307- Request-level `network_policy` further restricts access.307- Request-level `network_policy` further restricts access.
308- Requests fail if `allowed_domains` includes domains outside your org allow list.308- Requests fail if `allowed_domains` includes domains outside your org allow list.
309 309
310## Data retention and container lifecycle
311
312Hosted containers used by Hosted Shell and Code Interpreter may write temporary application state to the container filesystem (backed by ephemeral block storage) while the container is active. Container data is deleted when the container expires or is explicitly deleted.
313
314For more details on data controls, see [ZDR and data residency](https://developers.openai.com/api/docs/guides/your-data).
315
316### Download artifacts
317
318Hosted shell can produce downloadable files. Use the same container/files APIs as code interpreter to retrieve artifacts written under `/mnt/data`.
319
320### Additional data controls
321
322If you want to keep content and files ephemeral within the hosted lifecycle, you can inline files in the request and mount inline skills in the container.
323
324Use inline files and inline skills
325
326```javascript
327import fs from "fs";
328import OpenAI from "openai";
329
330const client = new OpenAI();
331
332const inlineZip = fs.readFileSync("csv_insights.zip").toString("base64");
333const reportCsv = fs.readFileSync("report.csv").toString("base64");
334
335const container = await client.containers.create({
336 name: "inline-skill-container",
337 skills: [
338 {
339 type: "inline",
340 name: "csv-insights",
341 description: "Summarize CSV files and produce a markdown report.",
342 source: {
343 type: "base64",
344 media_type: "application/zip",
345 data: inlineZip,
346 },
347 },
348 ],
349});
350
351const response = await client.responses.create({
352 model: "gpt-5.4",
353 tools: [
354 {
355 type: "shell",
356 environment: {
357 type: "container_reference",
358 container_id: container.id,
359 },
360 },
361 ],
362 input: [
363 {
364 role: "user",
365 content: [
366 {
367 type: "input_file",
368 filename: "report.csv",
369 file_data: \`data:text/csv;base64,\${reportCsv}\`,
370 },
371 {
372 type: "input_text",
373 text: "Use the csv-insights skill to summarize report.csv.",
374 },
375 ],
376 },
377 ],
378});
379
380console.log(response.output_text);
381```
382
383```python
384import base64
385from openai import OpenAI
386
387client = OpenAI()
388
389with open("csv_insights.zip", "rb") as f:
390 inline_zip = base64.b64encode(f.read()).decode("utf-8")
391
392with open("report.csv", "rb") as f:
393 base64_string = base64.b64encode(f.read()).decode("utf-8")
394
395container = client.containers.create(
396 name="inline-skill-container",
397 skills=[
398 {
399 "type": "inline",
400 "name": "csv-insights",
401 "description": "Summarize CSV files and produce a markdown report.",
402 "source": {
403 "type": "base64",
404 "media_type": "application/zip",
405 "data": inline_zip,
406 },
407 }
408 ],
409)
410
411response = client.responses.create(
412 model="gpt-5.4",
413 tools=[
414 {
415 "type": "shell",
416 "environment": {
417 "type": "container_reference",
418 "container_id": container.id,
419 },
420 }
421 ],
422 input=[
423 {
424 "role": "user",
425 "content": [
426 {
427 "type": "input_file",
428 "filename": "report.csv",
429 "file_data": f"data:text/csv;base64,{base64_string}",
430 },
431 {
432 "type": "input_text",
433 "text": "Use the csv-insights skill to summarize report.csv.",
434 },
435 ],
436 }
437 ],
438)
439
440print(response.output_text)
441```
442
443
444For follow-up requests, pass the same `container_id` with `container_reference`. The mounted skills and existing container files remain available while the container is active.
445
446### Proactively delete a container
447
448You can explicitly delete the container when the work is done instead of waiting for inactivity expiration.
449
450Delete a container
451
452```javascript
453import OpenAI from "openai";
454
455const client = new OpenAI();
456
457const deleted = await client.containers.delete("container_id");
458
459console.log(deleted);
460```
461
462```python
463from openai import OpenAI
464
465client = OpenAI()
466
467deleted = client.containers.delete("container_id")
468
469print(deleted)
470```
471
472
310## Domain secrets473## Domain secrets
311 474
312Use `domain_secrets` when a domain in your `allowed_domains` list requires private authorization headers, such as `Authorization: Bearer <token>`.475Use `domain_secrets` when a domain in your `allowed_domains` list requires private authorization headers, such as `Authorization: Bearer <token>`.
458```621```
459 622
460 623
461## Download artifacts
462
463Hosted shell can produce downloadable files. Use the same container/files APIs as code interpreter to retrieve artifacts written under `/mnt/data`.
464
465### Data retention and container lifecycle
466
467Hosted shell runs in OpenAI-hosted containers. If your org or project enables Zero Data Retention (ZDR), OpenAI-hosted containers aren't available, so hosted shell can't run in ZDR mode. Hosted shell and Code Interpreter work with Modified Abuse Monitoring (MAM) instead.
468
469If you need shell execution while using ZDR, use [local shell mode](#local-shell-mode) so commands run in infrastructure you manage.
470
471For data controls details, see [ZDR and data residency](https://developers.openai.com/api/docs/guides/your-data).
472
473Hosted shell uses the same container lifecycle as Code Interpreter. Treat hosted containers as ephemeral and store any data you need on your own systems. Files and artifacts created or uploaded in hosted shell live in the container filesystem (for example, under `/mnt/data`) while the container is active so you can reuse the container and download artifacts.
474
475- A hosted shell container expires after 20 minutes of inactivity. Download any files you need while the container is active. When the container expires, OpenAI discards the container data and you can't recover it.
476- You can't reactivate an expired container. Create a new container and upload files again.
477
478## Shell output in Responses624## Shell output in Responses
479 625
480Hosted shell and local shell use the same output item types. Shell runs are represented by paired output items:626Hosted shell and local shell use the same output item types. Shell runs are represented by paired output items:
