claude-docs-20260910T230716Z / documentation capture

What this read moved

1–12 of 12

claude-tag/admins/add-connections Changed · +8 / -6 lines

from line 121

121121* **The entry matches the exact host.** A wildcard like `*.example.com` doesn't cover `example.com` itself, and `www.example.com` and `example.com` are different hosts.

122122* **The request didn't move to another host.** If the page redirects, or loads from a CDN or a sign-in host, allow that host too; Claude names the host it was blocked on.

123123* **The port is listed.** Needed only when the service listens on something other than 443.

124* **A minute has passed since you saved.** Agent Proxy picks up a new entry within about a minute, in existing threads as well as new ones, so retry in the same thread after a short wait.

124* **A minute has passed since you saved the entry.** Agent Proxy picks up a new entry within about a minute, in existing threads as well as new ones, so retry in the same thread after a short wait.

125* **The bundle was attached before the thread started.** A bundle you attach after a thread started isn't guaranteed to reach that thread, so start a fresh thread to use its entries.

125126* **The request came from a channel, not a DM.** A bundle attached to a channel doesn't apply in DMs.

126127

127128Typical entries are hosts the work calls without a key, such as a docs site or a public API. Common package registries are usually already reachable through the [environment's Trusted access default](#broad-web-access-through-the-environment), and a host that needs a credential belongs in a [connection](#add-a-connection) instead. Entries appear below the form, and each one can be edited or removed from its row.

from line 199

198199

199200After saving, you can narrow a connection. Select **Edit** on the connection's row in the bundle's **Credentials** tab. The **Edit connection** dialog lets you rename the connection and, where the connection has an allow rule, restrict it by HTTP method and path, for example to allow `GET` but not `DELETE`.

200201

201Agent Proxy starts applying a change within about a minute of your saving it, in existing threads as well as new ones. It evaluates connections and Domains entries from the most specific scope outward (channel, then workspace, then organization), and within a scope by priority; the first match decides. A request that matches no connection, no Domains entry, and nothing in the environment's network access is blocked. Private IP ranges and cloud metadata endpoints stay blocked regardless.

202Agent Proxy starts applying an edit to a connection or a Domains entry within about a minute after you save it, in existing threads as well as new ones. It evaluates connections and Domains entries from the most specific scope outward (channel, then workspace, then organization), and within a scope by priority; the first match decides. A request that matches no connection, no Domains entry, and nothing in the environment's network access is blocked. Private IP ranges and cloud metadata endpoints stay blocked regardless.

202203

203204### Connections vs claude.ai connectors

204205

from line 209

208209

209210A connection grants access; a plugin teaches Claude how to use it well. A plugin is a packaged set of skills: reusable instructions for working with a specific tool or following a specific process. Attach a plugin to the same Access bundle or scope that carries the connection, so the credential arrives with directions for using it.

210211

211A Datadog API key, for example, makes the API reachable, and a Datadog plugin tells Claude which endpoints answer which questions. Sessions in covered channels pick up attached plugins automatically, with nothing to install or enable. Channel members can also add plugins available to your organization from the channel's [Configure page](/docs/claude-tag/users/good-habits#configure-claude-for-a-channel), unless an admin has [restricted editing to admins](/docs/claude-tag/admins/attach-to-scope#restrict-who-can-set-channel-instructions).

212A Datadog API key, for example, makes the API reachable, and a Datadog plugin tells Claude which endpoints answer which questions. Once you turn a plugin on for a bundle or add it to a scope, sessions in the channels that bundle or scope covers pick up the plugin automatically. Nobody in those channels has to turn that plugin on. A channel member can also add a plugin available to your organization, by asking Claude in the channel or from the channel's [Configure page](/docs/claude-tag/users/good-habits#configure-claude-for-a-channel), unless an admin has [restricted editing to admins](/docs/claude-tag/admins/attach-to-scope#restrict-who-can-set-channel-instructions).

212213

213214Anthropic provides plugins for common tools and processes, and you can add your own from a [skills repository](/docs/claude-tag/admins/skills-repo). To give Claude organization-wide skills, package them as a plugin.

214215

215Plugins attach in two places, and the two behave differently:

216Admins and channel members turn plugins on in different places:

216217

217218* A plugin added directly on a scope (the plugin chips on the scope's panel) is enabled there as soon as you add it.

218219* A bundle's **Plugins** tab lists the plugins available to your organization, each off until you toggle it on.

220* A channel member can ask Claude to add a plugin to their channel, unless an admin has [restricted editing to admins](/docs/claude-tag/admins/attach-to-scope#restrict-who-can-set-channel-instructions). Claude proposes the change and adds the plugin only after someone in that channel selects **Confirm**.

219221

220Registering a plugin at the organization level makes it available, not active. It takes effect only where a bundle enables it or a scope adds it directly.

222Adding a plugin for your whole organization makes it available, not active. The plugin takes effect only where a bundle enables it or a scope adds it directly.

221223

222224Adding or removing plugins and skills applies to new threads only. A thread already running keeps the set it began with; start a fresh thread to pick up changes. See [What survives between replies](/docs/claude-tag/concepts/how-it-works#what-survives-between-replies).

223225

claude-tag/admins/troubleshooting Changed · +14 / -6 lines

from line 630

630630

631631## Session start errors

632632

633Session start errors appear before any work begins. Some are transient and clear on retry; the rest point to capacity or environment configuration rather than credentials.

633Session start errors appear before any work begins. Some are transient and clear on retry; the rest point to capacity or environment configuration.

634634

635635### Still waiting for available capacity

636636

from line 642

642642

643643**What it means**

644644

645Compute capacity is temporarily busy. The session starts on its own once capacity frees up.

645The session was created and is still waiting for compute to run it. Why it waits differs between Anthropic-hosted and self-hosted [environments](/docs/claude-tag/admins/customize#configure-the-environment-for-a-scope).

646646

647On an Anthropic-hosted environment, capacity is temporarily busy or the session is taking longer than usual to start, and the session normally starts on its own. If it never starts, Claude posts the [Session failed to start](#session-failed-to-start-the-session-container-never-connected) message instead.

648

649On a [self-hosted environment](https://code.claude.com/docs/en/self-hosted-environments), no runner has claimed the session yet. The session waits in the environment's queue until a runner claims it.

650

647651**How to resolve**

648652

649Wait; no action is needed. Users should reply in the same thread if they have more to add, since starting a new thread only queues a second session behind the first.

653On an Anthropic-hosted environment, wait; no action is needed.

654

655On a self-hosted environment, use [Troubleshooting](https://code.claude.com/docs/en/self-hosted-environments-deploy#troubleshooting) in the self-hosted environments guide to find out why no runner is claiming the session. The queued session starts once a runner claims it.

656

657Either way, users should reply in the same thread if they have more to add, since starting a new thread only queues a second session behind the first.

650658

651659### Session failed to start: the session container never connected

652660

claude-tag/concepts/security-and-data Changed · +8 / -8 lines

from line 1

11# Security and data handling

3> Claude Tag runs in an isolated sandbox that holds no credentials. Covers sandbox isolation, credential storage, network egress, service accounts, isolating credentials between channels and what one channel can reach, who can open a published artifact, and which members can invoke Claude.

3> How Claude Tag keeps credentials out of the sandbox, limits where a channel session's requests can go, and controls who can see artifacts and invoke Claude.

55export const BetaNote = () => <Info>Claude Tag is in public beta. Features and behavior described here may change before general availability.</Info>;

from line 16

1616

1717Each Slack thread runs in its own sandbox. In an Anthropic-hosted environment, every outbound call from that sandbox passes through the same checkpoints.

1818

19<img className="block dark:hidden" src="https://mintcdn.com/claude-ai/5JFKyLlO7sHMMf5J/images/claude-tag/diagrams/request-path.svg?fit=max&auto=format&n=5JFKyLlO7sHMMf5J&q=85&s=e8776cf0edd1f3ec912b9b044c9cc838" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request matching no rule is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path.svg" />

19<img className="block dark:hidden" src="https://mintcdn.com/claude-ai/oY6LusJt4c576Dc3/images/claude-tag/diagrams/request-path.svg?fit=max&auto=format&n=oY6LusJt4c576Dc3&q=85&s=a7f2e0f4303072b4c8e4f3197bf0cb12" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request that no rule, domain entry, or environment network access setting allows is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path.svg" />

2020

21<img className="hidden dark:block" src="https://mintcdn.com/claude-ai/5JFKyLlO7sHMMf5J/images/claude-tag/diagrams/request-path-dark.svg?fit=max&auto=format&n=5JFKyLlO7sHMMf5J&q=85&s=a2ed5f3270989c3ed3d9b9a78a72bff0" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request matching no rule is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path-dark.svg" />

21<img className="hidden dark:block" src="https://mintcdn.com/claude-ai/oY6LusJt4c576Dc3/images/claude-tag/diagrams/request-path-dark.svg?fit=max&auto=format&n=oY6LusJt4c576Dc3&q=85&s=a728805c81b642004e1e15da96967e4c" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request that no rule, domain entry, or environment network access setting allows is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path-dark.svg" />

2222

23| Checkpoint | The guarantee |

24| :----------- | :------------------------------------------------------------------------------------------------------------- |

25| The sandbox | Holds no credentials |

26| Agent Proxy | Injects credentials from the credential store at request time, and blocks traffic to unlisted hosts by default |

27| Your systems | See the agent's own accounts, so its actions there are attributable |

23| Checkpoint | The guarantee |

24| :----------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

25| The sandbox | Holds no credentials |

26| Agent Proxy | Injects credentials from the credential store at request time, and blocks a request that no [connection](/docs/claude-tag/admins/add-connections#set-allowed-websites), [**Domains** entry](/docs/claude-tag/admins/add-connections#allow-a-host-without-a-credential), or [environment network access level](/docs/claude-tag/admins/add-connections#broad-web-access-through-the-environment) allows |

27| Your systems | See the agent's own accounts, so its actions there are attributable |

2828

2929### Compute and the sandbox

3030

claude-tag/users/proactivity Changed · +3 / -2 lines

from line 40

4040

4141## Routine recipes

4242

43Each recipe below sets up a complete routine with one message. Adapt the channel names, repositories, and times to your own, and name the timezone so the schedule fires when you expect.

43Each recipe below sets up a complete routine with one message. Adapt the channel names, repositories, and times to your own, and name the timezone in each message, since schedules run in UTC.

4444

4545### Daily standup summary

4646

from line 107

107107* A job runs with the channel's connections, the same as an interactive request.

108108* Claude can post a job's output into another public channel in the same workspace only if the job's own channel is public and Claude has been added to the target channel. It labels the message with the channel it came from.

109109* Claude doesn't post job output to private channels, DMs, group DMs, or externally shared channels, and doesn't message people directly. The one exception is the completion or failure notice it sends to whoever set up the routine, and only when that person's Slack account is connected to their Claude account.

110* Schedules default to UTC. When you say "every weekday at 9am," include the timezone (for example "9am Pacific") so Claude converts correctly; without one it may guess. Ask "what triggers do you have set up?" to confirm the time it actually scheduled.

110* Schedules run in UTC. Name the timezone when you set a schedule, as in "every weekday at 9am Pacific". With no timezone in your message, Claude uses the one on your Slack profile. To confirm the time Claude set, send [`@Claude !routines`](/docs/claude-tag/users/commands#list-the-routines-in-a-channel), which lists schedules in UTC.

111* A routine runs at a fixed UTC time, so each daylight saving change shifts its local time by an hour, in the same direction the clocks move. A routine running at 9am Pacific moves to 10am after the clocks go forward, or to 8am after they go back. Ask Claude to reschedule the routine to the local time you want.

111112* A scheduled job that touches a github.com repository uses the same GitHub connection your admin set up for interactive work. See [Configure GitHub access](/docs/claude-tag/admins/configure-github#scheduled-work-uses-the-same-connection).

112113

113114## Related resources

claude-tag/users/troubleshooting Changed · +8 / -6 lines

from line 81

8181

8282**What it means**

8383

84The first message means compute capacity is temporarily busy; the session starts on its own once capacity frees up. The second means the session didn't start at all, which is transient.

84The first message means the session was created and is waiting for capacity to run it. The session normally starts on its own within a few minutes. If your organization runs Claude's sessions on its own infrastructure, the wait lasts until that infrastructure starts the session.

8585

86The second message means the session didn't start at all. The failure is usually temporary.

8688**How to resolve**

8789

88* For the capacity message, wait a few minutes; no action is needed. To add context while waiting, @-mention Claude in the same thread rather than starting a new one. A new thread only queues a second session behind the first.

89* For the failed-start message, mention Claude in the same thread to retry. If the retry worked, the session starts and Claude begins the task.

90* For the capacity message, wait a few minutes. To add context while waiting, @-mention Claude in the same thread rather than starting a new one. A new thread only queues a second session behind the first. If Claude still hasn't started after several more minutes, send your admin [Still waiting for available capacity](/docs/claude-tag/admins/troubleshooting#still-waiting-for-available-capacity).

91* For the failed-start message, mention Claude in the same thread to retry. If the session fails to start again, ask an admin.

9092

9193### Claude didn't react to a message I edited

9294

cowork/changelog Changed · +39 / -0 lines

from line 2

33> Release notes for Claude Desktop

5<Update label="v1.52386.0" description="2026-09-10">

6 **General**

8 * Changed messages sent while Claude is still replying: they now wait in the conversation looking like sent messages, the first waiting message offers Send now, which stops the current reply and sends it next, and Cmd/Ctrl+Enter sends your message right away, interrupting the current turn.

9 * Fixed Claude's clicks, scrolls, screenshots, and page scripts in the built-in browser stalling, timing out, or failing while the browser pane was hidden or the window was minimized or in the background, and fixed elements picked with "Select element" reaching Claude without their styles or with a screenshot of the wrong part of the page.

10 * Fixed links that open a new tab to claude.ai or claude.com pages, like Upgrade buttons and some download links, doing nothing when clicked; they now open in your default browser.

11 * Fixed sites on a private network (VPN, Tailscale, or intranet hosts) loading without their styles, scripts, or data in the built-in browser, and URLs typed into the built-in browser in cloud sessions not reaching private-network sites.

12 * Fixed the app window resetting to its default size and position after an update or a display change.

14 **Code**

16 * Added a default transcript view setting: choose whether new sessions open in the Normal, Thinking, or Verbose view from Settings > Claude Code, or make the current view the default from a session's Transcript view menu. The choice syncs between the desktop app and claude.ai.

17 * Fixed Code sessions failing to start on some Microsoft Store and MSIX installs on Windows.

18 * Fixed model changes being refused with a message saying organization-managed hooks could not be checked, for some organizations that manage Claude Code plugins; the session now restarts on the chosen model.

19 * Fixed sessions failing to start on Windows for accounts with many plugins installed.

20 * Fixed several SSH session issues: hosts slowly accumulating leftover background processes from older Claude versions until sessions there failed, sessions from one computer being ended when a second computer's Claude app cleaned up the same host, sessions permanently losing their connection after a saved connection's host or port was edited, and history missing its newest messages soon after launch or a reconnect.

21 * Fixed WSL sessions failing to start or reconnect, including background reconnects right after an app update, when WSL reported a momentary error.

23 **Cowork**

25 * Changed what happens on Windows PCs where a Windows update released September 8, 2026 prevents Claude's workspace from reaching your files: the app now names that cause instead of deleting and reinstalling the workspace. This does not fix the underlying problem, and the workspace still can't reach your files on those PCs.

26 * Fixed a session appearing stuck running when an organization hook blocked a message; the reason the message was blocked is now shown.

27 * Fixed an Office file preview sometimes showing "Failed to load PDF document" instead of the spreadsheet or document view.

28 * Fixed long-running tasks failing with an authentication error after the app renewed your sign-in in the background; the running task now picks up the renewed sign-in and continues.

30 **3P**

32 * Added `coworkVmIpv6Enabled`, which gives the Cowork workspace VM an IPv6 address and route so the agent's tools can reach IPv6-only hosts through the device's own IPv6 connectivity. macOS and Windows; off by default. `coworkEgressAllowedHosts` still decides which hostnames the tools may reach.

33 * Added `sshTransport` (beta), which chooses the SSH engine that carries Code sessions: `system-openssh` runs the OpenSSH `ssh` program on the device, so connections can use the organization's own SSH setup (for example Kerberos sign-in, including on Windows), and `builtin` uses the app's built-in SSH library. Unset or `auto` keeps the build's default.

34 * Added a deprecation notice in the Setup window under any managed-configuration setting that is scheduled to stop being accepted, naming the date and what to use instead.

35 * Added organization-set session retention. `chatSessionRetentionDays`, `coworkSessionRetentionDays`, and `codeSessionRetentionDays` each delete that surface's idle sessions from the device, along with their files, after the set number of days (1 to 3650) without activity; unset deletes nothing. `sessionRetentionHold` suspends all automatic deletion for the users it is set for, as a legal hold. Projects, Spaces, and memory stay, and a Code session's uncommitted work stays on disk.

36 * Changed app launch to open the home composer on the last-used Chat or Cowork mode instead of always opening Cowork first.

37 * Changed hooks from organization plugins to also run in Chat, matching Cowork and Code.

38 * Deprecated an undocumented client-certificate fallback setting that releases from 1.49585.0 on no longer read, since the app now presents TLS client certificates natively. Deployments that still set it keep working unchanged, but their users see an in-app deprecation warning from November 3, 2026, and the setting stops being accepted on November 17, 2026; remove it once every device is on 1.49585.0 or later.

39 * Improved the connection test against gateways that require a TLS client certificate the device does not have: the result now says the device did not present a certificate the gateway accepts, so the connection could not be tested, instead of reporting the model as rejected.

40 * Fixed Claude Code's own settings on the device (a user's or project's settings file, or a `managed-settings.json`) being able to turn OpenTelemetry trace export back on after an administrator configured a collector with traces off; with `otlpEndpoint` set, Cowork and Code sessions now export traces only when `otlpTracesEnabled` is `true`.

41 * Fixed Amazon Bedrock and Google Vertex AI sessions retrying a failed request for several minutes when the AWS or Google Cloud credential on the machine had expired; the request now stops after one retry so the credential can be refreshed.

42</Update>

544<Update label="Known issue: Cowork on Windows" description="2026-09-10">

645 A Windows update released September 8, 2026 (including KB5124008) stops Cowork from reaching your files when it runs on your Windows PC, so tasks fail or the workspace does not start. This affects Cowork on third-party inference deployments and Cowork sessions that run on your computer rather than in the cloud. Cloud sessions and Claude Code, including the Code tab, are not affected. Chat cloud sessions are not affected, but Chat sessions on third-party inference deployments are affected if using advanced file analysis. The cause is a change in Windows, so restarting or reinstalling Claude does not help. We are investigating and working to resolve this as quickly as possible.

746</Update>

third-party/claude-desktop/configuration Changed · +55 / -8 lines

### Session retention

from line 83

8383| Re-run helper for silent refresh `inferenceCredentialHelperSilentRefreshEnabled` | `boolean` | MDM + Bootstrap | `true` | On credential expiry, re-run the helper (CLAUDE\_HELPER\_CONTEXT=mid-session-refresh) to recover silently. Turn off if the helper can’t run non-interactively. Defaults to `true`. |

8889| Credential kind `inferenceCredentialKind` | `enum` | MDM + Bootstrap | — | Selects the credential source. When set, only that source is used (no fallback). One of: `static`, `helper-script`, `interactive`, `vendor-profile`, `workforce`. Deprecated: `inferenceCredentialKind: "oauth" (Vertex AI)` (accepted until October 7, 2026); use "interactive" — the same Google sign-in under its new name (in hosted or nested documents, switch once every desktop is on a release that knows the Vertex "interactive" kind). If it is still present after that, "oauth" will no longer be a Vertex AI credential kind: the value will be reported as invalid and ignored — the device will then derive the kind from the credential fields present (Google sign-in when an OAuth client id is set), and the hosted editor will refuse to save the configuration until the kind is changed. Deprecated: `inferenceCredentialKind: "interactive" together with inferenceVertexWorkforceAudience (Vertex AI)` (accepted until October 7, 2026); use "workforce" — or remove inferenceVertexWorkforceAudience if Google sign-in ("interactive") is what is meant. If it is still present after that, the audience will no longer imply Workforce Identity: the kind will stay "interactive" (Google sign-in), which needs inferenceVertexOAuthClientId — without it the configuration will be reported as incomplete and inference will not start. |

from line 130

129130 On macOS and Windows the Cowork workspace VM is handed a copy of the script when it starts and evaluates it for each request itself; there `myIpAddress()` returns the VM's internal address rather than the device's, so a script that chooses by client subnet gives the VM its off-network answer (if that download fails, the VM connects directly). The Claude Code engine behind Chat, Cowork, and Code cannot evaluate a PAC file, so the app hands it one proxy (whichever the script returns for your inference endpoint) plus a bypass for loopback and `*.local` names. If the script answers `DIRECT` or only `SOCKS` for that endpoint, the engine uses no proxy at all, so have it return an HTTP `PROXY host:port` entry there; if the engine needs different rules, set `HTTPS_PROXY` and `NO_PROXY` in Claude Code managed settings, which win for the engine on the host.

130131 </Accordion>

131132

133 <Accordion title="coworkVmIpv6Enabled details">

134 When set to `true`, the Cowork workspace VM on macOS and Windows gets a static IPv6 address (a unique local `fd…` address) and an IPv6 default route on its virtual network next to its IPv4 address, and the VM's gateway forwards that traffic over the device's own IPv6 connectivity, as it already does for IPv4. Use it when the tools the agent runs in the VM (shell commands, package installs, `git`, plugin commands, and the whole engine under `requireCoworkFullVmSandbox`) must reach IPv6-only destinations. The VM's resolver then also returns IPv6 (AAAA) answers. A connection the VM makes over IPv6 succeeds only where the device's own IPv6 does; on a device without working IPv6, destinations that have both keep working over IPv4 and IPv6-only destinations stay unreachable. Because the VM's address is unique-local, most tools in it keep preferring IPv4 for destinations that have both, so IPv6 mostly carries traffic to IPv6-only destinations.

135

136 This is a reachability setting, not an egress control: `coworkEgressAllowedHosts` keeps deciding which hostnames the agent's tools may reach, by name, over either protocol, and IPv6 literals are still not accepted there. Hosts your policies allow must also be reachable, and filtered the way you intend, over IPv6 on your network.

137

138 Unset (default): the VM is IPv4-only and its resolver returns no IPv6 answers. A change takes effect the next time the workspace VM starts, typically at the next app launch. Does not apply to the Cowork workspace VM on Linux or to Code sessions, which use the device's own network stack.

139 </Accordion>

140

132141 <Accordion title="inferenceProvider details">

133142 The app activates 3P mode only when this is set and the required credential keys for the selected provider are present and valid; otherwise it launches in standard mode. Keys for providers other than the selected one are ignored. Each provider's required keys are documented on its dedicated page under Inference providers.

134143 </Accordion>

from line 385

376385

377386### Code surface

378387

380| ------------------------------------------------------------------------------------------ | ---------- | ---------------------- | ------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------- |

389| ------------------------------------------------------------------------------------------ | ---------- | ---------------------- | ------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

393| SSH connection engine `sshTransport` | `enum` | MDM + Bootstrap · Beta | — | Which SSH engine carries Code sessions: the OpenSSH ssh program on the device, or the app’s built-in SSH library. Unset or auto: the build’s default. One of: `auto`, `system-openssh`, `builtin`. |

384394

385395<AccordionGroup>

386396 <Accordion title="sshHostAllowlist details">

from line 404

394404 </Accordion>

395405

396406 <Accordion title="sshClientPath details">

397 Pins which OpenSSH client the app runs wherever it starts `ssh`: evaluating the user's SSH configuration (`ssh -G`), making the SSH connection and its channels, and the Code tab's terminal. `ssh-keygen` and `ssh-add` are taken from the same directory when they exist there, otherwise from PATH. The program must be OpenSSH 7.6 or newer; on macOS and Linux a wrapper script that ends in one is accepted, on Windows it must be a native `.exe` (not a .cmd, .bat or .ps1 script). When this key is set and the program is missing, cannot be run, or is too old, SSH sessions fail with an error telling the user to ask their IT administrator (the configured path is in its details) — the app never falls back to another ssh. On Windows, where SSH connections do not yet run through an ssh program, the key currently affects configuration evaluation and the terminal only.

407 Pins which OpenSSH client the app runs wherever it starts `ssh`: evaluating the user's SSH configuration (`ssh -G`), making the SSH connection and its channels, and the Code tab's terminal. `ssh-keygen` and `ssh-add` are taken from the same directory when they exist there, otherwise from PATH. The program must be OpenSSH 7.6 or newer (on Windows, Win32-OpenSSH 9.4 or newer to carry the connection); on macOS and Linux a wrapper script that ends in one is accepted, on Windows it must be a native `.exe` (not a .cmd, .bat or .ps1 script). When this key is set and the program is missing, cannot be run, or is too old, SSH sessions fail with an error telling the user to ask their IT administrator (the configured path is in its details) — the app never falls back to another ssh. The connection itself runs through this program on the system-OpenSSH engine, which the SSH connection engine setting's `system-openssh` value selects on every platform, including Windows; when the app's built-in SSH library makes the connection instead, this key still governs configuration evaluation (`ssh -G`), host-key lookups (`ssh-keygen`) and the terminal.

398408 </Accordion>

409

410 <Accordion title="sshTransport details">

411 `system-openssh`: the app makes every SSH connection by running an OpenSSH `ssh` program — the one `sshClientPath` names, otherwise the first `ssh` on the user's PATH (on Windows, a Win32-OpenSSH `ssh.exe`: the PATH one, else the in-box or Microsoft-installed client) — so the organization's own OpenSSH build, with its Kerberos/GSSAPI, certificate and `ssh_config` support, is what authenticates. The program must be OpenSSH 7.6 or newer (Windows: Win32-OpenSSH 9.4 or newer). When `sshClientPath` is set and that program cannot be used, sessions fail with an error telling the user to ask their IT administrator rather than falling back; when it is unset and Windows has no usable client, the built-in library is used.

412

413 `builtin`: the app's built-in SSH library makes the connection, whatever the build's default.

414

415 `auto` or unset: the build's default engine.

416

417 An explicit value applies to new connections (sessions already connected keep their engine) and overrides the build's default in both directions, including any remote switch-off Anthropic ships for the OpenSSH engine — so with `system-openssh` set, switching back is done here, by setting `builtin`.

418 </Accordion>

399419</AccordionGroup>

400420

401421### Cowork surface

from line 724

704724 <Accordion title="updateViaUpdatesHost details">

705725 By default the app asks `api.anthropic.com` which version to install. That host also serves the model APIs, so organizations that block un-approved LLM endpoints at the network edge end up blocking updates too.

706726

707 Turn this on to read the same feed from `releases.claude.com`, a hostname that serves only the desktop update-check route and carries no model API. `api.anthropic.com` can then stay blocked without breaking auto-update. Rollout behavior is unchanged; the installer download still comes from `downloads.claude.ai` as before.

727 Turn this on to read the same feed from `releases.claude.com`, a hostname that carries no model API. `api.anthropic.com` can then stay blocked without breaking auto-update. Rollout behavior is unchanged; the installer download still comes from `downloads.claude.ai` as before.

708728 </Accordion>

709729</AccordionGroup>

710730

from line 793

773793 </Accordion>

774794

775795 <Accordion title="otlpTracesEnabled details">

776 Enables Claude Code's session tracing (`CLAUDE_CODE_ENHANCED_TELEMETRY_BETA=1` + `OTEL_TRACES_EXPORTER=otlp`) in spawned Cowork tasks and Code sessions. Each user interaction exports a trace whose spans and events carry `trace_id`/`span_id`, enabling end-to-end correlation in your observability backend (metrics do not carry trace context; correlate those via `session.id`). Traces go to the collector endpoint and protocol configured above. The span structure may evolve between Claude Code releases; see the [Claude Code monitoring docs](https://code.claude.com/docs/en/monitoring-usage).

796 Enables Claude Code's session tracing (`CLAUDE_CODE_ENHANCED_TELEMETRY_BETA=1` + `OTEL_TRACES_EXPORTER=otlp`) in spawned Cowork tasks and Code sessions. Each user interaction exports a trace whose spans and events carry `trace_id`/`span_id`, enabling end-to-end correlation in your observability backend (metrics do not carry trace context; correlate those via `session.id`). Traces go to the collector endpoint and protocol configured above. When `otlpEndpoint` is set, this key alone decides whether those sessions export traces: leaving it unset or `false` keeps traces off even if Claude Code's own settings or managed settings (for example a `managed-settings.json` on the device) turn tracing on. Without `otlpEndpoint` it has no effect. The span structure may evolve between Claude Code releases; see the [Claude Code monitoring docs](https://code.claude.com/docs/en/monitoring-usage).

