ML.ai
ML.ai is a coding agent for delegating software work from a repository rather than asking for isolated code suggestions. The vendor says it reads the codebase, routes each task step to an appropriate model, runs the project’s tests, and presents a diff for review before making changes. It is available for macOS, Linux, and Windows and is currently described as a private beta launching in India first.
What it does
ML.ai’s Infinite agent is designed for end-to-end engineering tasks such as bug fixes, test writing, authentication migrations, dependency audits, and CI triage. According to the vendor, it can inspect a repository, reproduce an issue, edit multiple files, run existing commands and tests, and open a pull request with a plain-language summary. It does not merge changes automatically.
The product supports common development environments and tools listed by the vendor, including Git, GitHub, Docker, pnpm, pytest, Rust/Cargo, Terraform, Kubernetes, AWS, Prisma, Playwright, and GNU Make. These are presented as supported workflows rather than a guarantee that every repository or configuration will work without setup.
Who it helps
ML.ai is aimed at solo developers, professional engineers, and teams with work that benefits from background execution or parallelization. Its examples include long-running migrations, recurring dependency checks, issue triage, and monitoring flaky tests. The agent can be used from the desktop or web app, an IDE extension for VS Code and JetBrains, or a terminal CLI.
Teams can define project conventions in an MLAI.md file and create reusable Skills. The vendor says these instructions are applied to later tasks, allowing teams to encode standards for migrations, testing, or other repeatable work. Team and Enterprise offerings also advertise shared Skills and routines, usage and approval controls, SSO, audit logs, and optional self-hosted or VPC deployment; availability and final terms should be confirmed with ML.ai.
Workflow
A typical task starts with a request such as fixing a GitHub issue. ML.ai reads relevant files and searches the repository, then divides the work into steps such as planning, reproducing the failure, editing code, testing, and opening a pull request. Its automatic routing chooses models per step—for example, a faster model for routine work and a higher-effort model for more complex debugging—while users set the outcome and constraints rather than selecting a model at every stage.
The vendor’s example shows a double-charge bug being reproduced against a test charge, changes made to client and server files, pnpm test passing, and a pull request opened. Users review the generated diff before approval; the product states that nothing writes to disk until the user selects Allow. Multiple agents can run in separate sandboxes, making it possible to work on several repositories or tasks concurrently.
Strengths and limits
ML.ai’s notable strength is its review-oriented, verification-focused loop: inspect, change, test, summarize, and propose a pull request. Scheduled routines and model routing may be useful for teams that want engineering tasks to continue without manually managing each model invocation. The vendor also markets Infinite around a flat subscription with no visible token counter, subject to fair-use and concurrency rules.
The product is still in private beta, and the vendor says pricing shown on the site is India launch pricing and may change before general availability. Individual plans list ₹699, ₹1,499, and ₹3,999 per month with different concurrency levels; Team pricing is listed at $29 per seat monthly when billed annually, with a five-seat minimum. Actual performance will depend on the repository, tests, permissions, and task complexity. Vendor-published benchmark figures are not a guarantee of results on a particular codebase.