Argus Deployment Guide¶

This guide covers deploying the full Argus E2E Testing Agent stack.

Architecture Overview¶

┌─────────────────────────────────────────────────────────────────────────┐
│                           DASHBOARD (Vercel)                             │
│                      Next.js + Tailwind + Clerk Auth                     │
└─────────────────────────────────────┬───────────────────────────────────┘
                                      │
                    ┌─────────────────┼─────────────────┐
                    │                 │                 │
                    ▼                 ▼                 ▼
┌───────────────────────┐  ┌───────────────────────┐  ┌───────────────────┐
│    ARGUS BRAIN        │  │   BROWSER WORKER      │  │    MCP SERVER     │
│  (Railway/Render)     │  │   (Cloudflare)        │  │   (Cloudflare)    │
├───────────────────────┤  ├───────────────────────┤  ├───────────────────┤
│ • LangGraph Orchestr. │  │ • Playwright Browser  │  │ • AI IDE Integr.  │
│ • Code Analysis       │  │ • Self-Healing        │  │ • Claude Code     │
│ • Test Planning       │  │ • Screenshot Capture  │  │ • Cursor/Windsurf │
│ • Visual AI           │  │ • Pattern Learning    │  │                   │
│ • Report Generation   │  │ • Smart Waits         │  │                   │
└───────────────────────┘  └───────────────────────┘  └───────────────────┘
        Python                    TypeScript                TypeScript

Components¶

1. Deploy Argus Brain (Python Backend)¶

The Brain handles orchestration, code analysis, and advanced AI features.

Option A: Railway (Recommended)¶

# 1. Install Railway CLI
npm install -g @railway/cli

# 2. Login to Railway
railway login

# 3. Create a new project
railway init

# 4. Deploy
railway up

# 5. Set environment variables
railway variables set ANTHROPIC_API_KEY=sk-ant-...
railway variables set BROWSER_WORKER_URL=https://argus-api.samuelvinay-kumar.workers.dev

# 6. Get your deployment URL
railway status
# Example: https://argus-brain-production.up.railway.app

Option B: Render¶

# 1. Connect your GitHub repo to Render
# 2. Create a new Web Service
# 3. Select Docker deployment
# 4. Set environment variables in Render dashboard
# 5. Deploy

Or use the Blueprint:

# render.yaml is already configured
# Just connect repo and deploy

Option C: Fly.io¶

# 1. Install Fly CLI
curl -L https://fly.io/install.sh | sh

# 2. Login
fly auth login

# 3. Launch (uses fly.toml config)
fly launch

# 4. Set secrets
fly secrets set ANTHROPIC_API_KEY=sk-ant-...

# 5. Deploy
fly deploy

Option D: Local Development¶

# 1. Install Python dependencies
pip install -e .

# 2. Set environment variables
export ANTHROPIC_API_KEY=sk-ant-...
export BROWSER_WORKER_URL=https://argus-api.samuelvinay-kumar.workers.dev

# 3. Run the server
python -m uvicorn src.api.server:app --host 0.0.0.0 --port 8000

2. Connect Dashboard to Brain¶

Update your dashboard environment:

# dashboard/.env.local

# Point to your deployed Brain
ARGUS_BACKEND_URL=https://your-argus-brain.railway.app

# Browser Worker (already deployed)
E2E_WORKER_URL=https://argus-api.samuelvinay-kumar.workers.dev

Then redeploy the dashboard:

cd dashboard
vercel --prod

3. Verify Deployment¶

Test Brain Health¶

curl https://your-argus-brain.railway.app/health
# Should return: {"status": "healthy", "version": "0.1.0", ...}

Test Browser Worker Health¶

curl https://argus-api.samuelvinay-kumar.workers.dev/health
# Should return: {"status": "ok", ...}

Test Full Integration¶

In the dashboard chat, type:

Check system status

You should see both components connected: - Argus Brain (Python/LangGraph): connected - Argus Worker (Cloudflare/Browser): connected

Environment Variables Reference¶

Argus Brain (Python)¶

Dashboard (Next.js)¶

Browser Worker (Cloudflare)¶

Set via wrangler secret put:

Deployment Checklist¶

• Deploy Argus Brain to Railway/Render/Fly.io
• Set all required environment variables
• Test Brain health endpoint
• Update dashboard with Brain URL
• Redeploy dashboard
• Test full integration via chat
• Verify MCP Server works with Claude Code

Troubleshooting¶

Brain Not Connecting¶

1. Check if Brain is deployed: curl $ARGUS_BACKEND_URL/health
2. Verify CORS is enabled (should be by default)
3. Check environment variables are set correctly

Browser Actions Failing¶

1. Check Worker health: curl $E2E_WORKER_URL/health
2. Verify Anthropic API key is set in Worker
3. Check Cloudflare dashboard for errors

MCP Server Not Working¶

1. Verify MCP config in Claude Code/Cursor
2. Test SSE endpoint: curl https://argus-mcp.samuelvinay-kumar.workers.dev/sse
3. Check MCP server logs in Cloudflare dashboard

Cost Considerations¶

Total infrastructure cost: ~$5-20/month (excluding API usage)

Scaling¶

Horizontal Scaling¶

• Brain: Railway/Render auto-scale based on traffic
• Workers: Cloudflare edge locations globally
• Dashboard: Vercel edge functions

Performance Tips¶

1. Enable self-healing caching in Browser Worker
2. Use Vectorize for pattern learning (already configured)
3. Set appropriate timeouts for long-running tests
4. Use background jobs for large test suites