One read of Claude Documentationclaude-docs-20261029T173708Z
12 pages moved out of 280 read.
Pages moved
12
significant first
Pages read
280
in this capture
Captured
17:37 UTC
Corpus hash
f20e0a21946d
corpus-hash
What this read moved
1-12 of 12office-agents/excel Changed · +8 / -3 lines
from line 219
219219Data is cached for a number of hours after deletion so users can access
220220context in recently closed workbooks.
221221
222Chat history is stored locally in your browser using IndexedDB.
223Conversations are not stored on Anthropic's servers, are not synced
224across devices, and can be cleared from Settings at any time.
222Chat history is stored locally in your browser using IndexedDB. It is
223not synced across devices and can be cleared from Settings at any time.
225224Reinstalling the add-in or switching between Claude add-ins does not
226225remove it. See [Data storage and retention](https://claude.com/docs/office-agents/data-storage.md)
227226for where it sits on disk and how long it is kept.
227
228When you sign in with a Claude account, the add-in sends your
229conversations to Anthropic, including the content Claude reads from your
230workbook and any files you attach. Clearing chat history removes only
231the copy in your browser. It does not delete the messages or attached
232files that Anthropic keeps.
228233
229234Claude for Excel does not inherit custom data retention settings your
230235organization might have set. Activity is not included in Enterprise
office-agents/outlook Changed · +13 / -8 lines
from line 428
428428
429429## Chat history
430430
431Chat history is stored locally in your browser using IndexedDB.
432Conversations are not stored on Anthropic's servers and are not synced
433across devices or browsers. You can clear all chat history from Settings
434at any time. The local store is also cleared when you clear your browser
435data. Chat history is specific to the combination of the add-in surface,
436your user ID, and your organization ID, so your Excel and Outlook
437histories are separate. Reinstalling the add-in or switching between
438Claude add-ins does not remove it. See
431Chat history is stored locally in your browser using IndexedDB. It is
432not synced across devices or browsers. You can clear all chat history
433from Settings at any time. The local store is also cleared when you
434clear your browser data. Chat history is specific to the combination of
435the add-in surface, your user ID, and your organization ID, so your
436Excel and Outlook histories are separate. Reinstalling the add-in or
437switching between Claude add-ins does not remove it. See
439438[Data storage and retention](https://claude.com/docs/office-agents/data-storage.md) for where it
440439sits on disk and how long it is kept.
440
441When you sign in with a Claude account, the add-in sends your
442conversations to Anthropic, including the content Claude reads from your
443mailbox and calendar, and any files you attach. Clearing chat history
444removes only the copy in your browser. It does not delete the messages
445or attached files that Anthropic keeps.
441446
442447## Data retention and audit
443448
office-agents/powerpoint Changed · +8 / -3 lines
from line 252
252252Data is cached for a number of hours after deletion so users can access
253253context in recently closed presentations.
254254
255Chat history is stored locally in your browser using IndexedDB.
256Conversations are not stored on Anthropic's servers, are not synced
257across devices, and can be cleared from Settings at any time.
255Chat history is stored locally in your browser using IndexedDB. It is
256not synced across devices and can be cleared from Settings at any time.
258257Reinstalling the add-in or switching between Claude add-ins does not
259258remove it. See [Data storage and retention](https://claude.com/docs/office-agents/data-storage.md)
260259for where it sits on disk and how long it is kept.
260
261When you sign in with a Claude account, the add-in sends your
262conversations to Anthropic, including the content Claude reads from your
263presentation and any files you attach. Clearing chat history removes
264only the copy in your browser. It does not delete the messages or
265attached files that Anthropic keeps.
261266
262267Claude for PowerPoint does not inherit custom data retention settings
263268your organization might have set. Activity is not included in Enterprise
office-agents/third-party-platforms Changed · +95 / -25 lines
### Skill limits for a bootstrap endpoint
from line 5
55Organizations using Amazon Bedrock, Google Cloud Vertex AI, Azure AI
66Foundry, or an LLM gateway can deploy Claude's Office add-ins without
77requiring individual Claude accounts. The add-in connects through your
8organization's infrastructure, keeping prompts and responses within your
9trust boundary.
8organization's infrastructure, sends prompts to the provider your
9organization chooses, and receives Claude's responses from it. See
10[What Anthropic collects](#what-anthropic-collects) for what reaches
11Anthropic on Amazon Bedrock, Google Cloud Vertex AI, and Azure AI
12Foundry.
1013
1114## Connection paths
1215
from line 112
109112| `oauth2.googleapis.com` | If using Vertex AI direct | Google OAuth token exchange and refresh. |
110113| `aiplatform.googleapis.com` | If using Vertex AI direct | Vertex AI global inference endpoint. |
111114| `<region>-aiplatform.googleapis.com` | If using Vertex AI direct | Vertex AI regional inference endpoint; replace `<region>` with your GCP region. |
115| `aiplatform.us.rep.googleapis.com`, `aiplatform.eu.rep.googleapis.com` | If using Vertex AI direct with region `us` or `eu` | Vertex AI multi-region inference endpoints. |
112116| `<resource>.services.ai.azure.com` | If using Foundry direct | Azure AI Foundry inference endpoint; replace `<resource>` with your resource name. |
113117| Your Vertex AI gateway URL | If using Vertex AI direct through your own gateway | The gateway or proxy set in `gcp_base_url`. Connections made while it is set do not call `aiplatform.googleapis.com` or `<region>-aiplatform.googleapis.com`. Google sign-in still uses `accounts.google.com` and `oauth2.googleapis.com`. |
114118| Your Foundry gateway URL | If using Foundry direct through your own gateway | The gateway or proxy set in `azure_base_url`. Connections made while it is set do not call `<resource>.services.ai.azure.com`. |
from line 126
122126`claude.ai`, and the add-in reads your organization's settings from
123127`api.anthropic.com`. The `claude.ai` sign-in and approval pages run an
124128hCaptcha browser check. Inference still goes only to the gateway or cloud
125provider those settings name.
129provider the add-in connects to.
126130
127131## Deploy the add-in for your organization
128132
from line 294
290294`/claude-for-msft-365-install:access-policies` commands document both
291295keys in full.
292296
297### Skill limits for a bootstrap endpoint
298
299A bootstrap endpoint can return a `skills` array of admin-provisioned
300skills. The add-in applies these limits to each response.
301
302| Limit | Value |
303| - | - |
304| Skills kept | The first 20 skill names, in array order. No setting changes this number. Uploaded and built-in skills don't count toward it. |
305| Repeated names | One skill for each name: the last entry with that name, in the position of the first. |
306| Skill `url` | HTTPS, no redirects, CORS for `https://pivot.claude.ai`, and 15 seconds for each download. |
307| Bootstrap call | 10 seconds in total to get the user's Microsoft Entra token, when the add-in uses one, and to receive the response. This limit applies to each bootstrap call, except the call made when the add-in sets up a new connection, for example at first use or after a sign-out. |
308
309If a skill fails to load, the add-in doesn't replace it with a later
310skill. To see which skills failed to load, select the + button at the
311bottom of the chat pane, and then select Skills. A note at the top of the
312list names up to three of those skills and gives the number of any others.
313
314If a bootstrap call takes more than 10 seconds when the add-in
315loads with an existing connection, the add-in has no admin-provisioned
316skills until a later bootstrap call succeeds.
317
318To stay within these limits:
319
320* Put the most important skills first.
321* Return a different list for each app. The `X-Claude-User-Agent` header
322 on each bootstrap call starts with `claude-word`, `claude-excel`,
323 `claude-powerpoint`, or `claude-outlook`.
324* Use `url` instead of inline `content` for large skills. Inline content
325 is part of the bootstrap response, so it counts toward the 10 seconds.
326 The add-in downloads skill URLs after the bootstrap call, and only for
327 the skills that it keeps.
328
329If a response holds more than 20 different skill names, the add-in sends
330the warning log record `Bootstrap skill count exceeds cap — truncating`
331to your [OpenTelemetry collector](https://claude.com/docs/office-agents/opentelemetry.md) at
332`{your_url}/v1/logs`. Your collector might not receive this record for
333the bootstrap call made when the add-in sets up a new connection. On that
334record, `count` is the number of different skill names in the response,
335and `cap` is the limit, 20.
336
293337### Deploy to Outlook
294338
295339Outlook requires a separate manifest file from Excel, PowerPoint, and
from line 406
362406Claude organization directly, in place of manifest parameters, Microsoft
363407Entra ID attributes, or a bootstrap endpoint. Members sign in with the
364408add-in's standard "Log in" button and their Claude account. The add-in
365then reads the organization's settings from Anthropic and connects to
366the gateway or cloud provider those settings name. Prompts and responses
367still travel only to that provider, never to Anthropic.
409then reads the organization's settings from Anthropic and fills in the
410connection screen with them. Prompts and responses still travel only to
411the gateway or cloud provider the add-in connects to. What reaches
412Anthropic then depends on the provider, as
413[What Anthropic collects](#what-anthropic-collects) describes.
368414
369415This option is in preview. It works in Anthropic's preview environments
370416and is not yet enabled for production organizations. Members need the
from line 432
386432 verify your organization's sign-in."
3874333. After the reload, the add-in reads the organization's settings from
388434 `api.anthropic.com` and opens the connection screen with the served
389 values filled in, such as the gateway URL, API format, authorization
390 header, and available models. When the served settings include every
391 value the connection needs, the add-in connects without further
392 input. Otherwise the member enters the missing value, typically the
393 gateway token from your IT team, and connects.
435 values filled in, such as the gateway URL. When the served settings
436 include every value the connection needs, the add-in connects without
437 further input. Otherwise the member enters the missing value,
438 typically the gateway token from your IT team, and connects.
3944394. While the member stays signed in, the add-in reads the served
395440 settings again at each launch and periodically while it runs, so
396441 changes an admin makes apply without redeploying the manifest.
from line 448
403448[Admin feature controls](#admin-feature-controls). A few rules are
404449specific to this path:
405450
406* **Single source**: for a member signed in this way, the served
407 document is the only configuration source. The add-in does not merge
408 it with manifest parameters, Entra ID attributes, or a bootstrap
409 endpoint, and nothing from the task pane URL fills a key the served
410 document leaves out.
451* **Connection values**: for a member signed in this way, the add-in
452 fills in the connection screen from the served settings, not from
453 manifest parameters, Entra ID attributes, or a bootstrap endpoint.
454* **Other connection paths**: on the connection screen, the member can
455 select another [connection path](#connection-paths) and enter their
456 own values. The add-in then sends prompts to the provider the member
457 chose.
458* **Supported providers**: served settings can name an LLM gateway,
459 Bedrock, Vertex AI, or Foundry as the inference provider.
460* **Unsupported settings**: if served settings name Bedrock Mantle or
461 set a Bedrock base URL, the add-in makes no new connection. A member
462 who is already connected keeps that connection until they sign out. A
463 member who isn't connected yet sees the connection screen without its
464 fields, with Try again and Sign out actions. After you change the
465 served settings, the member selects Try again to load them. Until
466 then, the screen shows one of these messages:
467 * Bedrock Mantle: "This add-in can't use your organization's inference
468 provider. Let your IT team know."
469 * Bedrock base URL: "This add-in can't use your organization's Bedrock
470 base URL. Let your IT team know."
411471* **Applied as delivered**: the add-in applies served settings the same
412472 way it applies manifest configuration, with no per-setting consent
413473 prompt. The Claude organization admin who edits served settings can
from line 949
889949add-in experience. They do not transmit your prompts or Claude's
890950responses.
891951
892Anthropic collects information in accordance with Amazon Bedrock, Google
893Cloud Vertex AI, or Microsoft Azure's terms, consistent with Anthropic's
894arrangements with customers. Anthropic does not have access to a
895customer's AWS, Google, or Microsoft instance, including prompts or
896outputs it contains. Anthropic does not train generative models with
897such content or use it for other purposes. Anthropic can access
898metadata such as tool use and token counts, and uses such metadata for
899analytic and product-improvement purposes.
952When the add-in uses Amazon Bedrock or Google Cloud Vertex AI, directly
953or through a gateway, Anthropic collects information in accordance with
954that platform's terms, consistent with Anthropic's arrangements with
955customers. Anthropic does not have access to a customer's AWS or Google
956instance, including prompts or outputs it contains. Anthropic does not
957train generative models with such content or use it for other purposes.
958Anthropic can access metadata such as tool use and token counts, and
959uses such metadata for analytic and product-improvement purposes.
900960
961When the add-in uses Azure AI Foundry, directly or through a gateway,
962Anthropic operates the Claude models as an independent processor for
963Microsoft, and Anthropic's data use terms apply. For deployments hosted
964on Azure, prompts and responses stay within Azure, and only usage
965metadata and content flagged by Anthropic's safety systems reach
966Anthropic. For deployments hosted on Anthropic, prompts and responses are
967processed on Anthropic's infrastructure. See
968[Claude in Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry)
969for details.
970
901971For details on what your organization's gateway or cloud provider logs,
902972contact your IT team.
903973
from line 1115
10451115third-party platforms yet. Support is being added. A member who signs in
10461116with a Claude account to an organization whose settings Anthropic
10471117serves is in the third-party platform column too, because inference
1048goes to the organization's provider.
1118goes to a gateway or cloud provider.
10491119
10501120| Feature | Claude account | Third-party platform |
10511121| - | - | - |
office-agents/word Changed · +8 / -3 lines
from line 302
302302Data is cached for a number of hours after deletion so users can access
303303context in recently closed documents.
304304
305Chat history is stored locally in your browser using IndexedDB.
306Conversations are not stored on Anthropic's servers, are not synced
307across devices, and can be cleared from Settings at any time.
305Chat history is stored locally in your browser using IndexedDB. It is
306not synced across devices and can be cleared from Settings at any time.
308307Reinstalling the add-in or switching between Claude add-ins does not
309308remove it. See [Data storage and retention](https://claude.com/docs/office-agents/data-storage.md)
310309for where it sits on disk and how long it is kept.
310
311When you sign in with a Claude account, the add-in sends your
312conversations to Anthropic, including the content Claude reads from your
313document and any files you attach. Clearing chat history removes only
314the copy in your browser. It does not delete the messages or attached
315files that Anthropic keeps.
311316
312317Claude for Word does not inherit custom data retention settings your
313318organization might have set. Activity is not included in Enterprise
office-agents/work-across-apps Changed · +8 / -3 lines
from line 124
124124the [Compliance API](https://platform.claude.com/docs/en/manage-claude/compliance-api)
125125enabled, add-in sessions are included in the Compliance API. This
126126coverage is in public beta and requires no additional setup: the same
127Compliance Access Keys apply. Chat history is stored locally in your
128browser, not on Anthropic's servers, and can be cleared from Settings at
129any time.
127Compliance Access Keys apply.
128
129Chat history is stored locally in your browser and can be cleared from
130Settings at any time. The add-ins also send your conversations to
131Anthropic, including the content Claude reads from your open workbooks,
132documents, presentations, email, and calendar, and any files you attach.
133Clearing chat history removes only the copy in your browser. It does not
134delete the messages or attached files that Anthropic keeps.
130135
131136## Current limitations
132137
third-party/claude-desktop/extensions Changed · +59 / -0 lines
## Skill scanning ### Turn on skill scanning ### How a scan works ### What skill scanning doesn't cover
from line 470
470470
471471`userPluginMarketplacesEnabled` doesn't govern Claude Code in a terminal. To limit the marketplaces Claude Code can use in a terminal, set [marketplace restrictions in its managed settings](https://code.claude.com/docs/en/plugins/org#restrict-what-users-can-install). If you deploy Claude Code managed settings to the device, restrict marketplaces there too, because in Code sessions those settings can take precedence over the app's restrictions.
472472
473## Skill scanning
474
475> [!NOTE]
476>
477> Skill scanning is in beta. It's off until an administrator turns it on.
478
479Skill scanning checks the [skills](https://claude.com/docs/skills/overview.md) that users create or upload in the app. It looks for signs that a skill is malicious, such as hidden code execution, credential exposure, or data exfiltration. With scanning on, a scan starts each time a user creates, uploads, or edits a skill in the app. The user can't use the skill until the scan finishes, and the app blocks a skill that fails the scan.
480
481Claude Desktop doesn't scan plugins, or the skills inside them. To stop users from adding their own plugins, see [Controlling user extensions](#controlling-user-extensions).
482
483### Turn on skill scanning
484
485In the Enterprise Admin Console:
486
4871. In **Organization settings**, go to the **Capabilities** page under **Desktop 3P**.
4882. Under **General restrictions**, turn on the **Scan user-created skills** switch.
4893. Click **Save changes**.
490
491The console shows the **Scan user-created skills** switch only while **Allow user-created skills** is on.
492
493A [group's permission policy](https://claude.com/docs/third-party/claude-desktop/admin-console.md#per-group-permission-policies) has the same switch, and its setting overrides the organization's setting for that group.
494
495You can also turn on scanning in these ways:
496
497* Set [`skillScanningEnabled`](https://claude.com/docs/third-party/claude-desktop/configuration#skillscanningenabled) to `true` in the [MDM profile](https://claude.com/docs/third-party/claude-desktop/mdm.md) or the [bootstrap response](https://claude.com/docs/third-party/claude-desktop/bootstrap.md). A device that the console or a bootstrap server manages [ignores](https://claude.com/docs/third-party/claude-desktop/admin-console.md#which-settings-apply-on-a-device-with-a-managed-profile) the MDM profile's value for this setting.
498* Turn on the same switch under **Capabilities** in the [in-app configuration window](https://claude.com/docs/third-party/claude-desktop/in-app-configuration.md), then click **Apply Changes**. The change takes effect after the app relaunches.
499
500Changes from the console, a bootstrap server, or MDM take effect at the app's next [configuration check](https://claude.com/docs/third-party/claude-desktop/admin-console.md#configuration-updates), and users don't have to relaunch the app.
501
502### How a scan works
503
504The app sends the text of the skill's files to a Claude model on your organization's own inference provider, using the same credential that the user's sessions use.
505
506How well a scan detects a malicious skill depends on the model that runs it. A scan runs on one model from your organization's [model list](https://claude.com/docs/third-party/claude-desktop/models.md#model-list-and-default-model). Claude Desktop prefers an Opus model, then a Sonnet model, then a Haiku model, but only among the models it can recognize. It recognizes a model when the model ID contains "opus", "sonnet", or "haiku", or when you set `anthropicFamilyTier` on that model in [`inferenceModels`](https://claude.com/docs/third-party/claude-desktop/configuration#inferencemodels). If your model IDs are a gateway's route names, Claude Desktop may not recognize them, so set `anthropicFamilyTier` on each model. If Claude Desktop recognizes none of your models, it still runs the scan on another model from the list.
507
508Your organization pays its inference provider for scans. A scan's tokens count toward the user's [token limit](https://claude.com/docs/third-party/claude-desktop/configuration.md#inferencemaxtokensperwindow).
509
510A scan ends in one of these results:
511
512* **Pass**: the user can use the skill.
513* **Caution**: the scan found something that could be unsafe but couldn't confirm that the skill is unsafe. The user can use the skill.
514* **Blocked**: the skill failed the scan, and the user can't use it. The user can correct the skill and upload it again, which starts a new scan.
515* **Not reviewed**: the scan didn't produce a result, and the user can't use the skill. This happens, for example, when the provider returns an error or the user is over their token limit. Once the cause is fixed, the user can click **Retry** on the skill's page.
516
517Neither a user nor an administrator can approve a blocked skill. Turning scanning off doesn't unblock it. But while scanning is off, a user who edits a **Blocked** or **Not reviewed** skill, or uploads it again, can use it, because the app no longer scans it.
518
519The Enterprise Admin Console doesn't show scan results. To collect them, set [`otlpEndpoint`](https://claude.com/docs/third-party/claude-desktop/configuration#otlpendpoint). Claude Desktop then sends scan records to your own [OpenTelemetry collector](https://claude.com/docs/third-party/claude-desktop/telemetry.md#sending-telemetry-to-your-own-collector) as log records whose body is `custom3p_skill_scan_completed`. It sends no scan records while [`otlpDesktopLogLevel`](https://claude.com/docs/third-party/claude-desktop/configuration#otlpdesktoploglevel) is `off`. A scan record identifies the user and includes the skill's name, the result, and the model's summary of the skill when there is one. The summary can quote the skill, and Claude Desktop sends it even when [content capture](https://claude.com/docs/third-party/claude-desktop/telemetry.md#content-capture) is off. To keep the summary out of your backend, drop the `scan_summary` attribute in the collector.
520
521### What skill scanning doesn't cover
522
523Besides plugins, the app doesn't scan these:
524
525* Skills that users saved before you turned on scanning. The app scans such a skill the next time the user saves it.
526* Changes that someone makes directly to a skill's files on disk.
527* Files inside a skill that the scan skips because of their file names, for example PNG and JPEG images, PDFs, and dependency lock files such as `package-lock.json`. The scan doesn't look at what is in these files, so it doesn't read a script that is saved as `logo.png`.
528* Skills that aren't saved in the app at all, for example [Claude Code skills](https://code.claude.com/docs/en/skills) in `.claude/skills` folders on the device. A [Code session](https://claude.com/docs/third-party/claude-desktop/code.md) in Claude Desktop loads those skills too.
529
530A pass means that the scan found no sign that the skill is malicious. It isn't a guarantee that the skill is safe.
531
473532## Related topics
474533
475534> **[Code](https://claude.com/docs/third-party/claude-desktop/code.md)**
third-party/claude-desktop/admin-console Changed · +1 / -1 lines
from line 93
9393| Page | What you configure there |
9494| - | - |
9595| **Connection** | The inference provider ([gateway](https://claude.com/docs/third-party/claude-desktop/gateway.md), [Amazon Bedrock](https://claude.com/docs/third-party/claude-desktop/bedrock.md), [Bedrock Mantle](https://claude.com/docs/third-party/claude-desktop/mantle.md), [Google Cloud's Agent Platform](https://claude.com/docs/third-party/claude-desktop/vertex.md), or [Microsoft Foundry](https://claude.com/docs/third-party/claude-desktop/foundry.md)), its endpoint, region, or project, how users authenticate to it, custom request headers, and, under **Models**, the model list, default model, model discovery, and cost-estimate rates. **Desktop sign-in** on this page holds the **Require this organization in Claude Desktop** switch described under [Users in more than one Claude organization](#users-in-more-than-one-claude-organization). |
96| **Capabilities** | Whether Chat, Cowork, and Code are each available, [Unified Claude](#turn-on-unified-claude), the folders and network hosts the app may use, permission modes and built-in tool policy, the [built-in browser](https://claude.com/docs/third-party/claude-desktop/browser.md#manage-the-built-in-browser-from-the-enterprise-admin-console) and its site permissions, whether users may add their own skills and plugins, organization instructions, and [**Claude.ai data import**](https://claude.com/docs/third-party/claude-desktop/import.md) |
96| **Capabilities** | Whether Chat, Cowork, and Code are each available, [Unified Claude](#turn-on-unified-claude), the folders and network hosts the app may use, permission modes and built-in tool policy, the [built-in browser](https://claude.com/docs/third-party/claude-desktop/browser.md#manage-the-built-in-browser-from-the-enterprise-admin-console) and its site permissions, whether users may add their own skills and plugins, [skill scanning](https://claude.com/docs/third-party/claude-desktop/extensions.md#skill-scanning), organization instructions, and [**Claude.ai data import**](https://claude.com/docs/third-party/claude-desktop/import.md) |
9797| **Connectors** | Managed MCP servers, including the [built-in connectors](https://claude.com/docs/third-party/claude-desktop/built-in-connectors.md), whether users may add their own MCP servers, and desktop extension policy |
9898| **Telemetry & updates** | Which telemetry categories go to Anthropic, whether users' apps report [usage analytics](#usage-analytics) to your organization, OpenTelemetry export to your collector, update policy, the [configuration relaunch window](#configuration-updates), the configuration re-check interval, and [how much of the configuration devices keep on disk](#configuration-kept-on-devices) |
9999| **Limits** | A per-user token limit and its window |
third-party/claude-desktop/configuration Changed · +1 / -0 lines
This page is larger than the 256 KiB this site keeps, so one side of the diff below stops where the stored text does.
Nothing in the body moved in this read. What changed is above.
third-party/claude-desktop/feature-matrix Changed · +1 / -0 lines
from line 63
6363> * [Admin console](https://claude.com/docs/third-party/claude-desktop/admin-console.md)
6464> * [Endpoint or gateway configuration](https://claude.com/docs/third-party/claude-desktop/configuration.md#connection)
6565> * [Skills, hooks, and plugins distribution](https://claude.com/docs/third-party/claude-desktop/extensions.md#organization-plugins-admin)
66> * [Skill scanning](https://claude.com/docs/third-party/claude-desktop/extensions.md#skill-scanning)
6667> * [MCP server allowlist](https://claude.com/docs/third-party/claude-desktop/extensions.md#controlling-user-extensions)
6768> * [Feature toggles](https://claude.com/docs/third-party/claude-desktop/configuration.md#capabilities)
6869> * [Auto-updates](https://claude.com/docs/third-party/claude-desktop/installation.md#updates)
third-party/claude-desktop/overview Changed · +1 / -1 lines
from line 36
3636
3737* **Conversation content goes only to your configured endpoint.** The app sends prompts, responses, files, and tool outputs only to your configured inference endpoint and stores them only on the local machine. If you enable [SSH remote sessions](https://claude.com/docs/third-party/claude-desktop/ssh-remote-sessions.md), Claude Code runs those sessions on a remote host, which also receives that content, sends it to your endpoint, and stores it. What happens to that content at the endpoint depends on the provider, as described under [Data handling by provider](#data-handling-by-provider).
3838* **Sandboxed tool execution.** In Cowork, shell commands run in the hardened Cowork VM; file access is scoped to your allowed folders and web fetches to your egress allowlist. For Code sessions, see [Applied as managed policy](https://claude.com/docs/third-party/claude-desktop/code.md#applied-as-managed-policy).
39* **Auditable telemetry.** Crash reports and product analytics are scrubbed of conversation and user data before being sent to Anthropic, and can be fully disabled via configuration keys. Independently, you can export session activity to your own OpenTelemetry collector. The export is metadata only by default, with prompt and tool content available as an explicit opt-in.
