2 2
3> For the complete documentation index, see [llms.txt](/llms.txt). Markdown versions of documentation pages are available by appending `.md` to the page URL.3> For the complete documentation index, see [llms.txt](/llms.txt). Markdown versions of documentation pages are available by appending `.md` to the page URL.
4 4
5Agents can plan and complete tasks using tools, work with other agents, and maintain context across steps. Choose a runtime based on where you want orchestration to run and who should manage the state between tasks.5For new agent applications, start with the **[Agents API](https://developers.openai.com/api/docs/guides/agents-api/overview)**. OpenAI runs the Codex harness and manages orchestration, context compaction, and durable sessions. You build the surrounding application, connect tools, and choose where execution happens.
6
7Follow the [Agents API quickstart](https://developers.openai.com/api/docs/guides/agents-api/quickstart) to run a task, stream progress, and continue a session.
6 8
7## Choose your starting point9## Choose your starting point
8 10
9| You want to | Start here |11| You want to | Start here |
10| ------------------------------------------------------------------------------------ | ------------------------------------------------------ |12| --- | --- |
11| Run an agent with the Codex harness managed by OpenAI | [Agents API](https://developers.openai.com/api/docs/guides/agents-api/quickstart) |13| Build a new agent application with a managed runtime | [Agents API quickstart](https://developers.openai.com/api/docs/guides/agents-api/quickstart) |
12| Control the agent loop in your application with reusable agents, tools, and handoffs | [Agents SDK](https://developers.openai.com/api/docs/guides/agents/quickstart) |14| Run the Codex harness in infrastructure you operate | [Codex SDK](https://developers.openai.com/codex/codex-sdk) |
13| Work directly with model responses and control your integration | [Responses API](https://developers.openai.com/api/docs/guides/migrate-to-responses) |15| Call models directly or own the agent loop | [Responses API](https://developers.openai.com/api/docs/guides/migrate-to-responses) |
14| Add an embedded chat experience | [ChatKit](https://developers.openai.com/api/docs/guides/chatkit) |
15 16
16<a id="agents-sdk-vs-responses-api"></a>17<a id="agents-sdk-vs-responses-api"></a>
17 18
19 20
20## Compare agent runtime options21## Compare agent runtime options
21 22
22| | Agents API | Agents SDK | Responses API |23Choose based on what OpenAI manages and what your application needs to control.
23| ------------------------ | ------------------------------------------------------------------------------- | ------------------------------------------------------------------- | --------------------------------------------------------- |
24| **Use for** | Long-running tasks where OpenAI manages the agent and saves its progress | Building agents with custom tools and workflows in your application | Calling models directly or building an agent from scratch |
25| Where the agent runs | OpenAI runs a managed Codex harness | The SDK runs inside your application | Your application, with optional hosted orchestration |
26| Agent integration effort | Low | Medium | High |
27| State between tasks | Saved session configuration, turns, and items | Your storage and SDK sessions, or Responses conversation state | Manual history, response chaining, or Conversations |
28| Tool execution | Service-connected tools, application function handlers, and an optional sandbox | Tools and integrations configured in your application | Hosted tools and tools your application runs |
29| Execution environment | OpenAI hosted sandbox, self-hosted sandbox, or no sandbox | Your runtime and sandbox provider integrations | Your own execution environment |
30| Start here | [Agents API overview](https://developers.openai.com/api/docs/guides/agents-api/overview) | [Agents SDK overview](https://developers.openai.com/api/docs/guides/agents/sdk) | [Responses guide](https://developers.openai.com/api/docs/guides/migrate-to-responses) |
31 24
32The Agents API runs the Codex harness and manages the underlying agent infrastructure so you can focus on what your agents do. It includes automatic context compaction, multi-agent orchestration, programmatic tool calling, and support for MCP servers. See [Architecture](https://developers.openai.com/api/docs/guides/agents-api/architecture).25| Option | What it manages | What you operate |
26| --- | --- | --- |
27| **Agents API** | Hosted Codex harness, orchestration, and durable session state | Your application, tool integrations, and choice of execution environment |
28| **Codex SDK** | Codex harness running in your environment | The harness process, hosting, and application lifecycle |
29| **Responses API** | Model responses and configured hosted capabilities | Application logic and any agent loop you build around the API |
33 30
34The Agents SDK gives your application control over deployment, storage, approvals, and runtime integration. Its runner handles the agent loop and handoffs. See [Running agents](https://developers.openai.com/api/docs/guides/agents/running-agents).31Use the [Agents API architecture guide](https://developers.openai.com/api/docs/guides/agents-api/architecture) to understand the boundary between the hosted harness and your execution environment. For direct model integrations, the Responses API also offers hosted tools and state through response chaining or Conversations; follow its guides for the capabilities you use.
35 32
36 33
37 34
46- Read [Prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching) for shared caching behavior, then [Agents API observability and usage](https://developers.openai.com/api/docs/guides/agents-api/observability) for session accounting.43- Read [Prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching) for shared caching behavior, then [Agents API observability and usage](https://developers.openai.com/api/docs/guides/agents-api/observability) for session accounting.
47 44
48An Agents API session, an SDK session, a Responses conversation, and a sandbox are different resources. Follow the state and cleanup instructions for the runtime you choose.45An Agents API session, an SDK session, a Responses conversation, and a sandbox are different resources. Follow the state and cleanup instructions for the runtime you choose.
46
47## If you use the Agents SDK
48
49The [Agents SDK](https://developers.openai.com/api/docs/guides/agents/sdk) is **feature complete**: major new features are not planned, but maintenance, security fixes, critical bug fixes, and compatibility work continue. You can continue using it for existing applications. See also: [SDK support policy](https://developers.openai.com/api/docs/guides/agents/sdk#important-notice).
50
51For new agent applications, start with the [Agents API](https://developers.openai.com/api/docs/guides/agents-api/quickstart).
