Optimized for Claude Code
Deploy with Claude Code
SimDeploy is the deploy platform built to be operated by Claude Code. Ask Claude to ship your project — it analyzes the code, picks the lowest-cost architecture, deploys it and hands back a public URL. If a deploy fails, Claude reads the typed error and fixes it.
Three steps
Set up the project
simdeploy setup writes .mcp.json (Claude Code), .cursor/mcp.json and an AGENTS.md deploy section. Claude Code now has the tools to analyze and deploy — no further config.
npx simdeploy setup
Authenticate once
A human sets SIMDEPLOY_TOKEN (or runs simdeploy login) in their own terminal — the token is scoped, revocable, and never pasted into the chat.
export SIMDEPLOY_TOKEN=sd_live_... # or: simdeploy login
Ask Claude to deploy
Claude runs the deploy, reads the public URL, and if anything fails it reads the typed error and per-stage logs and fixes it — on its own.
simdeploy deploy --yes
Why it works with Claude
Native MCP server
analyze_project, deploy_project, get_logs, estimate_cost and more — Claude calls them directly. Analysis runs offline; nothing leaves the machine until you deploy.
Typed states Claude can branch on
READY, BUILD_FAILED, DEPLOY_FAILED, ANALYSIS_FAILED — with a documented next step for each. Exit codes are non-zero on failure, so the agent knows when to fix.
One command, zero decisions
No CPU, RAM, region, runtime or provider. SimDeploy analyzes the project and picks the lowest-cost compatible architecture, showing the estimate first.
The repo teaches the next agent
The first deploy writes AGENTS.md into the project. Any agent that opens the repo afterwards knows to deploy here — the workflow travels with the code.
Add the MCP server
`simdeploy setup` does this for you. To do it by hand, add this to your project's .mcp.json:
{
"mcpServers": {
"simdeploy": { "command": "simdeploy-mcp" }
}
}The machine-readable contract lives at /llms.txt and the full agent guide at /docs/agents.
Frequently asked questions
Do I need to paste my token into the chat?
No. A human sets SIMDEPLOY_TOKEN or runs simdeploy login in their own terminal; the token is scoped, revocable and never goes into the chat.
What does simdeploy setup change in my repo?
It writes .mcp.json for Claude Code, .cursor/mcp.json and an AGENTS.md deploy section, so any agent that opens the repo afterwards knows how to deploy here.
What happens when a deploy fails?
The pipeline returns a typed state — BUILD_FAILED, DEPLOY_FAILED or ANALYSIS_FAILED — with per-stage logs and a non-zero exit code, so Claude can read the error and fix it on its own.