9commands, previews, and resumable work all need an environment the agent can9commands, previews, and resumable work all need an environment the agent can
10inspect and change.10inspect and change.
11 11
12Sandbox agents are currently only available in the Python Agents SDK.12Sandbox agents are available in the TypeScript and Python Agents SDKs. They
13 are in beta, so API details, defaults, and supported capabilities may change.
13 14
14Use sandboxes when the agent needs to manipulate files, run commands, mount a15Use sandboxes when the agent needs to manipulate files, run commands, mount a
15data room, produce artifacts, expose a service, or continue stateful work16data room, produce artifacts, expose a service, or continue stateful work
71## What sandboxes add72## What sandboxes add
72 73
73`SandboxAgent` is still an `Agent`. It keeps the usual agent surface, including74`SandboxAgent` is still an `Agent`. It keeps the usual agent surface, including
74`instructions`, `prompt`, `tools`, `handoffs`, `mcp_servers`, `model_settings`,75`instructions`, `prompt`, `tools`, `handoffs`, MCP servers, model settings,
75`output_type`, guardrails, and hooks. What changes is the execution boundary:76structured output, guardrails, and hooks. What changes is the execution boundary:
76the runner prepares the agent against a live sandbox session that owns files,77the runner prepares the agent against a live sandbox session that owns files,
77commands, ports, and provider-specific isolation.78commands, ports, and provider-specific isolation.
78 79
83| Capabilities | Sandbox-native behavior attached to the agent | Which sandbox tools, instructions, or runtime behavior does this agent need? |84| Capabilities | Sandbox-native behavior attached to the agent | Which sandbox tools, instructions, or runtime behavior does this agent need? |
84| Sandbox client | The provider integration | Where should the live workspace run: Unix-local, Docker, or a hosted provider? |85| Sandbox client | The provider integration | Where should the live workspace run: Unix-local, Docker, or a hosted provider? |
85| Sandbox session | The live execution environment | Where do commands run, files change, ports open, and provider state live? |86| Sandbox session | The live execution environment | Where do commands run, files change, ports open, and provider state live? |
86| `SandboxRunConfig` | Per-run sandbox session source, client options, and fresh inputs | Should this run inject, resume, or create the sandbox session? |87| Sandbox run config | Per-run sandbox session source, client options, and fresh inputs | Should this run inject, resume, or create the sandbox session? |
87| Saved state | `RunState`, `session_state`, and snapshots | How should later runs reconnect to work or seed a new workspace? |88| Saved state | `RunState`, serialized session state, and snapshots | How should later runs reconnect to work or seed a new workspace? |
88 89
89Sandbox-specific defaults belong on `SandboxAgent`. Per-run sandbox-session90Sandbox-specific defaults belong on `SandboxAgent`. Per-run sandbox-session
90choices belong in `SandboxRunConfig`.91choices belong in the run's sandbox configuration.
91 92
92Sandbox agents also don't change what a turn means. A turn is still a model93Sandbox agents also don't change what a turn means. A turn is still a model
93step, not a single shell command or sandbox action. Some work may stay inside94step, not a single shell command or sandbox action. Some work may stay inside
112| Manifest input | Use it for |113| Manifest input | Use it for |
113| ------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------- |114| ------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------- |
114| `File`, `Dir` | Small synthetic inputs, helper files, or output directories. |115| `File`, `Dir` | Small synthetic inputs, helper files, or output directories. |
115| `LocalFile`, `LocalDir` | Host files or directories to materialize into the sandbox. |116| Local file or directory | Host files or directories to materialize into the sandbox. |
116| `GitRepo` | A repository to fetch into the workspace. |117| Git repo | A repository to fetch into the workspace. |
117| `S3Mount`, `GCSMount`, `R2Mount`, `AzureBlobMount`, `BoxMount`, `S3FilesMount` | External storage to make available inside the sandbox. |118| `S3Mount`, `GCSMount`, `R2Mount`, `AzureBlobMount`, `BoxMount`, `S3FilesMount` | External storage to make available inside the sandbox. |
118| `environment` | Environment variables the sandbox needs when it starts. |119| `environment` | Environment variables the sandbox needs when it starts. |
119| `users` and `groups` | Sandbox-local OS accounts and groups for providers that support account provisioning. |120| `users` and `groups` | Sandbox-local OS accounts and groups for providers that support account provisioning. |
178| [`Memory`](#persist-memory-across-runs) | Follow-on runs should read or generate memory artifacts. | Requires `Shell`; live memory updates also require `Filesystem`. |179| [`Memory`](#persist-memory-across-runs) | Follow-on runs should read or generate memory artifacts. | Requires `Shell`; live memory updates also require `Filesystem`. |
179| `Compaction` | Long-running flows need context trimming. | Adjusts model behavior and input handling after compaction items. |180| `Compaction` | Long-running flows need context trimming. | Adjusts model behavior and input handling after compaction items. |
180 181
181By default, a `SandboxAgent` uses `Capabilities.default()`, which includes182By default, a `SandboxAgent` includes filesystem, shell, and compaction
182`Filesystem()`, `Shell()`, and `Compaction()`. If you pass a `capabilities`183capabilities. If you pass a `capabilities` list, it replaces the default list,
183list, it replaces the default list, so include any default capabilities the184so include any default capabilities the agent still needs.
184agent still needs.
185 185
186Prefer built-in capabilities when they fit. Write a custom capability only when186Prefer built-in capabilities when they fit. Write a custom capability only when
187you need a sandbox-specific tool or instruction surface that the built-ins don't187you need a sandbox-specific tool or instruction surface that the built-ins don't
193the agent starts. Use the `Skills` capability so the agent can discover that193the agent starts. Use the `Skills` capability so the agent can discover that
194working context during the run.194working context during the run.
195 195
196Load skills
197
198```typescript
199import {
200 Capabilities,
201 SandboxAgent,
202 gitRepo,
203 skills,
204} from "@openai/agents/sandbox";
205
206const agent = new SandboxAgent({
207 name: "Tax prep assistant",
208 instructions: "Use the mounted skill before preparing the return.",
209 capabilities: [
210 ...Capabilities.default(),
211 skills({
212 from: gitRepo({
213 repo: "owner/tax-prep-skills",
214 ref: "main",
215 }),
216 }),
217 ],
218});
219```
220
196```python221```python
197from agents.sandbox import SandboxAgent222from agents.sandbox import SandboxAgent
198from agents.sandbox.capabilities import Capabilities, Skills223from agents.sandbox.capabilities import Capabilities, Skills
207)232)
208```233```
209 234
235
210Choose the skill source based on how you want it materialized:236Choose the skill source based on how you want it materialized:
211 237
212- Use `Skills(lazy_from=LocalDirLazySkillSource(...))` for larger local skill directories when you want the model to discover the index first and load only what it needs.238- Use a lazy local directory source for larger local skill directories when you want the model to discover the index first and load only what it needs.
213- Use `Skills(from_=LocalDir(src=...))` for a small local bundle to stage up front.239- Use a local directory source for a small local bundle to stage up front.
214- Use `Skills(from_=GitRepo(repo=..., ref=...))` when the skills bundle has its own release cadence or many sandboxes use it.240- Use a Git repo source when the skills bundle has its own release cadence or many sandboxes use it.
215 241
216### Expose previews and ports242### Expose previews and ports
217 243
2301. Build a `Manifest` that describes the workspace.2561. Build a `Manifest` that describes the workspace.
2312. Create a `SandboxAgent` with the capabilities the model needs.2572. Create a `SandboxAgent` with the capabilities the model needs.
2323. Choose a sandbox client for the environment where work should run.2583. Choose a sandbox client for the environment where work should run.
2334. Run the agent with `RunConfig(sandbox=SandboxRunConfig(...))`.2594. Run the agent with the per-run sandbox configuration.
2345. Inspect, copy, resume, or snapshot the artifacts that matter to your application.2605. Inspect, copy, resume, or snapshot the artifacts that matter to your application.
235 261
236Start with Unix-local for local development on macOS or Linux. It gives you the262Start with Unix-local for local development on macOS or Linux. It gives you the
237smallest local loop because the runner can create a temporary workspace from the263smallest local loop because the runner can create a temporary workspace from the
238agent's `default_manifest` and clean it up after the run.264agent's default manifest and clean it up after the run.
265
266Run a Unix-local sandbox agent
267
268```typescript
269import { run } from "@openai/agents";
270import {
271 Manifest,
272 SandboxAgent,
273 file,
274 shell,
275} from "@openai/agents/sandbox";
276import { UnixLocalSandboxClient } from "@openai/agents/sandbox/local";
277
278const manifest = new Manifest({
279 entries: {
280 "account_brief.md": file({
281 content:
282 "# Northwind Health\\n\\n" +
283 "- Segment: Mid-market healthcare analytics provider.\\n" +
284 "- Renewal date: 2026-04-15.\\n",
285 }),
286 "implementation_risks.md": file({
287 content:
288 "# Delivery risks\\n\\n" +
289 "- Security questionnaire is not complete.\\n" +
290 "- Procurement requires final legal language by April 1.\\n",
291 }),
292 },
293});
294
295const agent = new SandboxAgent({
296 name: "Renewal Packet Analyst",
297 model: "gpt-5.5",
298 instructions:
299 "Review the workspace before answering. Keep the response concise, " +
300 "business-focused, and cite the file names that support each conclusion.",
301 defaultManifest: manifest,
302 capabilities: [shell()],
303});
304
305const result = await run(
306 agent,
307 "Summarize the renewal blockers and recommend the next two actions.",
308 {
309 sandbox: {
310 client: new UnixLocalSandboxClient(),
311 },
312 },
313);
314
315console.log(result.finalOutput);
316```
239 317
240```python318```python
241import asyncio319import asyncio
251 entries={329 entries={
252 "account_brief.md": File(330 "account_brief.md": File(
253 content=(331 content=(
254 b"# Northwind Health\n\n"332 b"# Northwind Health\\n\\n"
255 b"- Segment: Mid-market healthcare analytics provider.\n"333 b"- Segment: Mid-market healthcare analytics provider.\\n"
256 b"- Renewal date: 2026-04-15.\n"334 b"- Renewal date: 2026-04-15.\\n"
257 )335 )
258 ),336 ),
259 "implementation_risks.md": File(337 "implementation_risks.md": File(
260 content=(338 content=(
261 b"# Delivery risks\n\n"339 b"# Delivery risks\\n\\n"
262 b"- Security questionnaire is not complete.\n"340 b"- Security questionnaire is not complete.\\n"
263 b"- Procurement requires final legal language by April 1.\n"341 b"- Procurement requires final legal language by April 1.\\n"
264 )342 )
265 ),343 ),
266 }344 }
293asyncio.run(main())371asyncio.run(main())
294```372```
295 373
296For a complete local example, see [`unix_local_runner.py`][sdk-example-unix-local-runner].374
375For complete local examples, see the TypeScript [sandbox agent quickstart][sdk-js-example-basic] and Python [`unix_local_runner.py`][sdk-example-unix-local-runner].
297 376
298### Switch providers377### Switch providers
299 378
304This example uses Docker for local container isolation. Hosted providers follow383This example uses Docker for local container isolation. Hosted providers follow
305the same pattern with their own client classes and options.384the same pattern with their own client classes and options.
306 385
386Switch to Docker
387
388```typescript
389import { run } from "@openai/agents";
390import { SandboxAgent } from "@openai/agents/sandbox";
391import { DockerSandboxClient } from "@openai/agents/sandbox/local";
392
393const agent = new SandboxAgent({
394 name: "Workspace reviewer",
395 model: "gpt-5.5",
396 instructions: "Inspect the sandbox workspace before answering.",
397});
398
399const result = await run(agent, "Inspect the workspace.", {
400 sandbox: {
401 client: new DockerSandboxClient({
402 image: "node:22-bookworm-slim",
403 }),
404 },
405});
406
407console.log(result.finalOutput);
408```
409
307```python410```python
308from docker import from_env as docker_from_env411from docker import from_env as docker_from_env
309 412
328)431)
329```432```
330 433
331For runnable examples, see [`basic.py`][sdk-example-basic] for provider434
332selection, [`docker_runner.py`][sdk-example-docker-runner] for Docker, and435For runnable examples, see the TypeScript [sandbox clients guide][sdk-js-sandbox-clients] and [basic example][sdk-js-example-basic], plus Python [`basic.py`][sdk-example-basic] for provider selection, [`docker_runner.py`][sdk-example-docker-runner] for Docker, and [`main.py`][sdk-example-dataroom-qa] for a data-room flow in the SDK repository.
333[`main.py`][sdk-example-dataroom-qa] for a data-room flow in the SDK
334repository.
