The deploy platform for AI-built software
SimDeploy is an AI deployment platform built for coding agents: every capability — analyze, deploy, read logs — exists on the CLI, the REST API and a native MCP server, so an agent can ship and debug software without clicking through a dashboard.
Code generated by agents needs a target that agents know how to operate. Every SimDeploy capability exists on four surfaces: dashboard, API, CLI and MCP — nothing critical depends on clicking through screens.
How it works
1. The agent runs simdeploy analyze --json and reads the recommended architecture 2. simdeploy deploy --yes publishes and returns a URL or a typed error 3. simdeploy logs feeds the agent's fix loop
- —Typed deployment states agents handle programmatically
- —API with structured errors {code, message} and per-token scopes
- —llms.txt, AGENTS.md and /docs/agents maintained as a contract
Why agents need a different target
Software generated by an agent needs a deploy target the agent can operate without a human in the loop. On SimDeploy every capability — analyze, estimate cost, deploy, read logs — exists on four surfaces: the dashboard, the REST API, the non-interactive CLI and a native MCP server. Nothing critical hides behind a screen that only a person can click. Deploys move through typed states the agent branches on, the API returns structured errors as {code, message}, and per-token scopes keep each agent's access narrow and revocable.
What runs today, and the contract
Static builds deploy end-to-end today; server-rendered and serverless targets are planned and return a typed error rather than a silent partial deploy, so an agent gets a clear signal. The machine-readable contract — llms.txt, AGENTS.md and /docs/agents — documents the exact flow and is maintained deliberately, so any agent can discover how to ship without trial and error.
| Step | Agent + generic host | Agent + SimDeploy |
|---|---|---|
| Choosing infrastructure | The agent (or you) guesses runtime, region and plan | The platform picks the lowest-cost compatible architecture from code analysis |
| Cost visibility | Discovered on the bill | Projected before the deploy, labeled an estimate |
| Reading a failed deploy | Parse free-form console output | Typed states plus structured {code, message} errors |
| Interface the agent uses | Click-through dashboard or scraped UI | CLI, REST API and native MCP — all non-interactive |
| Knowing when to retry | Ambiguous | Non-zero exit code and per-stage logs |
FAQ
What is an AI deployment platform?
It is a deploy target built to be operated by coding agents: every capability — analyze, deploy, read logs — exists on the CLI, the API and a native MCP server, so an agent can ship and debug without clicking through screens.
How do agents handle a failed deploy?
Deploys return typed states and the API returns structured errors as {code, message}. Agents branch on them and read per-stage logs to fix and retry.
Is there a machine-readable contract for agents?
Yes. llms.txt, AGENTS.md and /docs/agents document the flow and are maintained as a contract any agent can rely on.
What can SimDeploy deploy today?
Static builds ship today. Server-rendered and serverless targets are planned.