MENLO PARK, Calif. — Corepass, an artificial intelligence company based here, is opening Remy, its AI agent for software development, to teams in waves. Developers can request early access on the Remy website; once in, they can start on a free plan.
Remy takes a task described in plain language, writes the code and then checks its own work. The company sums up the product in five words, “Build it. Ship with confidence,” and the second half of that line is where Remy is trying to set itself apart. After it changes code, the agent runs the relevant tests, type checks or build. If something fails, it fixes the cause, reruns the same check and tells the developer what ran, how it went and what it did not verify.
After it changes code, the agent runs the relevant checks — and says plainly what it did not verify.
A model for each task
Remy is not tied to one model. Developers can choose among open-source and proprietary models — the documentation lists models from OpenAI and Anthropic alongside the Kimi, GLM, DeepSeek, Grok and Qwen families — or select Dynamic, in which Remy picks a model for each task and moves up to a stronger one if the work gets stuck. Dynamic comes in three tiers: Fast, Default and Deep.
Users do not need a separate account with a model provider or their own API key, according to the company’s answers to common questions.
From a spec to a pull request
On its product page, Corepass walks through a typical job. Remy reads the spec in Confluence and the design in Figma, builds the change, has Bugbot review the diff for what the tests missed and opens a pull request, after which the team hears about it in Slack.
- DescribeRemy reads the spec and the design
- BuildIt writes the change and checks it
- ReviewBugbot reads the diff for what tests missed
- ShipThe PR opens and the team hears in Slack
Two features carry that work past the developer’s own machine. A Cloud Agent can take over a local task and its context and keep going with the laptop closed; developers can add requests and answer Remy’s questions from a phone, then review the result in any browser. And in Slack, mentioning @Remy in a thread hands off what the thread is asking for. Remy posts its progress in the thread and shares the pull request when the work is ready.
Where it runs
Remy is available as a desktop app for macOS, Windows and Linux, as a command-line tool installed through npm, on the web and on phones, all under one account.
Corepass describes Remy’s audience as individual developers and engineering teams, with enterprise engineering teams as a longer-term focus. Teams get a shared workspace with roles, usage reporting and spend limits, and the company’s Trust Center sets out, plan by plan, whether a customer’s work may be used to improve future models.