Pay-per-use code execution for AI agents. HTTP 402, no account, no API key.
agent-exec runs untrusted source code for software agents and returns stdout, stderr and the exit code from an isolated, network-less sandbox. There is no signup, no account and no API key: it answers HTTP 402 Payment Required with an x402 challenge, and possession of a valid payment is the entire authorisation model, so an agent can discover the price, pay it and get a result with no human in the loop. Every endpoint is priced per call in USDC on Base, from $0.0010 for POST /execute. It also offers the small paid primitives agents need to coordinate: object storage, name claims, attestations, handoffs and dead-man switches. The whole surface is machine-readable up front - x402 manifest, OpenAPI 3 and llms.txt.
-
Accountless: no signup, no API key, no dashboard - a valid x402 payment is the only credential
-
HTTP 402 native, priced per call in USDC on Base from $0.0010
-
Isolated, network-less sandbox returning stdout, stderr and exit code
-
Machine-readable discovery: x402 manifest, OpenAPI 3 and llms.txt
-
Paid coordination primitives: storage, claims, attestation, handoff, dead-man switch
-
An autonomous agent needs to run code it just wrote and read the real output
-
Executing untrusted or model-generated code away from your own host
-
Paying for compute machine-to-machine without provisioning an account
-
Giving a short-lived agent scratch storage and a name claim it can pay for per use
-
Integrating a priced tool purely from its published discovery documents