> ## Documentation Index
> Fetch the complete documentation index at: https://docs.omnara.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Concepts

> Key concepts to understand when using Omnara

Omnara provides a stateful API for interacting with agents. This page provides an overview of the core entities you should understand when interacting with the dashboard or API.

## Organizations and Projects

When you join Omnara, you join or create an **organization**. Organizations manage [memberships](/organization/members), own shared team resources, and contain **projects**. Projects contain [agents](/agents/overview), their [configs and profiles](/agents/configuration), and the grants that control which shared resources those agents may use.

## Resources

Shared resources are owned at the org, project, or user scope:

* **[Machines and machine pools](/machines/overview)** (org owned): existing computers you [connect](/machines/connect) (BYO) or [pools](/machines/pools) of ephemeral sandboxes that agents run commands on.
* **[Model providers and models](/organization/model-providers)** (org owned): the model endpoints and configured models agents can use.
* **[Skills](/tools/skills)** (org, project, or user owned): reusable instructions and tools for agents.
* **[Secrets](/organization/secrets)** (org, project, or user owned): credentials referenced by model providers, machine and their environments, and [MCP servers](/tools/mcp).

## Grants

A project can use a resource only if it owns it or holds an explicit **grant**. Org and user ownership alone don't make a resource visible to a project — granting is the deliberate step that shares a model, pool, machine, secret, or skill with a project. Own each resource at the narrowest scope that needs it, and grant outward.

## Agents

[Agents](/agents/overview) are the core feature of Omnara. Define an agent's model, [tools](/tools/permissions), and instructions in a [config or profile](/agents/configuration). Then Omnara runs the agent loop and runtime.

Once an agent has started, interact with it by:

* Reading or [streaming](/events/streaming) it's event log
* [Sending inputs](/events/sending-input), such as messages or file attachments
* Resolving [questions and approvals](/events/interactions)

## Continue

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Write a config, create the grants, and launch an agent with the dashboard, the API, or the CLI
  </Card>

  <Card title="API reference" icon="code" href="/api/overview">
    Conventions, authentication, errors, and every endpoint
  </Card>
</CardGroup>