Getting started
Orfyn is an AI forward deployed engineer: a platform that runs coding agents inside your own environment to do the work that decides whether an AI project ever ships — discovery, system archaeology, field mapping, verification against the real rules, and the change requests that never stop after go-live.
It is an enterprise product, not a public download. Every deployment is licensed and provisioned by us, and access to the builds, the workspace, and the team features is granted per engagement. If you have reached these docs without a deployment, start by booking a call — bring one workflow you are trying to ship, and we will walk through what it takes to get it live.
How access works
Orfyn is licensed under terms that permit use only within a granted deployment. There is no self-serve signup and no public installer: the owner-admin provisions your organisation, your builds, and your accounts. That is deliberate. The product runs with real credentials against real systems, and the access model is the first of its guardrails, not an inconvenience in front of them.
Once your deployment is provisioned you receive:
- The desktop app — bundles the daemon that runs and supervises agents on your machine or server. Open it, sign in with the account provisioned for you, and it is running.
- Mobile and web clients — drive the same daemon from anywhere. Pair a phone from Settings → your host → Connections → Pair a device.
- A Teams workspace — the coordination layer: assignments, issues, transcripts, and review. Your files, code, and agent instructions stay on your machines; Teams carries only the coordination.
What a deployment looks like
- Scope. One workflow, chosen together. Not a platform rollout — a specific piece of work with a definition of done.
- Deploy. The daemon runs on your infrastructure — a laptop, a company server, or a container inside your network. Model calls leave on your own credentials. Nothing routes through us.
- Discover and build. Agents read the systems involved, map the fields that matter, and build the connection where it cannot hurt anything, with every change reviewable before it lands.
- Verify. Against your real rules — the approval branches, permission boundaries, and edge cases a demo never exercises.
- Operate. After go-live, the same loop absorbs the change requests. The second change costs less than the first, because the context stays in the system rather than in someone's head.
The full argument for this shape is on the site: how it works, use cases, and security.
Where next
These docs describe how the product works once you have it:
- Why Orfyn — the problem it exists to solve.
- Workspaces — the project, workspace, and session model everything is built around.
- Providers — Orfyn runs the agent CLIs you already use, on your own accounts.
- Orchestration — one agent delegating work to others.
- Security — the boundary model, end to end.
- CLI reference — every command.
Prerequisites
Orfyn manages coding agents; it does not ship a model. A deployment needs at least one provider CLI installed and authenticated with your own credentials — see Supported providers. For PR-aware worktrees and several orchestration features, the GitHub CLI (gh) should be installed and authenticated as well.