AmpleRun documentation

AmpleRun connects people and AI agents who need compute with independent hardware hosts. Choose a workload, compare eligible machines, fund a prepaid balance, and rent within an explicit budget. The meter begins at verified readiness, not when you request a machine.

Start with your task

You want to…	Start here
Run a model or use a GPU from the website	Rent your first machine
Integrate compute into an application	Build with the API
Let an AI agent register and rent for its work	Set up an agent account
Offer your hardware for rentals	Become a host

Understand the workflow

1. Discover: use the public catalog to find a template and a compatible offer.
2. Fund: read your account’s current deposit instructions and wait for spendable credit.
3. Quote: get a time-limited price snapshot. A quote does not reserve a machine or charge your balance.
4. Rent: reserve using the quote and a spending cap. Wait for RUNNING before using access details.
5. Finish: stop the workload, retrieve your results, and read the itemized receipt.

Read catalogs and quotes, deposits, and budgets and receipts before automating paid work.

Know which address to use

Address	Purpose
https://amplerun.com	Human-facing website, account controls, and hosting setup
https://amplerun-api.spiry.org	API origin for machine clients; append /api/v1/...
https://docs.amplerun.com	This documentation; not an API or inference endpoint

The API reference describes request and response fields. Download the OpenAPI contract for tooling. Hosted MCP and the Agent Skill provide agent-oriented entry points.

Availability comes from the service

Catalog inventory, accepted deposit networks, feature availability, and quoted prices change. Read the current API response rather than treating an example as an available machine, accepted payment route, or fixed price.