39* **Auditable telemetry.** Crash reports and product analytics are scrubbed of conversation and user data before being sent to Anthropic, and can be fully disabled via configuration keys. Independently, you can export session activity to your own OpenTelemetry collector. The export is metadata only by default, with prompt and tool content available as an explicit opt-in. If you turn on [skill scanning](https://claude.com/docs/third-party/claude-desktop/extensions.md#skill-scanning), the export also carries records of scan results. A record's summary can quote the skill.
4040* **Centrally managed.** Configuration is delivered through your existing MDM (Jamf, Intune, Workspace ONE, Group Policy), a [bootstrap server](https://claude.com/docs/third-party/claude-desktop/bootstrap.md), or the [admin console](https://claude.com/docs/third-party/claude-desktop/admin-console.md). End users can't edit a configuration that MDM delivers. They can leave it for standard Claude Desktop by signing in to claude.ai. On Claude Desktop 2.37337.0 or later, a user who signs in to any organization that uses the admin console can move the device to that organization's configuration. To prevent both, set [`disableDeploymentModeChooser`](https://claude.com/docs/third-party/claude-desktop/configuration#disabledeploymentmodechooser) to `true` in the MDM profile, as described under [Keep a device on its managed profile](https://claude.com/docs/third-party/claude-desktop/admin-console.md#keep-a-device-on-its-managed-profile).
4141
4242For a detailed treatment of the threat model, sandbox boundaries, and data flows, request access to the [Claude Cowork Desktop Security Architecture Overview](https://trust.anthropic.com/resources?s=2a7bbzo1lyymvdt551q7kl\&name=claude-cowork-desktop-security-architecture-overview) on Anthropic's Trust Center. For architecture, telemetry, and controls information specific to Claude Desktop on 3P, see the [Claude Desktop Security Overview (Third-party platforms)](https://trust.anthropic.com/resources?s=0c8rx4s7mm5ierz8ppetfs\&name=claude-cowork-security-overview-\(third-party-platforms\)) on the Trust Center.
third-party/claude-desktop/telemetry Changed · +1 / -1 lines
from line 58
5858
5959Independently of what's sent to Anthropic, you can export session activity to your own OpenTelemetry collector by setting `otlpEndpoint`. This is the recommended way to retain an audit trail in environments that disable Anthropic-bound telemetry.
6060
61For third-party deployments, the export includes session metadata (event names, durations, token counts, result counts, errors) by default, but not message content. It also identifies the signed-in user; see [User attribution](#user-attribution). See [Monitoring](https://claude.com/docs/cowork/monitoring.md) for the event schema and the [`otlp*` keys](https://claude.com/docs/third-party/claude-desktop/configuration#otlpendpoint) in the configuration reference.
61For third-party deployments, the export includes session metadata (event names, durations, token counts, result counts, errors) by default, but not message content. It also identifies the signed-in user; see [User attribution](#user-attribution). Records of skill scans are an exception to leaving out message content. If you turn on [skill scanning](https://claude.com/docs/third-party/claude-desktop/extensions.md#skill-scanning), the export also carries `custom3p_skill_scan_completed` records of scan results, and a record's summary can quote the skill. Claude Desktop sends the summary even when [content capture](#content-capture) (`otlpContentCapture`) is off. See [Monitoring](https://claude.com/docs/cowork/monitoring.md) for the event schema and the [`otlp*` keys](https://claude.com/docs/third-party/claude-desktop/configuration#otlpendpoint) in the configuration reference.
6262
6363The export carries logs and metrics. Cowork sessions, Code sessions, and the desktop application's own events arrive under the `service.name` values `cowork`, `claude-code-desktop`, and `claude-desktop` respectively. With [Unified Claude](https://claude.com/docs/third-party/claude-desktop/unified-claude.md) on, a Cowork session started from the home screen also arrives under `cowork`. The export isn't checked against the [agent egress allowlist](#required-egress-paths), so `otlpEndpoint` doesn't need an entry in `coworkEgressAllowedHosts`; your perimeter firewall still needs to allow the host.
6464