pulsemcp/content/start-dev-server

Start Dev Server

Start the Linear Clone demo application for local development. Handles both agent container and standalone clone environments.

Checklist

• [ ] Detect environment (agent container vs standalone clone)
• [ ] Check if dev server is already running
• [ ] Start Docker Compose services (if standalone) or run.sh (if agent container)
• [ ] Run setup (install deps, migrate, seed)
• [ ] Wait for health check to pass
• [ ] Report status with URLs and log locations

Pre-requisites

• [ ] Docker and Docker Compose are available (standalone clone only)
• [ ] The linear-clone-demo/ directory exists in the repo

If any pre-requisite is not met, STOP immediately and tell the user.

Procedure

Step 1: Detect environment

if [ -f /app/.agent-containers/run.sh ]; then
  echo "agent-container"
else
  echo "standalone-clone"
fi

• Agent container: Source code is already mounted at /app, database is running, deps may be installed.
• Standalone clone: Need to start Docker Compose from the host.

Step 2: Check if already running

Check if the dev server is already up:

curl -sf http://localhost:3000/health

If the health check passes, skip to Step 5 (report status). Do not restart a healthy server.

Also check for PID files:

• Agent container: /app/.logs/app.pid
• Standalone: check docker compose ps output

Step 3: Start services

Agent container path

cd /app
bash .agent-containers/setup.sh
bash .agent-containers/run.sh

Standalone clone path

Run from the linear-clone-demo/ directory:

cd linear-clone-demo
 
# Start Docker Compose
docker compose -f .agent-containers/docker-compose.dev.yml -p linear-clone-dev up -d --build
 
# Wait for db to be healthy
docker compose -f .agent-containers/docker-compose.dev.yml -p linear-clone-dev exec app bash -c "until pg_isready -h db -U app 2>/dev/null; do sleep 1; done"
 
# Run setup inside the container
docker compose -f .agent-containers/docker-compose.dev.yml -p linear-clone-dev exec app bash .agent-containers/setup.sh
 
# Start the dev server
docker compose -f .agent-containers/docker-compose.dev.yml -p linear-clone-dev exec -d app bash .agent-containers/run.sh

Use docker compose -f .agent-containers/docker-compose.dev.yml -p linear-clone-dev port app 3000 to discover the mapped host port.

Step 4: Wait for health check

Poll the health endpoint until it responds:

for i in $(seq 1 24); do
  if curl -sf http://localhost:${HOST_PORT}/health > /dev/null 2>&1; then
    echo "Health check passed"
    break
  fi
  sleep 5
done

• For agent containers, HOST_PORT is 3000
• For standalone clones, discover it via docker compose port app 3000
• Timeout after 120 seconds (24 attempts x 5s)

If health check fails after all retries, check logs and report the error.

Step 5: Report status

Report to the user:

• App URL: http://localhost:${HOST_PORT}
• Health: http://localhost:${HOST_PORT}/health
• API: http://localhost:${HOST_PORT}/api/issues
• Logs: .logs/app.log (agent container) or docker compose logs (standalone)

Example: "Dev server running at http://localhost:32771. Health check passing. 5 seeded issues available."

Output

• Running Docker Compose stack with PostgreSQL and the Node.js app
• Health endpoint responding at /health
• Seeded database with sample issues
• Status summary with URLs

Important

• Never use run_in_background: true for the docker compose or server start commands -- they need to complete in the foreground so you can check for errors.
• Always check if the server is already running before starting. Don't restart a healthy server.
• For standalone clones, always use -p linear-clone-dev as the compose project name for consistency.
• If the port is dynamically assigned (standalone), always discover it via docker compose port rather than hardcoding.
• If setup.sh or migrations fail, check if the database is ready. The depends_on with health check should handle this, but occasionally the first connection attempt races the readiness check.