Skip to main content

What PlatformSmith is

PlatformSmith gives you isolated containers running coding agents against your code, on infrastructure you can bring yourself.

You point it at a repository (or start from a plain base image), it starts a container on a machine you control, and a coding agent — Claude Code or Codex — runs inside that container with your repository checked out. You talk to the agent, it edits and runs code, and the container is yours until you throw it away.

Three things follow from that sentence, and they are what most of this site explains:

  • The agent runs on your model account. You bring an Anthropic or OpenAI credential; PlatformSmith never supplies a model key and never bills you for tokens. → Credentials and secrets
  • The container runs on your infrastructure. A Docker host or a Kubernetes cluster you register with the platform. PlatformSmith schedules work onto it; it does not own it. → Environments and controllers
  • Everything is scoped to your organization. Workspaces, projects and environments are real, enforced permission boundaries, not display grouping. → Security and tenancy

The shape of it

A project is what code. An environment is where it runs. They are siblings, not nested: you launch a runtime for a project into an environment. Getting that one relationship right makes the rest of the model fall into place — Concepts at a glance does it properly, with the rest of the nouns.

Where to go next

If you want to…Go to
Get from a fresh account to a running agentGet started — about 15 minutes
Understand the nouns before you click anythingConcepts at a glance
Know what a coding-agent credential is and who can see itCredentials and secrets
Evaluate the isolation and permission modelSecurity and tenancy
Automate against the platformAPI and automation
What this site does not cover

This is the product documentation. It describes what you can do in PlatformSmith and how the pieces fit together. It is not an internal engineering reference, and it deliberately documents only the capabilities enabled on the build you are reading it for — see Not yet enabled for what is staged behind that.