AI-native deploy platform
You build. AI chooses
where it runs.
SimDeploy analyzes your project, picks the lowest-cost compatible architecture and ships it in one command — for you and for your coding agent.
Works with Claude Code, Codex, Cursor and your terminal.
$ npx simdeploy deploy --yes SimDeploy OK Project detected: Vite (static) OK 128 files analyzed OK Architecture: static site Served on SimDeploy's own origin Estimated cost: USD 0.00/month (projection from price tables) Building OK Build completed Deploying OK Health check OK (HTTP 200) OK Deployment ready https://my-project.simdeploy.com
Works with your stack and your agent
What is SimDeploy?
SimDeploy is an AI-native deploy platform: it analyzes the project, automatically picks the lowest-cost compatible architecture, deploys it with a verified health check and meters real usage per project.
How it works
Ship in three steps
SimDeploy analyzes the project and picks the lowest-cost architecture it can ship — and shows the estimate before anything deploys.
01
Point it at your project
One command in your repo. The analyzer reads the framework, routes, data needs and static share — deterministic and fully offline. Nothing leaves your machine.
02
AI picks the architecture
The analyzer prices the compatible architectures and picks the lowest-cost one it can ship today (static). You see the estimate before anything deploys.
03
Deploy in one command
Build on your machine, publish only the output, verify with a real HTTP health check. A public URL at the end — or a typed error your agent can fix.
Start your way
Four ways in. One deploy out.
Deploy with a coding agent
Claude Code, Codex or Cursor drive the whole flow through the native MCP server.
simdeploy-mcpSet up your agent →Deploy from your terminal
One non-interactive command. Honest exit codes, JSON output, typed states.
npx simdeploy deploy --yesGet the CLI →Estimate before you ship
See the monthly cost projection offline — no account, nothing uploaded.
npx simdeploy analyze --jsonSee how →Wire it into your repo
Writes .mcp.json and an AGENTS.md deploy section so any agent knows how to ship.
npx simdeploy setupRead the guide →The mechanism: Autopilot Infrastructure
No questions about CPU, RAM, region or runtime. This is the full structure that runs on every deploy:
- 1
Your project
your code, exactly as it is
- 2
AI Project Analyzer
deterministic scan of the repository
- 3
Framework detection
Next.js, Vite, React, Astro, Express, Node, static
- 4
Requirements detection
API routes, database, cron, workers, storage, env vars
- 5
Resource estimation
static share, memory and timeout per architecture
- 6
Architecture selection
static, edge, serverless, hybrid or node-server — by compatibility
- 7
Cost estimation
monthly projection per provider, free allowances applied
- 8
Build
on your machine — code never runs on platform servers
- 9
Deploy
artifact upload, publish and verified HTTP health check
- 10
Monitoring
requests, bandwidth and storage measured per project
- 11
Usage report
measured consumption shown against your plan limits
01
Analyze
The AI Project Analyzer scans your project: framework, API routes, database, workers, static share. Deterministic and offline — nothing leaves your machine.
02
Route
The AICloudRouter computes the cost of every compatible architecture from provider price tables and picks the cheapest. You see the estimate before anything ships.
03
Deploy
Build on your machine, upload only the output, publish with a verified health check. A public URL at the end — or a typed error with logs to fix.
04
Measure
Real usage — requests, bandwidth and storage — is metered per project, so you and your agent see actual consumption against the plan's limits.
What runs today
Honest compatibility, no surprises
We show exactly what deploys today, what is in beta and what is on the roadmap — so your agent never promises a deploy the platform cannot finish.
SSR, serverless and multi-cloud are planned — the pipeline returns a clear, typed error when a project depends on them today.
Zero infrastructure decisions
No choosing CPU, RAM, region, runtime or provider. The platform picks the lowest-cost architecture compatible with your project.
Cost before the deploy
Every deploy shows the monthly projection per architecture, computed from public price tables — always labeled as an estimate, never a surprise.
Built for coding agents
Non-interactive CLI (--yes, --json), typed deployment states, structured per-stage logs and a native MCP server for Claude Code, Codex and Cursor.
Transparent pipeline
Full timeline for every deploy: analysis, queue, build, upload, publish and a real HTTP health check — with an event and log for each transition.
Secure by design
Your code never runs on the platform's servers: builds happen on your machine and only the output is published, validated before extraction.
Predictable pricing
Plans with explicit limits, real measured usage (requests and bandwidth) and Stripe billing. What is included is visible before you pay.
Agent-first
Built for agents to operate alone
Everything the dashboard does exists as API, CLI and MCP. JSON output, honest exit codes, typed failure states and per-stage logs — the agent publishes, diagnoses and fixes without human intervention.
{
"mcpServers": {
"simdeploy": { "command": "simdeploy-mcp" }
}
}
// tools: analyze_project, estimate_cost,
// deploy_project, get_deployment, get_logs,
// list_projects, get_project, create_projectPredictable pricing
Explicit limits, measured usage, no surprise bills.
Frequently asked questions
What do I need to do to publish a project?
Install the CLI, authenticate once and run simdeploy deploy at the project root. Analysis, architecture selection and publishing are automatic.
Which frameworks are detected?
Next.js, Vite, React (CRA), Astro, Express, generic Node and static sites. Deploys currently cover static builds (Vite, CRA, static Astro, Next with output export); SSR and serverless are planned — the pipeline fails with a clear, typed message when a project depends on them today.
How does it work with Claude Code, Codex or Cursor?
The CLI takes --yes and --json for automation, and there is an MCP server with tools for analysis, deploys and log reading. The agent publishes, reads the typed error if something fails and fixes it on its own.
How much does it cost?
There is a free plan to start. Paid plans have a fixed monthly price with explicit limits — and the infrastructure cost estimate appears before every deploy.
How does billing work?
Subscription via Stripe (card), with a portal to change plans or cancel at any time. Online purchases carry a 7-day right of withdrawal (article 49 of the Brazilian Consumer Code).
Built by you, or your agent
Start free. The cost estimate shows up before any deploy — and your coding agent can operate everything on its own.
Point your coding agent at simdeploy.com/docs/agents and let it ship for you.