Ask for CPU, RAM and a region. Get a cloud VM.
Routing finds the machine that fits your ask at the best price across providers. The meter runs per second at raw provider cost, and the machine dies on your TTL, your budget, or a DELETE.
give this to your agent
set up https://reachpad.dev/router/skill.md
The skill covers keys, paying, provisioning, exec and teardown. An agent with a funded key takes it from there.
Build every call
Everything lives under https://router.reachpad.dev/v1 and answers JSON. Open an endpoint, pick its values, copy the call. Payment is USDC over x402, with one 5% fee at top-up; machines spend at raw provider cost, metered per second, until ttl_seconds, max_spend, or a DELETE ends them.
machines
accessssh is a key login and a hostname; exec runs one command over HTTP, with neither.
ssh_public_keyssh-ed25519 AAAA... you@hostA placeholder. Paste your own key before running.
ttl_secondsDestroys the machine at expiry. Omit it and the API uses 4 hours. Some offers cap TTL, so a shorter one routes to more of them.
max_spendDestroys the machine when accrued spend reaches it. Offers that cannot run 5 minutes inside it never route.
max_rate and client_ref
max_rateDrops offers whose hourly rate is above it.
client_refA retried create carrying the same ref returns the machine it already made instead of buying a second one.
POST /v1/machines
# $KEY is your API key: POST /v1/accounts mints one$ curl -sS -X POST https://router.reachpad.dev/v1/machines \ -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \ -d '{"cpu":2,"ram_gb":4,"disk_gb":8, "region":"us-west","access":["ssh"], "ssh_public_key":"ssh-ed25519 AAAA... you@host", "ttl_seconds":3600,"max_spend":"1.50"}'