335 436
336### Advanced patterns437### Advanced patterns
337 438
357Keep three state concepts separate:458Keep three state concepts separate:
358 459
359| State surface | Restores | Use when |460| State surface | Restores | Use when |
360| --------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ |461| ------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ |
361| `RunState` | Harness-side state such as model items, tool state, approvals, and active agent position. | The runner should carry the workflow forward across pauses. |462| `RunState` | Harness-side state such as model items, tool state, approvals, and active agent position. | The runner should carry the workflow forward across pauses. |
362| `session_state` | A serialized sandbox session that a client can reconnect to. | Your app or job system stores provider session state directly. |463| Session state | A serialized sandbox session that a client can reconnect to. | Your app or job system stores provider session state directly. |
363| `snapshot` | Saved workspace contents used to seed a fresh sandbox session. | A new run should start from saved files and artifacts, not an empty workspace. |464| `snapshot` | Saved workspace contents used to seed a fresh sandbox session. | A new run should start from saved files and artifacts, not an empty workspace. |
364 465
365In practice, the runner resolves the sandbox session in this order:466In practice, the runner resolves the sandbox session in this order:
366 467
3671. If you pass `run_config.sandbox.session`, the runner reuses that live sandbox session directly.4681. If you pass a live sandbox session, the runner reuses that session directly.
3682. Otherwise, if the run is resuming from `RunState`, the runner resumes from the stored sandbox session state.4692. Otherwise, if the run is resuming from `RunState`, the runner resumes from the stored sandbox session state.
3693. Otherwise, if you pass `run_config.sandbox.session_state`, the runner resumes from that explicit serialized sandbox state.4703. Otherwise, if you pass explicit serialized sandbox state, the runner resumes from that state.
3704. Otherwise, the runner creates a fresh sandbox session. For that fresh session, it uses `run_config.sandbox.manifest` when provided, or `agent.default_manifest` if not.4714. Otherwise, the runner creates a fresh sandbox session. For that fresh session, it uses the per-run manifest when provided, or the agent's default manifest if not.
371 472
372The sandbox resume example serializes the stopped session state, resumes it473The sandbox resume example serializes the stopped session state, resumes it
373through the same client, and then passes the resumed session back into the next474through the same client, and then passes the resumed session back into the next
374run:475run:
375 476
477Serialize and resume sandbox state
478
479```typescript
480import { run } from "@openai/agents";
481import { Manifest, SandboxAgent } from "@openai/agents/sandbox";
482import { UnixLocalSandboxClient } from "@openai/agents/sandbox/local";
483
484const manifest = new Manifest();
485const client = new UnixLocalSandboxClient({
486 snapshot: { type: "local", baseDir: "/tmp/my-sandbox-snapshots" },
487});
488const agent = new SandboxAgent({
489 name: "Workspace builder",
490 model: "gpt-5.5",
491 instructions: "Inspect the sandbox workspace before answering.",
492});
493
494const session = await client.create({ manifest });
495let conversation: any[] = [];
496let frozenSessionState;
497
498try {
499 const firstResult = await run(agent, "Build the first version of the app.", {
500 maxTurns: 20,
501 sandbox: { session },
502 });
503
504 conversation = firstResult.history;
505 frozenSessionState = await client.serializeSessionState?.(session.state);
506} finally {
507 await session.close?.();
508}
509
510if (!frozenSessionState || !client.deserializeSessionState || !client.resume) {
511 throw new Error("Sandbox client does not support session resume.");
512}
513
514const resumedSession = await client.resume(
515 await client.deserializeSessionState(frozenSessionState),
516);
517
518try {
519 conversation.push({
520 role: "user",
521 content: "Continue from the existing workspace and add tests.",
522 });
523
524 await run(agent, conversation, {
525 maxTurns: 20,
526 sandbox: { session: resumedSession },
527 });
528} finally {
529 await resumedSession.close?.();
530}
531```
532
376```python533```python
377async with session:534async with session:
378 first_result = await Runner.run(535 first_result = await Runner.run(
421 579
422This split lets the harness resume the agent loop while the sandbox provider580This split lets the harness resume the agent loop while the sandbox provider
423restores or recreates the workspace. Current sample code for these paths lives581restores or recreates the workspace. Current sample code for these paths lives
424in [`main.py`][sdk-example-sandbox-resume],582in the TypeScript [resume session state example][sdk-js-example-resume] and
583Python [`main.py`][sdk-example-sandbox-resume] and
425[`sandbox_agent_with_remote_snapshot.py`][sdk-example-remote-snapshot].584[`sandbox_agent_with_remote_snapshot.py`][sdk-example-remote-snapshot].
