Coming soon

Modal for AI agents

Serverless GPUs, functions and sandboxes.

What your agent can do with Modal

Modal: serverless GPUs, functions and sandboxes. Through AMXP your agent calls it like any other tool: no Modal account, no separate API key and no extra subscription. The call is priced up front and paid from your AMXP balance.

Modal is on the AMXP roadmap and not callable yet. Everything on this page describes how it will work once its connector ships.

How to call it

The same three calls work for every provider. Your agent does this on its own once AMXP is installed.

agent.ts

// Find the right Modal endpoint, check its price, then run it.
const [tool] = await amxp.discover({ provider: "modal" })
const spec   = await amxp.inspect(tool.id)      // parameters, price per call
const result = await amxp.execute(tool.id, { /* params from spec */ })

What Modal charges

Free tier, then pay as you go

$0.0000131 per CPU core/sec; GPUs from $0.000164/sec (T4)

Model: Subscription + usage
Currency: USD
Free tier: Yes

$30/month included compute, 3 seats, 100 containers, 10 GPU concurrency

Website ↗Pricing ↗

Prices as of 2026-11-02, source: https://modal.com/pricing.

Pricing through AMXP

0% markup, pay per call. Your agent sees Modal's own price from amxp.inspect() before it runs anything, the amount is reserved from your balance, and only the real cost is settled. No monthly fee, no Modal account. How pricing works

Alternatives in Virtual Machines

Your agent can compare these by fit and price with amxp.discover() and pick per task.