learningOS-MCP

FastMCP server that bridges Claude (Desktop / IDE / Colab) with a master_knowledge_graph.md file stored on Google Drive. Deployed to GCP Cloud Run with scale-to-zero.

Prerequisites

Tool	Purpose
uv	Python package & project manager
gcloud CLI	Deploy to Cloud Run, proxy tunnelling, logs
Docker	Local container builds (optional — Cloud Run builds from source)

Setup

# 1. Install dependencies
uv sync

# 2. Copy env template and fill in your values
cp .env.example .env

# 3. Authenticate with GCP
gcloud auth login
gcloud auth application-default login

Environment variables

See .env.example:

Variable	Description
GCLOUD_PROJECT_ID	GCP project ID
SERVICE_ACCOUNT	Service-account email bound to the Cloud Run revision
GDRIVE_FOLDER_ID	Google Drive folder containing the knowledge graph
GDRIVE_FILE_ID	File ID of master_knowledge_graph.md

Deploy

make deploy

Useful commands

make logs    # tail Cloud Run logs
make proxy   # tunnel Cloud Run service to localhost:8080