GitHub Gateway docs

Current Scope and Limitations

GitHub Gateway should be understood through its current scope, not through imagined future claims. The public product today is Self-hosted GitHub Gateway v1.3, a local release that demonstrates a repository write boundary on GitHub. That is already meaningful, but it is intentionally narrower than a complete hosted platform story.

What Is Available Today

The current public path is Self-hosted GitHub Gateway v1.3. It runs locally, uses a configured GitHub App as the write identity, creates Runner Keys for agent submission, and demonstrates guarded pull request creation, pull request reuse, and validated follow-up on Gateway-created pull requests.

That means the product can already be evaluated in terms of real repository impact. It is not just a mockup or a concept note.

What It Does Not Claim

GitHub Gateway does not claim to prove semantic correctness. It does not claim to review code quality, replace CI, replace branch protection, or auto-merge agent work. It does not claim to prevent every malicious action an agent might attempt in a broader environment. It does not claim to be a complete hosted governance platform.

Those limits are not weaknesses in the description. They are part of an honest public contract. The product claim is about controlling agentic write impact on GitHub repositories, not about proving that every generated change is good.

Why the Scope Is Deliberately Narrow

The narrower scope is what makes the product legible. A team can inspect whether blocked work creates no pull request. It can inspect whether allowed work becomes a reviewable pull request. It can inspect whether follow-up stays on the same Gateway-created review object.

Those are tractable, testable questions. Claims such as "the system makes agent coding safe" are not. GitHub Gateway is more credible when it stays anchored to the write boundary it actually implements.

What Kinds of Changes Fit Best

Self-hosted GitHub Gateway v1.3 is best for small, reviewable repository changes. It is well suited to configuration updates, bounded code edits, and demonstrations where the resulting pull request is easy for a human reviewer to inspect.

It is not described as the best tool for huge refactors, bulk generated assets, opaque binary changes, or unconstrained long-running repository automation. The boundary model is clearest when the requested change is small enough that the decision and resulting review object remain legible.

Why Hosted Availability Is Not the Current Story

A hosted GitHub Gateway may become part of the future product direction, but it is not the public path today. The serious path today is Self-hosted GitHub Gateway v1.3. That is where teams can actually inspect the boundary, keep the write identity under their control, and evaluate whether the model fits their needs.

This matters because vague hosted language can overstate what is currently available. The honest description is simpler: Self-hosted GitHub Gateway v1.3 is the current product surface, and any future hosted surface would still need to preserve the same core separation between agent proposals and external impact.

The Practical Reading of These Limits

The right way to read the current scope is not "too small to matter." It is "narrow enough to evaluate honestly." The product is focused on repository impact, reviewable pull requests, and the credential split between the agent and the Gateway write identity.

That focus is also what makes the next step clear. If Self-hosted GitHub Gateway v1.3 solves the repository-boundary problem for a team, then broader product surfaces can be considered later. If it does not, the evaluation still revealed something concrete rather than hiding behind inflated claims.

Where To Go Next

For technical detail on the current product surface, continue with: