codepulse

User guide

Getting started

First-time path: run the Python demo under codepulse, wire MCP, ask for hot paths.

Prerequisites

1. Clone and build the daemon

git clone https://github.com/detroitpro/codepulse.git
cd codepulse
cargo build -p codepulse-daemon

2. Start the daemon on the demo root

./target/debug/codepulse \
  --root examples/python-demo \
  --db .codepulse/dev.db \
  --listen 127.0.0.1:7420

Check: curl -s http://127.0.0.1:7420/health{"ok":true,...}

3. Run the Python demo with the agent

In another terminal:

python3 -m venv .codepulse/venv
source .codepulse/venv/bin/activate
pip install -e agents/python "fastapi>=0.115" "uvicorn>=0.32"

export CODEPULSE_ENDPOINT=http://127.0.0.1:7420
export CODEPULSE_ROOT="$PWD/examples/python-demo"
export CODEPULSE_SESSION_ID=demo

cd examples/python-demo
PYTHONPATH="$PWD/../../agents/python/src" python - <<'PY'
from codepulse_agent import install
install()
import uvicorn
from app import app
uvicorn.run(app, host="127.0.0.1", port=8000, log_level="info")
PY

Drive load:

python examples/python-demo/scenario.py

4. Wire the MCP server (Cursor)

cd packages/mcp && npm install && npm run build

Add to Cursor MCP settings (adjust path):

{
  "mcpServers": {
    "codepulse": {
      "command": "node",
      "args": ["/ABS/PATH/to/codepulse/packages/mcp/dist/index.js"],
      "env": {
        "CODEPULSE_ENDPOINT": "http://127.0.0.1:7420"
      }
    }
  }
}

5. First successful ask

curl -s "http://127.0.0.1:7420/v1/query/hot-paths?limit=5"

You should see PricingWorkflow.execute / inventory_reserve after the scenario. Or ask your agent: “What were the hottest functions in this session?”

One-command check

./scripts/e2e-python.sh