What is Faustine | Faustine Docs

Faustine is a platform that adds AI agents to a software product. It has two products. Web Agents puts an agent inside your web app, and Agent Gateway publishes your product's API to the AI clients people already use, such as Claude, ChatGPT and Cursor.

Both are built from your code and managed in the Faustine portal.

How it works

You give Faustine your code: a repository on GitHub, GitLab or Bitbucket, or a ZIP. Faustine runs an analysis over it and writes down what the product can do. For a web app the result is a set of actions. For an API it is a set of operations.

The analysis writes into a draft. You review the draft, change what you need and publish it as a numbered version. An environment, such as Production or Staging, serves one version at a time, so a change reaches your users when you decide.

The analysis, the chat and the Advanced editor run on a model provider that you connect to your workspace, and that provider bills you for the usage. Nothing that uses AI runs until a provider is connected. See Connect your model provider.

Web Agents

A Web Agent is an agent that lives inside your web app. The user talks to it in a widget on the page, and the agent does the work in the app for them, such as creating a deal or answering a question about an invoice.

Faustine builds the agent from the app's code. Each action is generated code that runs in the page, in the user's own session, so the agent can do what that user can do by hand and nothing else. Your app does not need a separate API for the agent.

You install the agent with a script tag, the @faustine-ai/web-agent npm package, or @faustine-ai/next in a Next.js app. The widget loads the published version from Faustine at runtime, so a version you deploy in the portal reaches your users without a new build of your app.

To add one to your app, see Quickstart: add an agent to your web app. The Web Agents section covers the rest.

Agent Gateway

Agent Gateway publishes your API to AI clients over MCP and REST.

It works in two parts:

An API is what Faustine knows about a service of yours: its operations with their input and output schemas, which ones write, which ones require confirmation, and how the service authenticates. Faustine analyzes it from code, a ZIP or a spec. REST, GraphQL and gRPC are supported. See APIs.
A gateway serves one or more APIs at one URL. You choose which operations it exposes, create an API key or a gateway user for each caller, and set limits. Every request is recorded in the gateway's logs. See Agent Gateway.

A client connects to the gateway's URL with its credential. The gateway checks the credential, applies the limits, asks for confirmation where an operation requires it, and calls your API through a connection that holds the base URL and the credential values.

To publish your first API, see Quickstart: let AI clients operate your API.

Apps

Two apps use your gateways.

App	What it is
Faustine Chat	A chat app for the people in your company. They sign in with a username and password and work with the gateways you give them access to. It runs on your model provider.
Faustine CLI	The `faustine` command-line client. It calls the operations of any gateway from a terminal, a script or a coding agent.

Where things are in the portal

Section	What you do there
Web Agents	Create, configure, publish and monitor your Web Agents.
APIs	Analyze an API and review its operations, recipes, authentication and connections.
Agent Gateway	Create gateways, choose operations, manage API keys and gateway users, read logs.
Faustine Chat, Faustine CLI	Set up the two apps.
Monitoring, Usage	Activity, logs and errors across the workspace, and the tokens each feature used.
Settings	Team, notifications, billing and the model provider.

Next steps

Connect your model provider: the first thing a new workspace needs.
Quickstart: add an agent to your web app: from an empty workspace to an agent on a page.
Quickstart: let AI clients operate your API: analyze an API and connect a client to it.
Concepts: the terms the portal and these docs use.
Take a Web Agent to production: what to set up before your users get the agent.