Introduction
A backend that lets AI agents do real work, with limits you control.
Authoric is a backend platform where every request — from a person, a script, or an AI agent — is authorized the same way, by policy, at the moment it happens.
You get the pieces you would expect from a backend: a database with an instant REST API, authentication, file storage, and custom server-side functions. What makes Authoric different is what happens when an AI agent uses them.
The problem it solves
Giving an agent access to your production data is usually all-or-nothing. You hand it an API key with broad permissions and hope its instructions hold. When it does something you did not intend, you find out afterwards.
Authoric inverts that. An agent has no tool list of its own — its abilities are derived from the policies bound to its role, the same policies that govern your human users. When it reaches a step you have marked high-risk, the run stops and waits for a person to approve it. That pause is the product.
The Halt
A halted run is not an error and not a timeout. It is a run that reached a boundary you set, recorded what it wanted to do, and is waiting. Approve it and it resumes from exactly where it stopped.
Where to start
Quickstart
Sign up, get a token, and make your first authenticated API call.
Core concepts
Organizations, apps, and the two planes — the model everything else assumes.
The governance model
Roles, policies, and halts. The thread that runs through the whole platform.
Agents & AI
Give an agent capabilities, connect your own LLM, and review what it does.
How the documentation is organized
Each capability is documented in the same shape, so you always know where to look:
- Overview — what it is and when you would reach for it.
- Set it up — the console steps, with screenshots.
- Use it — the API calls your application makes at runtime.
The API Reference is generated directly from the running server's contract, so it cannot drift from what the API actually accepts.