA_DB_PATH> -- uvx engrava-mcp in your terminal. To provide an embedding configuration, pass ENGRAVA_MCP_CONFIG pointing to your YAML file as well.

What does engrava do?

It runs an MCP server that exposes Engrava's persistent agent memory. Agents can store thoughts, query them via MindQL, and navigate contextual graph edges.

Why is semantic search inactive with just ENGRAVA_DB_PATH?

A bare database path does not declare an embedding provider. Vector scoring requires an engrava.yaml supplied through the ENGRAVA_MCP_CONFIG environment variable.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor