> ## Documentation Index
> Fetch the complete documentation index at: https://agno-v2-docs-align-with-readme.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# What is a Runtime?

> A runtime turns your agent into a production service.

<div className="text-center text-3xl font-mono my-10">
  <span>Runtime</span>
  <span className="mx-3 text-orange-500">=</span>
  <span>Agent Server</span>
</div>

The runtime is one of three layers of an [agent platform](/tutorials/agent-platform/overview): the SDK to build, the runtime to run, and the control plane to manage. To ship agents to real users, you need to run them as a horizontally scalable, durable service reachable over an API. That's what the runtime gives you.

AgentOS turns the `Agent(...)` into a FastAPI app with persistent sessions, JWT-based authentication, observability, and interfaces for Slack, Telegram, and more.

## What the runtime gives you

The runtime covers the ground between your agent code and a shipped service:

| Concern               | How AgentOS handles it                                                  |
| --------------------- | ----------------------------------------------------------------------- |
| **HTTP API**          | Auto-generated endpoints for every registered agent, team, and workflow |
| **State**             | Sessions and memory persisted to your `db`                              |
| **Streaming**         | SSE on every run endpoint; tokens and tool calls stream as they happen  |
| **Auth**              | JWT validation with RBAC scopes built-in                                |
| **Scheduling**        | In-process cron that polls the database and fires due jobs              |
| **Observability**     | OpenTelemetry tracing into the same `db`, queryable with SQL            |
| **Interfaces**        | Slack, Telegram, WhatsApp, Discord, A2A, AG-UI                          |
| **Human in the loop** | Pause runs for user confirmation, admin approval, or external execution |

The same `Agent(...)` runs in a script, a test, or an AgentOS service. AgentOS itself is stateless: state lives in `db`, so you can scale horizontally to handle load.

<Note>
  AgentOS also runs teams and workflows, not only single agents. See [Build a
  Product](/runtime/build-a-product).
</Note>

<Note>
  AgentOS can also serve agents built with the Claude Agent SDK, LangGraph and
  DSPy. See [Multi-Framework Support](/agent-os/multi-framework/overview).
</Note>

## What's in this section

| Page                                          | Covers                                                       |
| --------------------------------------------- | ------------------------------------------------------------ |
| [Serve as an API](/runtime/serve-as-api)      | The HTTP endpoints AgentOS generates and how to add your own |
| [Storage](/runtime/storage)                   | Database backends and what gets stored                       |
| [Context](/runtime/context)                   | Knowledge, dependencies, and live context providers          |
| [Human Approval](/runtime/human-approval)     | Pausing runs for confirmation, approval, or external systems |
| [Observability](/runtime/observability)       | Tracing, run history, audit logs                             |
| [Security & Auth](/runtime/security-and-auth) | JWT, RBAC scopes, request isolation                          |
| [Interfaces](/runtime/interfaces)             | Slack, Telegram, WhatsApp, A2A, AG-UI                        |
| [Scheduling](/runtime/scheduling)             | Cron jobs and multi-step workflows                           |
| [Deploy](/runtime/deploy)                     | Docker, Railway, AWS, GCP, scaling                           |
| [Build a Product](/runtime/build-a-product)   | From a working AgentOS to a shipped product                  |

## Next

[Serve as an API →](/runtime/serve-as-api)