777797 </Accordion>

778798</AccordionGroup>

779799

780800## Limits

801

802### Session retention

803

805| ------------------------------------------------------------------------------------------------- | --------- | --------------- | ------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

807| Cowork retention period `coworkSessionRetentionDays` | `integer` | MDM + Bootstrap | — | Delete Cowork tasks, with their uploads and outputs, after this many days without activity. Unset: kept until the user deletes them. Spaces and memory stay. Range: 1–3650. |

808| Code retention period `codeSessionRetentionDays` | `integer` | MDM + Bootstrap | — | Delete Code sessions, conversation included, after this many days without activity. Unset: kept until the user deletes them. Uncommitted work stays on disk. Range: 1–3650. |

810

811<AccordionGroup>

812 <Accordion title="chatSessionRetentionDays details">

813 The app deletes whole sessions in the background shortly after it fetches this configuration (at launch and on each re-poll), or some minutes after launch and then daily when the configuration comes from device management alone. Idle time runs from the session's last activity; viewing an old Code session counts as activity, viewing an old chat or Cowork task without continuing it does not durably. A session that is running or open on screen is skipped and checked again on the next pass, as is a chat or Cowork task whose folder changed on disk within the period. Pinned sessions are not exempt. Minimum 1 day; a value that cannot be read as a whole number of days deletes nothing. Set through the served configuration or in the device-management profile that carries the rest of the configuration. What it does not reach: another account's sessions on the device (evaluated when that account signs in), Code sessions on a remote machine (SSH/WSL), Cowork background (dispatch) tasks the sidebar does not list, a chat or Cowork task whose folder cannot be located, and Claude Code files not named by a session id (prompt history, plans, shell snapshots), which keep Claude Code's own retention. Honored only by Claude Desktop configured for a third-party model provider.

814 </Accordion>

815

816 <Accordion title="coworkSessionRetentionDays details">

817 The app deletes whole sessions in the background shortly after it fetches this configuration (at launch and on each re-poll), or some minutes after launch and then daily when the configuration comes from device management alone. Idle time runs from the session's last activity; viewing an old Code session counts as activity, viewing an old chat or Cowork task without continuing it does not durably. A session that is running or open on screen is skipped and checked again on the next pass, as is a chat or Cowork task whose folder changed on disk within the period. Pinned sessions are not exempt. Minimum 1 day; a value that cannot be read as a whole number of days deletes nothing. Set through the served configuration or in the device-management profile that carries the rest of the configuration. What it does not reach: another account's sessions on the device (evaluated when that account signs in), Code sessions on a remote machine (SSH/WSL), Cowork background (dispatch) tasks the sidebar does not list, a chat or Cowork task whose folder cannot be located, and Claude Code files not named by a session id (prompt history, plans, shell snapshots), which keep Claude Code's own retention. Honored only by Claude Desktop configured for a third-party model provider.

818 </Accordion>

819

820 <Accordion title="codeSessionRetentionDays details">

821 The app deletes whole sessions in the background shortly after it fetches this configuration (at launch and on each re-poll), or some minutes after launch and then daily when the configuration comes from device management alone. Idle time runs from the session's last activity; viewing an old Code session counts as activity, viewing an old chat or Cowork task without continuing it does not durably. A session that is running or open on screen is skipped and checked again on the next pass, as is a chat or Cowork task whose folder changed on disk within the period. Pinned sessions are not exempt. Minimum 1 day; a value that cannot be read as a whole number of days deletes nothing. Set through the served configuration or in the device-management profile that carries the rest of the configuration. What it does not reach: another account's sessions on the device (evaluated when that account signs in), Code sessions on a remote machine (SSH/WSL), Cowork background (dispatch) tasks the sidebar does not list, a chat or Cowork task whose folder cannot be located, and Claude Code files not named by a session id (prompt history, plans, shell snapshots), which keep Claude Code's own retention. Honored only by Claude Desktop configured for a third-party model provider.

822 </Accordion>

823

824 <Accordion title="sessionRetentionHold details">

825 Meant to be set per user or group, through the served configuration's group overrides or in the same device-management profile that carries the rest of the configuration (a profile carrying only this key makes the device profile-managed, like any policy key); a hold in the device's profile also counts when the served configuration does not restate it. Deletion stops at the first configuration fetch that carries this value, before it takes effect as configuration at the next relaunch; a device that cannot reach its configuration server deletes nothing.

826 </Accordion>

827</AccordionGroup>

781828

782829### Token limits

783830

third-party/claude-desktop/configuration-changelog Changed · +32 / -0 lines

from line 4

55Configuration keys by Claude Desktop release. Each section lists keys added in that release, with the MDM key name (for plist/registry deployment) and the equivalent JSON shape (for local-file or bootstrap remote configuration).

7<Update label="v1.52386.0" description="2026-09-10">

8 <div className="cfg-keys">

9 | MDM key | Type | Description |

10 | ---------------------------------------------------------------------------------------------------- | --------- | ------------------------------- |

11 | [`sshTransport`](/docs/third-party/claude-desktop/configuration#sshtransport) · Beta | `enum` | SSH connection engine |

12 | [`chatSessionRetentionDays`](/docs/third-party/claude-desktop/configuration#chatsessionretentiondays) | `integer` | Chat retention period |

13 | [`coworkSessionRetentionDays`](/docs/third-party/claude-desktop/configuration#coworksessionretentiondays) | `integer` | Cowork retention period |

14 | [`codeSessionRetentionDays`](/docs/third-party/claude-desktop/configuration#codesessionretentiondays) | `integer` | Code retention period |

15 | [`sessionRetentionHold`](/docs/third-party/claude-desktop/configuration#sessionretentionhold) | `boolean` | Suspend session deletion |