426 585
427## Persist memory across runs586## Persist memory across runs
436turn. Resume and snapshots preserve workspace state; memory preserves reusable595turn. Resume and snapshots preserve workspace state; memory preserves reusable
437guidance about work that happened in the workspace.596guidance about work that happened in the workspace.
438 597
598Enable sandbox memory
599
600```typescript
601import {
602 Manifest,
603 SandboxAgent,
604 filesystem,
605 memory,
606 shell,
607} from "@openai/agents/sandbox";
608
609const manifest = new Manifest();
610
611const agent = new SandboxAgent({
612 name: "Memory-enabled reviewer",
613 instructions:
614 "Inspect the workspace and retain useful lessons for follow-up runs.",
615 defaultManifest: manifest,
616 capabilities: [memory(), filesystem(), shell()],
617});
618```
619
439```python620```python
440from agents.sandbox.capabilities import Filesystem, Memory, Shell621from agents.sandbox.capabilities import Filesystem, Memory, Shell
441 622
447)628)
448```629```
449 630
450`Memory()` enables both reads and generation by default. Memory reads require631
451`Shell` so the agent can search and open memory files. By default, live memory632Memory enables both reads and generation by default. Memory reads require shell
452updates also require `Filesystem`, so the agent can repair stale memory or633access so the agent can search and open memory files. By default, live memory
634updates also require filesystem access, so the agent can repair stale memory or
453update memory when the user asks.635update memory when the user asks.
454 636
455Memory reads use progressive disclosure. The SDK injects `memory_summary.md` at637Memory reads use progressive disclosure. The SDK injects `memory_summary.md` at
457relevant, and it opens rollout summaries only when it needs more detail.639relevant, and it opens rollout summaries only when it needs more detail.
458 640
459| Memory mode | Use it when |641| Memory mode | Use it when |
460| ----------------------- | ----------------------------------------------------------------------- |642| -------------------- | ----------------------------------------------------------------------- |
461| `Memory()` | The agent should read existing memory and generate new memory. |643| Default read/write | The agent should read existing memory and generate new memory. |
462| `Memory(generate=None)` | The agent should read memory but not generate new memory after the run. |644| Read-only memory | The agent should read memory but not generate new memory after the run. |
463| `Memory(read=None)` | The run should generate memory without using existing memory. |645| Generate-only memory | The run should generate memory without using existing memory. |
464| `MemoryReadConfig` | You need to disable live updates with `live_update=False`. |646| Read config | You need to disable live updates. |
465| `MemoryGenerateConfig` | You need to tune generation, such as `extra_prompt`. |647| Generate config | You need to tune generation, such as the extra prompt. |
466| `MemoryLayoutConfig` | Agents need isolated memory layouts in the same sandbox workspace. |648| Layout config | Agents need isolated memory layouts in the same sandbox workspace. |
467 649
468By default, memory artifacts live in the sandbox workspace:650By default, memory artifacts live in the sandbox workspace:
469 651
491session state, starting from a snapshot, or mounting persistent storage such as673session state, starting from a snapshot, or mounting persistent storage such as
492S3.674S3.
493 675
494For multi-turn sandbox chats, use a stable SDK `Session` such as676For multi-turn sandbox chats, use a stable SDK session together with the same
495`SQLiteSession(...)` together with the same live sandbox session. Memory groups677live sandbox session. Memory groups runs by the explicit conversation ID, then
496runs by `conversation_id`, then SDK `Session.session_id`, then678the SDK session ID, then the run group ID, and finally a generated per-run ID.
497`RunConfig.group_id`, and finally a generated per-run ID. The sandbox session679The sandbox session ID identifies the live workspace; it's not the memory
498ID identifies the live workspace; it's not the memory conversation ID.680conversation ID.
499 681
500For runnable examples, see [`memory.py`][sdk-example-memory] for a local682For runnable examples, see the TypeScript [memory guide][sdk-js-sandbox-memory],
501snapshot flow, [`memory_s3.py`][sdk-example-memory-s3] for S3-backed memory683plus Python [`memory.py`][sdk-example-memory] for a local snapshot flow,
502storage, and [`memory_multi_agent_multiturn.py`][sdk-example-memory-multi-agent]684[`memory_s3.py`][sdk-example-memory-s3] for S3-backed memory storage, and
503for separate memory layouts across agents.685[`memory_multi_agent_multiturn.py`][sdk-example-memory-multi-agent] for separate
686memory layouts across agents.
