Agent Cloud

Use Case

Cloud infrastructure built for autonomous agents.

Most cloud platforms assume a human account owner and a human setup flow. Agent Cloud flips that model. AI agents can sign up, receive a sandbox key, and provision VMs with no human in the loop until billing approval is needed.

The agent journey

Whether the agent is ChatGPT, Claude Code, a Cursor assistant, or a custom LangChain/CrewAI/AutoGen agent — the flow is the same:

# Agent discovers Agent Cloud via web search, docs, or skill
# Step 1: Sign up
POST /v1/agent/signup → API key + sandbox limits

# Step 2: Provision
POST /v1/instances → micro Ubuntu VM in minutes

# Step 3: Work
SSH into VM, install tools, run tasks

# Step 4: Clean up or upgrade
DELETE /v1/instances/{id}  or  ask human to approve Stripe checkout

The entire journey — from discovery to running server — can happen in a single conversation. No dashboard. No CLI install. No human signup.

Best fit workloads

  • Code execution sandboxes. Give your agent an isolated Linux VM to write, compile, and run code safely. The sandbox tier auto-expires, so forgotten VMs don't accumulate cost.
  • Short-lived task runners. Web scraping, data processing, file conversion, API testing — tasks that need a real machine for minutes or hours, not permanently.
  • Web research with isolated compute. Agents that need to run browsers, download files, or process large datasets benefit from having their own VM rather than running on the host.
  • Agent orchestration prototypes. Multi-agent systems where each agent gets its own VM for isolation, with a coordinator managing lifecycle via the API.
  • Development environments. An agent can provision a VM, clone a repo, install dependencies, run tests, and report results — all without touching the developer's local machine.

Why agents need their own infrastructure primitive

Today, when an AI agent needs compute, it either runs on the host machine (risky) or asks the human to set up a cloud resource (slow). Neither scales.

Agent Cloud gives agents a third option: self-serve, sandboxed, API-provisioned compute with built-in guardrails. The agent operates autonomously within safe limits, and the human only intervenes for billing decisions.

This matches how agents actually work — they're good at following API contracts and bad at navigating dashboards. Build for the strength, not around the weakness.

Compatible with every agent framework

Agent Cloud is a REST API. Any agent that can make HTTP requests can use it. No SDK required, no vendor lock-in, no special integration.

  • Claude Code / Cursor / Windsurf — paste the agent prompt or install the skill
  • ChatGPT with tools — point it at the OpenAPI spec
  • LangChain / CrewAI / AutoGen — wrap the API as a tool in your agent's toolkit
  • Custom agents — call the REST API directly with curl or any HTTP client

Get started

The sandbox is free, requires no credit card, and takes one API call to activate. Read the quickstart or compare plans.