Deploy an Agent

Quick Start (30 seconds)

# Install CLI
npm install -g @moltghost/cli

# Deploy your first agent
moltghost deploy my-agent \
  --model llama3.1-8b-q4 \
  --gpu l40s \
  --memory 24gb

# Get instant endpoint
echo "✅ Agent ready: https://$(moltghost status my-agent --url)/v1/chat"

🚀 Deployment complete in 45s
📱 Endpoint: https://abc123.agent.moltghost.io
💰 Cost: $0.0017/min (A100 paused)

Complete Deployment Flow

graph TD
    A[CLI/UI/API Deploy] --> B[Validate Config]
    B --> C[Provision Pod: 10-30s]
    C --> D[Container Start: 5s]
    D --> E[Runtime Init: 15s]
    E --> F[Ollama + Model Load: 30-180s]
    F --> G[Health Check]
    G -->|✅| H[Endpoint Active]
    G -->|❌| I[Auto-Retry]
    
    style H fill:#90EE90

End-to-End Timeline:

Deployment Command Reference

moltghost deploy AGENT_NAME [options]

# Core Options
--model llama3.1-70b-q4           # LLM (20+ supported)
--gpu l40s \| a100 \| h100        # Compute tier
--memory 24gb \| 80gb             # System RAM
--storage 50gb                    # Persistent volume

# Advanced
--region ap-southeast-1           # Jakarta (low latency)
--replicas 3                      # HA deployment
--auto-pause 15m                  # Cost optimization
--skills ./my-skills/             # Private tools
--env KEY=secret                  # Environment vars

# Examples
moltghost deploy sales-agent --model qwen2.5-72b --gpu h100 --region jakarta
moltghost deploy dev-chat --model phi3-mini --cpu --memory 16gb --free-tier

Step-by-Step Process

1. Configuration (Instant)

# Define agent spec
cat > agent.yaml
name: sales-assistant
model: llama3.1-70b-q4
resources:
  gpu: a100
  memory: 80gb
  storage: 100gb
skills: ["crm", "slack"]
env:
  CRM_API_KEY: secret

2. Pod Provisioning (10-30s)

✓ Allocating GPU (A100-80GB)
✓ Mounting 80GB RAM  
✓ Creating 100GB NVMe volume
✓ Internal networking (WireGuard)
Pod abc123 ready ✓

3. Runtime Initialization (15s)

✓ OpenClaw framework v1.2.3
✓ Ollama server starting
✓ Skills loaded: crm_query, slack_notify
Runtime healthy ✓

4. Model Loading (30-180s)

⬇️ Downloading llama3.1-70b-q4... 42GB / 42GB
🔄 Loading to GPU memory (Q4_K_M 38GB)
⚡ Warming inference engine
Model ready ✓ 1.2s/token

5. Endpoint Activation (Instant)

✅ Agent fully deployed!
📱 https://abc123.agent.moltghost.io
🔗 OpenAI compatible: /v1/chat, /v1/completions
💰 Current rate: $0.0417/hour

Deployment States

Real-time Status:

moltghost status my-agent --watch
# 14:32:15  Initializing (87%)  Model loading 72%
# 14:32:45  ✅ Healthy  Endpoint active  Uptime 0m

Production Deployment Patterns

High Availability

moltghost deploy prod-agent \
  --replicas 3 \
  --region ap-southeast-1 \
  --auto-scale cpu>80 \
  --healthcheck /v1/health

Cost-Optimized Development

moltghost deploy dev-agent \
  --model phi3-mini-q4 \
  --cpu \
  --memory 16gb \
  --auto-pause 10m \
  --backup daily

Enterprise Multi-Region

moltghost deploy global-agent \
  --model llama3.1-405b \
  --gpu h100 \
  --replicas 2 \
  --region ap-southeast-1 \
  --region us-east-1 \
  --backup cross-region

Troubleshooting

Summary

Deploy → Ready in <3 minutes with production-grade infrastructure.

✅ One-command deployment (CLI/UI/API)
✅ 30s-5m total time (model dependent)
✅ Automatic HTTPS endpoints
✅ Cost controls built-in
✅ Production patterns (HA, multi-region)

From zero to intelligent agent in one command.

Next: Manage Agents → Scale, pause, update

Test Immediately:

curl -X POST https://YOUR-AGENT.agent.moltghost.io/v1/chat \
  -d '{"messages":[{"role":"user","content":"Hello!"}]}'