Gittifacts proposes a Git workflow where agents prove their work

Developer Jovi De Croock has launched gittifacts, a Git platform built around a neat division of labour: people approve what should happen, while machines provide evidence that a change meets the bar. Rather than asking reviewers to inspect raw diffs first, it routes changes into review lanes using checks such as tests, coverage and security results.

Watch Desk analysis

What happened

De Croock describes gittifacts as an agent-native platform for managing code changes. Its design includes isolated forks for agents, speculative merge queues and automated rollbacks linked to feature flags. These are features of the launched platform as described by its developer, not independently tested performance claims.

Read De Croock’s description of gittifacts.

Why it matters

AI coding agents can generate changes quickly; reviewing and integrating those changes is a different problem. Gittifacts’ approach tries to make evidence, rather than a human’s line-by-line inspection of every diff, the first stop in the process. That could change where engineering teams spend their attention, while leaving people responsible for approving intent.

The design also puts a practical question on the table: what happens when automated checks pass but miss a problem? A queue that can merge changes and roll them back needs trustworthy checks and clear ways for people to intervene. Automation may lighten the review load, but the evidence has to earn that trust.

Our read

This is a concrete attempt to redesign Git workflows for coding agents, not just bolt a chatbot onto an existing review page. The useful idea is separating approval of intent from checking whether a change conforms. Whether that is better than reviewing diffs depends on the quality of the checks and the control engineers keep over the merge.

What to watch

• Which review lanes and checks are available in the platform.
• How teams can inspect or override an agent’s proposed change.
• Whether automated rollbacks and speculative merges work reliably in real projects.

Discussion spark: If coding agents produce more changes than people can review line by line, should teams trust automated evidence to gate merges, or keep human review of every diff?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.