AgentCloudThe agent-native cloud

Take source code to a live HTTPS URL — with managed Postgres, Redis, and Mongo — in a single verb. Built for autonomous agents, usable by humans.

Quickstart

For Agents

API Reference

Open the Dashboard

🚀

Source to URL, zero config

Push a directory. AgentCloud autodetects the stack (Node, Python, Go, Rust…) with railpack, builds it, health-gates the rollout, and hands back a live HTTPS URL.

🗄️

Managed datastores

conway add postgres/redis/mongo provisions an engine, generates credentials, and injects the connection URL into every service. Browse and query them right from the API.

🤖

Agent-native by design

One token plus a fetchable skill.md is the entire onboarding. Identical semantics across REST, CLI, and an MCP server — so an agent deploys in a single tool call.

🔐

Scoped, auditable, isolated

Capability-scoped tokens with safe delegation, an append-only audit log, per-project isolation, and structured errors that tell you exactly what to do next.

What is AgentCloud?

AgentCloud is a platform-as-a-service that an AI agent can drive end to end with no human in the loop. You give it source code; it gives you a running service on a public URL, with the databases and environment that service needs.

It runs on bare-metal Kubernetes (Talos Linux), builds with BuildKit + railpack, serves through a Caddy edge with automatic TLS, and is programmed entirely through one REST API — exposed identically as a CLI, an MCP server, and this documentation's companion contract, skill.md.

conway init my-app
conway deploy            # → https://app-my-app.apps.agentcloud.ac
conway add postgres      # DATABASE_URL injected automatically

Everything you can do here, an agent can do through the same API — that's the whole point.

Building an agent?

Fetch /skill.md (the machine contract) or read the For Agents page. One token + that file is the entire onboarding. A terse index for LLMs lives at /llms.txt.