Introduction
What ML.ai Inference is, the four modules it's made of, and where to start.
ML.ai Inference is the harness behind ML.ai, a product of Pixis: it takes a task, decides which agent and which routing tier should handle it, and routes the work to the option that clears the quality bar for that step at the lowest cost. Everything documented under ML.ai is a module built on top of that same harness.
ML.ai Inference does not surface which underlying model handles a request. Work is routed to one of two tiers, ML.ai Standard or ML.ai High, and the routing behind each tier is ML.ai's own decision, not something you configure.
The four modules
ML.ai Inference ships as four separate products, each aimed at a different point where you'd want an agent working on your behalf.
| Module | What it's for | Status |
|---|---|---|
| ML.ai Code | An AI coding agent inside VS Code and Cursor: reads your repository, edits files, runs commands, and asks before anything irreversible happens. | Shipped |
| ML.ai CLI | The same harness from a terminal, for scripting and CI workflows. | Parked |
| ML.ai Market | Agent workflows built for marketing and growth teams. | Parked |
| ML.ai Voice | Voice-driven agent interaction. | Parked |
ML.ai Code is the only module documented here today. The other three aren't public yet: this page will grow a section for each as they ship, rather than promising specifics ahead of a release.