RepositoriesBuilt into Genosyn

Version control your team can actually use.

Code or documents, edited in the browser, reviewed before it ships.

Not every repository is a codebase. A Repository is a real git repository your company owns — clone one from GitHub, GitLab, Bitbucket, or a self-hosted server, or create one empty inside Genosyn with no git host at all. Starting empty is not a dead end: connect that repository to GitHub later and Genosyn creates it there through your existing GitHub Connection and pushes the history across, with no personal access token minted or pasted. Members browse, search, and edit files in the browser, read the diff, and commit under their own name. AI Employees get their own isolated worktree, edit through Genosyn-executed tools rather than a shell, and commit to their own branch — which a human reviews and merges. It runs on the standard Docker install, with no coding tools and no sandbox required.

Repositories · Live product storyRunning
Genosyn running a Software Engineer use case in Repositories.
  • Start empty inside Genosyn, or clone any git URL
  • Publish a local repository to GitHub later
  • Browser editor with search, diffs, and history
  • AI work sessions you review before merging
  • Credentials AES-256-GCM encrypted
Repositories in practice

Built around outcomes, not demos.

Start with a real role and a real handoff. Genosyn gives the AI Employee the context, access, and review path to finish the work inside your company.

EN
Engineering

Software Engineer

Investigate failures, prepare a tested patch, and hand the diff to a human reviewer.

A review-ready fix, with the evidence attached
CS
Support

Customer Support Specialist

Triage the inbox, gather customer context, and draft answers grounded in your docs.

A clean support queue and faster, consistent replies
SE
Security

Security Engineer

Review risky changes and recurring checks with a complete, auditable trail.

Evidence-rich reviews that keep humans in control
What ships

Repositories, end to end.

Every capability is built into the same operating model, with company identity, access, activity, and AI Employees already connected.

Code or documents

Mark a Repository as code or documents and the copy, editor defaults, and AI briefing follow. Underneath it is a plain git repository either way, so a quarter's strategy gets the same branches, diffs, and history as a service.

Start local, publish to GitHub later

Clone any HTTPS or SSH URL, or create one empty with git init and no remote at all. When it outgrows that, pick a connected GitHub Connection: Genosyn creates the repository on GitHub — private by default — and pushes the history. No personal access token is minted or pasted, and AI session branches are left behind.

A real editor in the browser

Tree, editor with syntax highlighting, create, rename, move, delete. The tree respects .gitignore, search is literal and includes uncommitted files, and the README renders on the Overview. Diffs per file or whole-tree, and commits attributed to the Member who made them.

Branches and history

Create a branch from any revision, switch between them, and read the log for the repository or one file, each commit with its own diff. Open a file as it was at an older commit. Fast-forward pull and push for owners and admins.

AI work sessions

Describe the work, pick a granted AI Employee, and it runs in its own git worktree on its own branch — editing through six Genosyn-executed tools, never a shell. It commits and reports. You read the diff and merge, or discard it.

Nothing ships unreviewed

The checkout that holds credentials is unreachable by any model process. Tokens and SSH keys are AES-256-GCM encrypted, never returned to the client, and used only in the push path — which only an owner or admin can trigger.

With AI Employees

An AI Employee that works like a colleague

Grant an employee a Repository and you can hand it a piece of work the way you would hand it to a person: describe the outcome, let it work in isolation, then read what it did before any of it lands.

01

Isolated by construction

Each session gets its own git worktree and branch. Two sessions never collide, and an employee cannot reach the shared checkout, write into .git, or point a symlink out of the tree.

02

No shell required

List, read, write, delete, search, commit — six tools Genosyn runs on the employee's behalf. Repository work no longer waits on coding tools, bubblewrap, or host execution being switched on.

03

A diff and a report

The session ends with commits on a branch and a short written report of what changed, what was left alone, and what could not be verified. Publish merges it; discard deletes the branch.

Questions

Frequently asked.

Is this only for source code?

No. A Repository is any version-controlled workspace — a service's source, a quarter's strategy, a set of operating policies. The kind field (code or documents) changes the copy, the editor defaults, and how an AI Employee is briefed; underneath, every Repository is a plain git repository.

Do I need a GitHub account, or any git host?

No. A local Repository is created empty inside Genosyn with git init and never leaves your server. A remote Repository clones any HTTPS or SSH URL — GitHub, GitLab, Bitbucket, or a self-hosted server. And a local one can be connected to GitHub whenever you want, which creates it there and pushes the history it already has.

What does connecting a repository to GitHub actually do?

It creates the repository on GitHub through the Connection your company already authorised under Settings → Integrations, then pushes the history the repository already has. Nobody creates or pastes a personal access token: the Connection's token is resolved per operation and never stored on the repository, and later pushes to that remote use the same Connection. Branches from AI work sessions are deliberately left out of that first push. If you would rather create the repository yourself, paste the clone URL of an empty one instead — a remote that already has commits is refused, not force-pushed. Owner or admin only.

Do I need coding tools or a sandbox enabled?

Not for this. The browser editor and AI work sessions run against a server-owned checkout with no shell involved, so they work on the standard Docker install. The separate per-employee checkout — the one an employee uses with ordinary git during open-ended chat and Routine work — still needs coding execution enabled.

Can an AI Employee push straight to my remote?

No. It commits to its own branch in its own worktree. A Member reviews the diff and merges, and only an owner or admin can push. Credentials are used exclusively in that push path and never enter a tree a model can reach.

Who is allowed to do what?

Browsing, searching, editing, committing, branching, and starting AI work sessions are open to any company Member. Pushing, pulling, connecting a repository to a remote, and repository configuration — clone URL, credentials, and AI grants — require an owner or admin, because a local commit can be undone and a push cannot be recalled.

How are my tokens and SSH keys stored?

Encrypted at rest with AES-256-GCM, the same protection as model API keys, and never shown back in plaintext — the UI reports only whether a credential is set. An SSH key is written to an App-private temporary directory for exactly one operation and removed afterwards.

What happens if I delete a repository in Genosyn?

It removes the grants, the work sessions, and the server-side checkout. A remote repository is never touched on its host. A local repository has no remote, so deleting it deletes the only copy of its history.

Put your first AI Employee to work.

Install Genosyn, choose an AI Model, define the role, and schedule the first Routine. The rest of the company can grow from there.