Code
Model Pond has a supported version of Codex preconfigured for our models and with all telemetry disabled.
What it is
A coding agent you can install and use without wiring it up yourself. The build is preconfigured for Model Pond's models, so model access comes through the gateway rather than from a separate account, and no API keys are needed beyond your Model Pond one.
It is a supported build, which means it is tested against the models served here and updated when they change, rather than left to drift against whatever the upstream project releases.
Telemetry
All telemetry is disabled. The agent does not report usage, crashes, prompts or code back to its upstream vendor, and there is no setting to turn reporting on.
This is separate from what the gateway records for billing, which is described on the gateway page: counters and token totals, attributed to an API key and a target model, with no record of content.
Where the work runs
Model requests go to the gateway, so the same routing, hosting and retention terms apply as to any other request you send us: handled on Model Pond-managed infrastructure, then forwarded to the provider serving the model you selected.
Code stays on your machine. What leaves it is the context the agent sends to the model, and nothing else.
What to read next
The parts of the platform people usually read after this page.