Connect Vibgrate AI Context to Cursor

This guide connects Vibgrate AI Context — the local, offline MCP server — to Cursor, so Cursor's AI answers from your real repository context instead of a frozen training set.

It is for Cursor users who want version-correct docs and code-map awareness in their editor.

Overview

Vibgrate AI Context serves three things over MCP: version-correct library docs, your code map, and your drift. Cursor supports MCP servers, so connecting them is a setup step plus a running server.

Step 1 — Install the wiring

From your project:

vg install

This writes the MCP configuration and advisory nudge for supported assistants — Cursor is explicitly supported. It is idempotent and repo-local, so re-running it is safe.

Step 2 — Start the local server

vg serve

Vibgrate AI Context starts locally and exposes its MCP tools. Nothing is uploaded; after a one-time local model download for semantic search the server is fully offline.

Step 3 — Ask Cursor grounded questions

With the server running, Cursor's AI can:

• Pull version-correct docs for the libraries in your lockfile via the Free Dev Docs Library.
• Traverse your code map to understand callers, callees, and structure.
• Read your DriftScore to reason about upgrade risk.

Keeping it current

Rebuild the code map after significant changes so served context matches your working tree:

vg build
vg status

Troubleshooting

• Cursor shows no Vibgrate tools. Verify the server is running and that vg install was run in this repository; re-run vg install to rewrite the config.
• Stale results. Run vg build to refresh the map.

Related

• Starting the local context server with vg serve
• Connecting Vibgrate AI Context to Claude and Claude Code
• Connecting Vibgrate AI Context to a generic MCP client