Computers & runtimes

Choose where work runs—and make that boundary legible.

M1 Project is designed to coordinate execution on computers your team connects: a laptop, development host, server, container, or controlled cloud runner with supported agent tools installed.

Pre-release noteRuntime connections and hosted execution are proposed capabilities under development. The marketing site does not connect to repositories or computers.

Request early access Discuss your use case

Separate four jobs cleanly

The task is the shared record. The agent defines how to work. The computer supplies files, credentials, network, and tools. The runtime performs the concrete AI session.

↗Planned runtime support for Codex, Claude Code, Gemini CLI, and compatible tools
↗Availability and concurrency visible before dispatch
↗Model usage remains on the connected tool or API account

Use temporary worktrees for parallel code changes

Repository-backed tasks can prepare separate Git worktrees so concurrent runs do not mutate one checkout. Isolation prevents workspace collisions; it does not remove real semantic merge conflicts, which reviewers must still resolve.

Treat credentials as scoped execution inputs

Repository access, tokens, and environment variables should be granted to the narrow agent role that needs them. The planned product will make connection and lifecycle controls part of pilot review before they are enabled.

Pre-release · No instant signup or payment

See whether M1 Project fits your team’s operating model.

Request a walkthrough