Tutorial: Deploy Vant
Deploy Vant to production
Where
Local
# Full runtime (layout check + health)
vant start
# Long-running servers
vant server # REST, port 3456
vant mcp # MCP, port 3457
vant all # Both in one process
Docker
The repo ships a remote-ready Dockerfile (non-root, deps installed, healthcheck) - use it rather than hand-rolling:
docker build -t vant .
docker run -d --name vant \
-p 3456:3456 -p 3457:3457 \
-v vant-models:/app/models \
--restart unless-stopped \
vant
Default command is the full runtime (bin/vant.js all: REST 3456 + MCP 3457, HEALTHCHECK on GET /health). There is no vant serve or vant start --daemon command - run servers under your supervisor (systemd, pm2, docker restart policy).
Vercel
npm i -g vercel
vercel deploy
Fly.io
fly launch
fly deploy
Environment
Required:
GITHUB_TOKEN=ghp_xxx
GITHUB_REPO=owner/repo
Optional:
VANT_SERVER_PORT=3456
VANT_MCP_PORT=3457
VANT_MCP_REQUIRE_KEY=true
Health
Check deployment:
curl https://your-domain.com/health
Response:
{
"status": "ok",
"version": "0.8.6",
"uptime": 3600
}
Scale
Horizontal
Multiple instances:
# Each instance gets unique agent ID
VANT_AGENT_ID=agent-1 vant server
VANT_AGENT_ID=agent-2 vant server
VANT_AGENT_ID=agent-3 vant server
Vertical
Memory per instance:
# 512MB default
docker run -m 512m vant
# 2GB for large brains
docker run -m 2g vant