Skip to content

Firestartr

Firestartr is an open source framework to build internal developer platforms tailored to your needs. It gives an organization one explicit map of its software — and of the people responsible for it — and then acts on that map to create and maintain the repositories, pipelines, and infrastructure the software needs.

Firestartr sits between a map of the organization — domains, systems, components, teams, features, and infrastructure — and the GitHub, AWS, and Azure resources it creates and maintains

When software outgrows the organization

It usually starts small: a few repositories, and everyone knows who owns what. As teams and products multiply, that shared understanding disappears.

  • Every repository grows its own CI/CD setup, with little convention across them.
  • The architecture becomes opaque: nothing describes how the pieces relate.
  • Permissions accumulate person by person, so responsibility turns diffuse.

Four people hold tangled admin, write, and read permissions over nine loose repositories, under clouds labeled opaque architecture, CI/CD per repo, duplicated configuration, ad-hoc solutions, and individual permissions

The software still runs, but the organization no longer has a model of it. Firestartr makes that model explicit before automating anything on top of it.

A map the business can recognize

Firestartr organizes software in layers that match how the business actually works.

Domains

A domain groups the systems that serve one part of the business. In an online store, a commerce domain covers the buying experience end to end.

Systems

A system is an explicit grouping inside a domain, with its own responsibility. The commerce domain splits into an inventory-system (catalog, search, and stock) and a checkout-system (cart, orders, and payments). The boundary is deliberate: each system owns a coherent slice of the product.

Components

A component is a unit of software together with its repository. The system it belongs to gives it a place on the map, so nine repositories become nine known components in two systems instead of nine loose ends.

Teams and ownership

Responsibility is part of the map. People are grouped into teams, and each team owns the components of a system: an inventory team owns the inventory components, a checkout team owns the checkout components. Ownership is exactly that — who answers for a component, its changes, and its operation.

Features: the stack each component needs

Components are not empty repositories. A feature adds the files and workflows a repository needs, so teams adopt a capability instead of hand-rolling it. Typical features cover Docker image builds and automated releases with release_please. Applying a feature is where the map becomes concrete: a component plus its features behaves the way the organization expects.

Supporting infrastructure

Software also depends on resources that are not code: a bucket for a frontend’s static files, a queue, a managed database. Firestartr keeps this supporting infrastructure next to the component that uses it and provisions it with Terraform. The component keeps its place, its owner, and its features, while the infrastructure remains part of the same picture instead of drifting into an unowned set of cloud resources.

Three control surfaces

People do not need another tool to log into. Firestartr meets them on the surfaces they already use:

  • GitHub — where code, reviews, and workflows live.
  • Backstage — where the catalog of services and ownership is browsed.
  • Skill — the conversational surface that turns a request into Firestartr instructions.

GitHub, Backstage, and Skill feed instructions into a central Firestartr card as its three control surfaces

All three drive the same platform, so a change made from any of them is reflected in the same model.

How the Skill carries instructions

The Skill does not act directly. A request travels a fixed path before anything changes:

  1. agent — the AI agent you talk to, with the Firestartr skill loaded.
  2. skill — interprets the request against the map and turns it into Firestartr instructions.
  3. fs-forge — the command-line tool the skill uses to hand those instructions to Firestartr.

Firestartr then applies the change where the software lives: GitHub, AWS, and Azure.

An agent running the Firestartr skill sends instructions through fs-forge to Firestartr, which acts on GitHub, AWS, and Azure

The skill interprets; the CLI executes. Because every stage works from the same map, the result stays consistent with the organization’s structure instead of drifting away from it.

From chaos to a clear map

Put together, the pieces replace the tangle from the start of this page with a structure anyone can read:

  • Every repository is a component with a place in a system.
  • Every system belongs to a domain of the business.
  • People belong to teams, and each team owns the components of a system.

People grouped into team-inventory and team-checkout, each team the owner of one system in the commerce domain, and each system listing its components

The same software, now with a model the organization can rely on — and one Firestartr can act on.

Where to go next

  • Deploying workloads — practical guides for the repositories Firestartr manages: the .firestartr configuration repository, the application and system-service state repositories, claim validation, and migration.
  • Providers — documentation for the providers that back claims, including the Terraform provider.
  • Features — the catalog of features you can apply to repositories.