AI
Availability: AvailableDeploy your Python agent, get a live endpoint
Write handler.py in the in-portal Cloud Editor, click Deploy, and your agent serves POST /invoke on a runtime we pin and roll for you.
What you get
Handler to endpoint
Your handler.py runs on a managed runtime and answers POST /invoke, with no Dockerfile or server code to maintain.
In-portal Cloud Editor
Edit handler.py, env vars, and model keys in a browser code editor, then Deploy without leaving the portal.
Agents launchpad
A sidebar entry lists every agent across all your projects and deep-links straight into each one's editor.
Bring your own model key
Point an agent at OpenAI, OpenRouter, Anthropic, Google AI Studio, or Vertex, or call the platform model with a minted sk- key.
Deploy from your own CI
Ship the same agent from your pipeline instead of the editor when you want deploys in source control.
Keys stay in your project
Provider keys you bring live only in your own project secret and are never shared across tenants.
How it works
- 01
Write handler.py
Open the Cloud Editor, write a Python handler.py, and set the env and model keys your agent needs.
- 02
Deploy
Click Deploy, or push from your own CI; the agent starts as a managed container service serving POST /invoke.
- 03
Invoke and iterate
Call POST /invoke from your app, find the agent in the launchpad, and redeploy edits anytime.
Pricing
Each service uses its published meter against prepaid credit: provisioned time, storage, tokens, or another listed unit. Estimate any setup in the calculator.
Open the pricing calculator- ComputeAn agent bills as container compute at the same rate as Containers.
- $6 / vCPU-mo plus $3 / GB-mo
- Platform inference tokensCharged only when the agent calls the platform model; bring-your-own-key is billed by your own provider.
- $0.10 / $0.30 per 1M in/out
- Welcome creditPrepaid uUSD balance, with $5 applied to new accounts.
- $5
Questions
What language and interface does my agent use?
Python. You write a handler.py that serves POST /invoke. Other languages, general serverless functions, and long-running background daemons are not supported.
Which models can my agent call?
Bring your own key for OpenAI, OpenRouter, Anthropic, Google AI Studio, or Vertex, or use the platform model (Qwen2.5-7B, 32k context, no embeddings).
How am I billed for an agent?
As container compute ($6/vCPU-mo, $3/GB-mo) plus inference tokens when you use the platform model. There is no per-agent or per-invocation fee.
Where do my provider API keys live?
In your own project secret, scoped to your tenant. When you bring your own key, that provider bills you directly, not us.
Related products
Deploy this in your first project.
Claim your $5 trial credit and go live today.