16 | [`coworkVmIpv6Enabled`](/docs/third-party/claude-desktop/configuration#coworkvmipv6enabled) | `boolean` | Enable IPv6 in the workspace VM |

17 </div>

19 **JSON (e.g. for non-MDM users or Bootstrap):**

21 ```json theme={null}

22 {

23 "codeSurface": {

24 "sshTransport": "<auto|system-openssh|builtin>"

25 },

26 "sessionRetention": {

27 "chatDays": "<integer>",

28 "coworkDays": "<integer>",

29 "codeDays": "<integer>",

30 "legalHold": "<boolean>"

31 },

32 "workspace": {

33 "vmIpv6Enabled": "<boolean>"

34 }

35 }

36 ```

37</Update>

739<Update label="v1.49585.0" description="2026-09-08">

840 **Changed:**

941

claude-tag/concepts/agent-identity Changed · +2 / -2 lines

from line 26

2626

2727The diagram below traces one request through this process.

2828

29<img className="block dark:hidden" src="https://mintcdn.com/claude-ai/5JFKyLlO7sHMMf5J/images/claude-tag/diagrams/request-path.svg?fit=max&auto=format&n=5JFKyLlO7sHMMf5J&q=85&s=e8776cf0edd1f3ec912b9b044c9cc838" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request matching no rule is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path.svg" />

29<img className="block dark:hidden" src="https://mintcdn.com/claude-ai/oY6LusJt4c576Dc3/images/claude-tag/diagrams/request-path.svg?fit=max&auto=format&n=oY6LusJt4c576Dc3&q=85&s=a7f2e0f4303072b4c8e4f3197bf0cb12" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request that no rule, domain entry, or environment network access setting allows is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path.svg" />

3030

31<img className="hidden dark:block" src="https://mintcdn.com/claude-ai/5JFKyLlO7sHMMf5J/images/claude-tag/diagrams/request-path-dark.svg?fit=max&auto=format&n=5JFKyLlO7sHMMf5J&q=85&s=a2ed5f3270989c3ed3d9b9a78a72bff0" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request matching no rule is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path-dark.svg" />

31<img className="hidden dark:block" src="https://mintcdn.com/claude-ai/oY6LusJt4c576Dc3/images/claude-tag/diagrams/request-path-dark.svg?fit=max&auto=format&n=oY6LusJt4c576Dc3&q=85&s=a728805c81b642004e1e15da96967e4c" alt="Diagram showing the request path across three zones, labeled your Slack workspace, Anthropic's infrastructure, and your systems. A task mentioned in the Slack workspace runs in a session sandbox in the middle zone, one sandbox per thread, holding no credentials. Outbound requests pass to Agent Proxy, which injects the credential drawn from the credential store; a request that no rule, domain entry, or environment network access setting allows is blocked. Credentialed requests reach your systems, like GitHub, a data warehouse, monitoring, or any HTTP API. A dashed return path shows results posting back in the thread, as Claude." width="1000" height="440" data-path="images/claude-tag/diagrams/request-path-dark.svg" />

3232

3333<Steps>

3434 <Step title="Tag Claude in a channel">

claude-tag/users/use-cases/marketing-team Changed · +1 / -1 lines

from line 91

9191@Claude every Monday at 9am Eastern, post last week's campaign metrics as a chart: signups by campaign, week-over-week change, and a two-line note on anything unusual.

9292```

9393

94Naming the timezone matters, since schedules default to UTC. To list or cancel scheduled work later, see [Manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

94Name the timezone in the message, since schedules run in UTC. To list or cancel scheduled work later, see [Manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

9595

9696### Save brand voice rules to channel memory

9797

claude-tag/users/use-cases/review-documents Changed · +1 / -1 lines

from line 72

7272@Claude every Monday at 9am Pacific, check the shared folder for documents added in the past week, review each against the review checklist, and post the findings here.

7373```

7474

75Including the timezone matters, since schedules default to UTC. To list or cancel scheduled work later, see [Manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

75Include the timezone in the message, since schedules run in UTC. To list or cancel scheduled work later, see [Manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

7676

7777Keep the checklist as a document Claude can read, in the connected tool or linked in the channel, rather than re-describing its contents in [channel memory](/docs/claude-tag/users/memory).

7878

claude-tag/users/use-cases/your-own-channel Changed · +1 / -1 lines

from line 54

5454@Claude every Friday at 3pm Pacific, post a digest of this channel: what closed this week, what's still open, and what hasn't moved in five days. Skip anything with a ✅ reaction.

5555```

5656

57Give the digest a concrete threshold, like five days without movement, and stalled work shows up without you asking for it. React ✅ to anything you consider done, and the digest drops it. Include the timezone, since schedules default to UTC. To list, edit, or cancel scheduled work later, see [manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

57Give the digest a concrete threshold, like five days without movement, and stalled work shows up without you asking for it. React ✅ to anything you consider done, and the digest drops it. Include the timezone in the message, since schedules run in UTC. To list, edit, or cancel scheduled work later, see [Manage standing work](/docs/claude-tag/users/proactivity#manage-standing-work).

5858

5959### Hand a thread to a teammate

6060