Run source code in 60+ languages via sandboxed Judge0, paid per call on Tempo L2.
What it does
This MPP endpoint wraps Judge0, an open-source online code execution system, behind Locus's Micropayment Protocol. It lets AI agents and applications execute arbitrary source code in over 60 programming languages (Python, JavaScript, C++, Java, and many more) inside a sandboxed environment. Each call is settled as a one-shot charge via the Tempo L2 network using pathUSD.
The service exposes five operations: execute-code (synchronous execution), submit-code (asynchronous submission that returns a token), get-submission (poll for async results), list-languages, and list-statuses. The execute-code and submit-code endpoints accept source code, a language ID, optional stdin, expected output for validation, compiler flags, command-line arguments, and resource limits (CPU time, wall time, memory, stack size). Network access inside the sandbox can be toggled per request. Execution costs $0.006 per call for execute-code and submit-code, and $0.005 per call for get-submission, list-languages, and list-statuses (assuming pathUSD with 6 decimals).
The endpoint is a POST-only service; HEAD/GET probes return 404, which is expected behavior for this routing style. The OpenAPI 3.1.0 spec is well-defined with full request schemas and payment metadata. Documentation references point to judge0.com and the Locus skill file at paywithlocus.com/skill.md.
Capabilities
Use cases
- —AI agents that need to run generated code and inspect output
- —Automated grading systems that evaluate student submissions
- —LLM tool-use pipelines requiring code verification or computation
- —Chatbots that execute user-provided code snippets safely
- —CI/CD or testing workflows needing ephemeral code execution
Fit
Best for
- —Agents needing on-demand code execution without managing infrastructure
- —Pay-per-call code execution with no subscription or API key
- —Running untrusted code in a sandboxed environment across many languages
Not for
- —Long-running or persistent server processes
- —Workloads requiring large file I/O or persistent storage
- —High-throughput batch execution where per-call pricing is prohibitive
Quick start
curl -X POST https://judge0.mpp.paywithlocus.com/judge0/execute-code \
-H "Content-Type: application/json" \
-H "Authorization: <MPP-Tempo-token>" \
-d '{"source_code": "print(\"Hello, world!\")", "language_id": "71"}'Example
Request
{
"stdin": "",
"language_id": "71",
"source_code": "print('Hello, world!')",
"memory_limit": 128000,
"cpu_time_limit": 5,
"redirect_stderr_to_stdout": true
}Endpoint
Quality
The OpenAPI spec is detailed with full request schemas, payment metadata, and language examples. However, no response schema is documented, the probe returned 404 on HEAD/GET (expected for POST-only but no 402 challenge was captured), no crawl pages yielded useful docs, and example responses are unavailable.
Warnings
- —No 402 MPP challenge was captured because the endpoint only accepts POST; HEAD/GET return 404. The endpoint is likely live but liveness could not be confirmed via the probe.
- —Response schema is not documented in the OpenAPI spec — output format must be inferred from Judge0's upstream API.
- —Currency address 0x20c000000000000000000000b9537d11c60e8b50 is assumed to be pathUSD with 6 decimals; if decimals differ, stated prices are wrong.
- —All crawled pages returned 404 JSON errors; no supplementary documentation was retrievable from the provider origin.
Citations
- —OpenAPI spec declares 60+ programming languages with sandboxed isolationhttps://judge0.mpp.paywithlocus.com
- —execute-code and submit-code cost 6000 base units; get-submission, list-languages, list-statuses cost 5000 base units via Tempo methodhttps://judge0.mpp.paywithlocus.com
- —Homepage documentation at judge0.com, API reference at ce.judge0.com, LLM docs at beta.paywithlocus.com/mpp/judge0.mdhttps://judge0.mpp.paywithlocus.com