504 687
505## Compose sandbox agents688## Compose sandbox agents
506 689
510workspace-heavy part of a workflow to a sandbox agent. The top-level run693workspace-heavy part of a workflow to a sandbox agent. The top-level run
511continues, but the sandbox agent becomes the active agent for the next turn.694continues, but the sandbox agent becomes the active agent for the next turn.
512 695
513Use `agent.as_tool()` when an outer orchestrator should call one or more696Use agents as tools when an outer orchestrator should call one or more sandbox
514sandbox agents as nested tools. Each sandbox tool-agent can have its own697agents as nested tools. Each sandbox tool-agent can have its own sandbox run
515`RunConfig(sandbox=SandboxRunConfig(...))`, sandbox client, manifest, and698configuration, sandbox client, manifest, and provider options.
516provider options.
517 699
518For examples, see [`handoffs.py`][sdk-example-handoffs] and700For examples, see [`handoffs.py`][sdk-example-handoffs] and
519[`sandbox_agents_as_tools.py`][sdk-example-agents-as-tools].701[`sandbox_agents_as_tools.py`][sdk-example-agents-as-tools].
533| Blaxel | `BlaxelSandboxClient` | <a href="https://docs.blaxel.ai/Sandboxes/Overview">Sandbox overview</a> |715| Blaxel | `BlaxelSandboxClient` | <a href="https://docs.blaxel.ai/Sandboxes/Overview">Sandbox overview</a> |
534| Cloudflare | `CloudflareSandboxClient` | <a href="https://developers.cloudflare.com/sandbox/">Sandbox documentation</a><br /><a href="https://docs.cloudflare.com/sandbox/tutorials/openai-agents/">OpenAI Agents tutorial</a><br /><a href="https://github.com/cloudflare/sandbox-sdk/tree/main/bridge/examples">Sandbox Bridge examples</a> |716| Cloudflare | `CloudflareSandboxClient` | <a href="https://developers.cloudflare.com/sandbox/">Sandbox documentation</a><br /><a href="https://docs.cloudflare.com/sandbox/tutorials/openai-agents/">OpenAI Agents tutorial</a><br /><a href="https://github.com/cloudflare/sandbox-sdk/tree/main/bridge/examples">Sandbox Bridge examples</a> |
535| Daytona | `DaytonaSandboxClient` | <a href="https://www.daytona.io/docs/en/sandboxes/">Sandbox documentation</a><br /><a href="https://www.daytona.io/docs/en/guides/openai-agents/openai-agents-sdk-with-sandboxes">OpenAI Agents SDK guide</a> |717| Daytona | `DaytonaSandboxClient` | <a href="https://www.daytona.io/docs/en/sandboxes/">Sandbox documentation</a><br /><a href="https://www.daytona.io/docs/en/guides/openai-agents/openai-agents-sdk-with-sandboxes">OpenAI Agents SDK guide</a> |
536| Docker | `DockerSandboxClient` | <a href="https://docs.docker.com/">Docker documentation</a><br /><a href="https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/docker/docker_runner.py">Docker SDK example</a> |718| Docker | `DockerSandboxClient` | <a href="https://docs.docker.com/">Docker documentation</a><br /><a href="https://github.com/openai/openai-agents-js/blob/main/examples/docs/sandbox-agents/docker-client.ts">TypeScript Docker SDK example</a><br /><a href="https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/docker/docker_runner.py">Python Docker SDK example</a> |
537| E2B | `E2BSandboxClient` | <a href="https://e2b.dev/docs">Sandbox documentation</a><br /><a href="https://e2b.dev/docs/agents/openai-agents-sdk">OpenAI Agents SDK guide</a><br /><a href="https://e2b.dev/blog/e2b-is-now-in-agents-sdk">Launch blog</a> |719| E2B | `E2BSandboxClient` | <a href="https://e2b.dev/docs">Sandbox documentation</a><br /><a href="https://e2b.dev/docs/agents/openai-agents-sdk">OpenAI Agents SDK guide</a><br /><a href="https://e2b.dev/blog/e2b-is-now-in-agents-sdk">Launch blog</a> |
