Getting Started

Podium stores reusable AI agent artifacts in a catalog and translates them into the file formats used by Claude Code, Cursor, Codex, OpenCode, and other harnesses.

Start with the filesystem quickstart. It uses a local directory and podium sync, so the basic authoring and materialization loop is visible before server concepts appear.


Reading Order

In order:

Page What it covers Time
Why Podium What Podium does, when it applies, when a simpler alternative is enough, and how it compares to adjacent tools. Read first when evaluating. ~10 minutes
Quickstart Install the CLI, write one skill, materialize it into Claude Code, and see it load. Filesystem mode has no daemon and no setup beyond a CLI. ~5 minutes
Concepts Vocabulary used throughout the docs: artifact, domain, layer, harness, materialization, visibility, and meta-tools. ~15 minutes
How it works Component overview, deployment modes, where state lives, what runs in your process versus on a server. ~15 minutes

Where to go next

After the quickstart, choose the role-specific guide that fits the task. Many workflows involve both authoring artifacts and consuming them.

Goal Next Why
Write artifacts Authoring guide Artifact types, frontmatter, how DOMAIN.md works, and when to use rule modes and hooks.
Use artifacts in your harness Consuming guide Configure Claude Code / Cursor / OpenCode / Pi / Hermes / Codex / Gemini / Claude Desktop, browse the catalog from the agent, work via the SDK.
Set up Podium for a team or org Deployment guide Pick your mode (filesystem / standalone / standard), then run it. Day-2 operations, progressive adoption of governance, OIDC cookbooks.
Build against the API Reference CLI, HTTP API, frontmatter schema, error codes, glossary.

Table of contents