Mainline

Mainline is Git-native intent records for the AI era. It lets agents save developer intent and decisions alongside code in a Git repository, so teams can keep engineering judgment with the work rather than in separate chat logs or memory systems. The project says it is for teams already using agents in real engineering workflows and for solo builders who want later agents to understand why earlier plans were abandoned.

Its records can include intent, tradeoffs, no-go zones, goals, pivots, validation, constraints, and key decisions. The site describes several uses for those records: reading repo history, decisions, and risks before starting work; saving a task’s goal and pivots on the current branch; binding merged commits to the decisions that produced them; and surfacing logic conflicts before Git conflicts. It also says future agents can see old traps and high-risk boundaries before editing, and that reviewers can inspect the original goal, reasoning path, and key decisions behind a change.

Mainline works through agent hooks, skills, and a CLI. It writes intents into Git refs and notes, and the site says those records move through fetch, branch, and merge. It also mentions a live intent Hub and a public alpha, with a snapshot showing historical decisions, risks and constraints, and work in progress. The product page includes an install command and states that users keep their existing AI coding workflow, asking for work, inspecting the diff, and reviewing as usual.

The service is described as Git-based and Git-native, with collaboration handled through Git’s existing model. It says data stays in the Git repo, that it does not save code or AI chats, and that it has no new memory database or platform lock-in.