What is Faros?
Understand what Faros brings together, when to use it, and how to get started.
Faros is an open-source control plane for applications, AI agents, and infrastructure. It brings capabilities such as application development, GitHub repositories, compute environments, and connected clusters into shared workspaces. People and automation can work with those resources through the console, CLI, APIs, and MCP tools.
These docs are for people building applications and agents, teams managing infrastructure, and operators providing a platform for others. You can begin with a single capability; you do not need to understand or enable every provider.
Get started with Faros
Choose a task that matches what you want to accomplish:
- Build and publish an application : create a project, work with an assistant, test a preview, and prepare a deployment.
- Create an AI agent : configure a model and instructions, then verify the first response.
- Connect a cluster or server : register an edge and verify access through the hub.
Start these tutorials in the Faros SaaS hub. Your workspace administrator can confirm the providers, connections, and permissions available to you. You can also use a self-hosted hub if your organization operates Faros on its own infrastructure.
When to use Faros
Faros is useful when a task crosses the boundaries between application code, AI tools, and the systems they use. For example, you might want to build an application against an existing data source, give an agent access to selected tools, or inspect resources across connected Kubernetes clusters.
| What you need | How Faros helps | Where to begin |
|---|---|---|
| Develop an application with an AI assistant | App Studio connects project conversations to repositories, development environments, previews, and publishing workflows. | App Studio |
| Run repeatable work with an agent | Configure instructions and tools, inspect runs, and add schedules or triggers when the task works. | AI agents |
| Reach infrastructure from a shared entry point | Edge agents connect outward to the hub; use those connections to access clusters, servers, and services. | Edges |
| Offer reusable environments to a team | Templates define the resources people can request; instances track the resulting environments. | Infrastructure |
Faros coordinates these systems; repositories still live in GitHub, source tables remain in Databricks, and workloads run in the configured infrastructure. Each provider overview explains its dependencies and data storage.
How Faros works
The hub is the shared control plane. It handles identity and workspace routing and connects requests to the providers installed by an operator.
An organization groups people and workspaces. A workspace is the context in which you create resources and use enabled capabilities. Selecting the intended workspace matters whether you are using the console, a CLI command, or an agent.
A provider adds a capability: its resource APIs, controllers, and, where implemented, console views and MCP tools. Providers can be enabled independently. App Studio, for example, uses Code and Infrastructure to connect application work to repositories and runtime environments. Edges connects existing clusters and servers through an agent that initiates an outbound connection to the hub.
Read organizations, workspaces, and providers for an introduction to these boundaries.
The provider service and the runtime may operate in different clusters. A resource in the control plane describes work; it is not necessarily where that work executes. MCP exposes enabled tools to compatible assistants, subject to the endpoint credential and provider authorization.
Resources, actions, and streams
Declarative resources express desired state and controllers reconcile it. Direct actions perform bounded requests; streams carry logs, terminals, or other live traffic. Not every action creates an auditable, reversible resource object.
Follow a provisioning request to see how an assistant request becomes a running workload.
Choose how you work
Use the console for interactive application and agent workflows. Use the CLI for terminal workflows, workspace selection, and edge connectivity. Connect an external AI assistant through MCP when you want it to discover and use tools exposed by your enabled providers. Each interface operates with its configured identity and authorization; connecting an assistant does not grant it unrestricted access.
Operators install and run the hub and providers. Provider authors extend the platform with new APIs, interfaces, and tools. These responsibilities have separate guides under Self-hosting and Extend Faros .
Next steps
Complete your first task , browse the available capabilities , or follow an application workflow using Databricks data to see multiple providers working together.