538| Modal | `ModalSandboxClient` | <a href="https://modal.com/docs/guide/sandboxes">Sandbox guide</a><br /><a href="https://modal.com/blog/building-with-modal-and-the-openai-agent-sdk">Integration blog</a><br /><a href="https://github.com/modal-labs/openai-agents-python-example">Example repo</a><br /><a href="https://github.com/modal-labs/openai-agents-python-example?tab=readme-ov-file#modal-extension-reference">Modal extension reference</a> |720| Modal | `ModalSandboxClient` | <a href="https://modal.com/docs/guide/sandboxes">Sandbox guide</a><br /><a href="https://modal.com/blog/building-with-modal-and-the-openai-agent-sdk">Integration blog</a><br /><a href="https://github.com/modal-labs/openai-agents-python-example">Example repo</a><br /><a href="https://github.com/modal-labs/openai-agents-python-example?tab=readme-ov-file#modal-extension-reference">Modal extension reference</a> |
539| Runloop | `RunloopSandboxClient` | <a href="https://docs.runloop.ai/docs/devboxes/overview">Devbox overview</a><br /><a href="https://docs.runloop.ai/docs/devboxes/tunnels">Tunnels</a> |721| Runloop | `RunloopSandboxClient` | <a href="https://docs.runloop.ai/docs/devboxes/overview">Devbox overview</a><br /><a href="https://docs.runloop.ai/docs/devboxes/tunnels">Tunnels</a> |
540| Unix-local | `UnixLocalSandboxClient` | <a href="https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/unix_local_runner.py">Local SDK example</a> |722| Unix-local | `UnixLocalSandboxClient` | <a href="https://github.com/openai/openai-agents-js/blob/main/examples/docs/sandbox-agents/basic.ts">TypeScript local SDK example</a><br /><a href="https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/unix_local_runner.py">Python local SDK example</a> |
541| Vercel | `VercelSandboxClient` | <a href="https://vercel.com/docs/vercel-sandbox">Sandbox documentation</a><br /><a href="https://vercel.com/kb/guide/building-an-agent-with-openai-agents-sdk-and-vercel-sandbox">OpenAI Agents SDK guide</a><br /><a href="https://vercel.com/templates/template/openai-agents-sdk-with-fastapi">FastAPI template</a><br /><a href="https://github.com/vercel-labs/openai-agents-fastapi-starter">Sample app</a> |723| Vercel | `VercelSandboxClient` | <a href="https://vercel.com/docs/vercel-sandbox">Sandbox documentation</a><br /><a href="https://vercel.com/kb/guide/building-an-agent-with-openai-agents-sdk-and-vercel-sandbox">OpenAI Agents SDK guide</a><br /><a href="https://vercel.com/templates/template/openai-agents-sdk-with-fastapi">FastAPI template</a><br /><a href="https://github.com/vercel-labs/openai-agents-fastapi-starter">Sample app</a> |
542 724
543[sdk-example-agents-as-tools]: https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/sandbox_agents_as_tools.py725[sdk-example-agents-as-tools]: https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/sandbox_agents_as_tools.py
554[sdk-example-sandbox-resume]: https://github.com/openai/openai-agents-python/tree/main/examples/sandbox/tutorials/sandbox_resume736[sdk-example-sandbox-resume]: https://github.com/openai/openai-agents-python/tree/main/examples/sandbox/tutorials/sandbox_resume
555[sdk-example-unix-local-runner]: https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/unix_local_runner.py737[sdk-example-unix-local-runner]: https://github.com/openai/openai-agents-python/blob/main/examples/sandbox/unix_local_runner.py
556[sdk-example-vision-clone]: https://github.com/openai/openai-agents-python/tree/main/examples/sandbox/tutorials/vision_website_clone738[sdk-example-vision-clone]: https://github.com/openai/openai-agents-python/tree/main/examples/sandbox/tutorials/vision_website_clone
739[sdk-js-example-basic]: https://github.com/openai/openai-agents-js/blob/main/examples/docs/sandbox-agents/basic.ts
740[sdk-js-example-resume]: https://github.com/openai/openai-agents-js/blob/main/examples/docs/sandbox-agents/resume-session-state.ts
741[sdk-js-sandbox-clients]: https://openai.github.io/openai-agents-js/guides/sandbox-agents/clients
742[sdk-js-sandbox-memory]: https://openai.github.io/openai-agents-js/guides/sandbox-agents/memory
