What this read moved
1–25 of 58This capture is too large to show at once. Changes 1-25 of 58 are below, significant first; the rest are on the following screens.
agents-and-tools/tool-use/code-execution-tool Changed · +15 / -43 lines
## Compatibility ## Tool versions ## Model compatibility ## Platform availability
from line 4
44description: Run Python and bash code in a sandboxed container to analyze data, generate files, and iterate on solutions.
55---
66
7## Compatibility
8- [ZDR](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention): not eligible
9- Supported models: `claude-fable-5`, `claude-mythos-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-opus-4-5-20251101`, `claude-sonnet-5`, `claude-sonnet-4-6`, `claude-sonnet-4-5-20250929`, `claude-haiku-4-5-20251001`
10- Platforms: Claude API, Claude Platform on AWS, Microsoft Foundry [1]; not available on Amazon Bedrock, Google Cloud
11- Every supported model accepts all three [tool versions](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#tool-versions). On Claude Haiku 4.5, programmatic tool calling and REPL state persistence aren't available, so the newer versions behave like `code_execution_20250825` there.
12- For [Claude Mythos Preview](https://anthropic.com/glasswing), code execution is supported on the Claude API and Microsoft Foundry.
131. On [Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry), code execution requires a [Hosted on Anthropic deployment](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry#additional-features-not-supported-when-hosted-on-azure).
14
715Claude can analyze data, create visualizations, perform complex calculations, run system commands, create and edit files, and process uploaded files directly within the API conversation. The code execution tool allows Claude to run Bash commands and manipulate files, including writing code, in a secure, sandboxed environment.
816
917**Code execution is free when used with web search or web fetch (`web_search_20260209`, `web_fetch_20260209`, or later).** When one of those tools is in your request, there are no additional charges for code execution in that request beyond standard token costs. This covers both the code execution behind dynamic filtering and any code Claude runs directly. Standard code execution pricing applies when they are not included.
from line 22
1422 Reach out through the [feedback form](https://forms.gle/LTAU6Xn2puCJMi1n6) to share your feedback on this feature.
1523</Note>
1624
17<Note>
18 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
19</Note>
25## Tool versions
2026
21## Model compatibility
27The code execution tool has three current versions, and every [supported model](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#compatibility) accepts all three. Each version builds on the previous one:
2228
23The code execution tool is available on the following models:
24
25| Model | Tool versions |
26| ---------------------------------------------- | ------------------------------------------------------------------------------- |
27| Claude Opus 5 (claude-opus-5) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
28| Claude Fable 5 (claude-fable-5) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
29| Claude Mythos 5 (claude-mythos-5) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
30| Claude Sonnet 5 (claude-sonnet-5) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
31| Claude Opus 4.8 (claude-opus-4-8) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
32| Claude Opus 4.7 (claude-opus-4-7) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
33| Claude Opus 4.6 (claude-opus-4-6) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
34| Claude Sonnet 4.6 (claude-sonnet-4-6) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
35| Claude Opus 4.5 (claude-opus-4-5-20251101) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
36| Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
37| Claude Haiku 4.5 (claude-haiku-4-5-20251001) | `code_execution_20250825`, `code_execution_20260120`, `code_execution_20260521` |
38
39Each tool version builds on the previous one:
40
4129* `code_execution_20250825` supports Bash commands and file operations.
4230* `code_execution_20260120` adds REPL state persistence and [programmatic tool calling](https://platform.claude.com/docs/en/agents-and-tools/tool-use/programmatic-tool-calling) from within the sandbox. Claude Haiku 4.5 accepts the `code_execution_20260120` and `code_execution_20260521` tool types, but programmatic tool calling and the REPL state persistence that depends on it aren't available on it, so the newer versions behave like `code_execution_20250825` there.
4331* `code_execution_20260521` is the same runtime as `code_execution_20260120`. The difference is that the tool description tells Claude about the 90-second wall-clock limit on each Python cell in programmatic tool calling, so Claude can budget long-running cells. A cell that exceeds the limit returns a normal code execution result with a non-zero `return_code` and a `detection_timeout` status message in its output. This is separate from the `execution_time_exceeded` [error code](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#errors), which the API returns when a whole tool invocation exceeds the maximum execution time.
from line 32
4432
4533None of the three tool versions requires an `anthropic-beta` header. The legacy code execution beta headers remain valid opt-ins.
4634
47The examples on this page use `code_execution_20250825`, which covers the Bash and file operations they demonstrate and behaves the same way on every model in the table; use `code_execution_20260120` or later when you need programmatic tool calling or REPL state persistence. The current [web search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool) and [web fetch](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-fetch-tool) tools (`web_search_20260209`, `web_fetch_20260209`, and later) require `code_execution_20260120` or later as their code execution version.
35The examples on this page use `code_execution_20250825`, which covers the Bash and file operations they demonstrate and behaves the same way on every supported model; use `code_execution_20260120` or later when you need programmatic tool calling or REPL state persistence. The current [web search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool) and [web fetch](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-fetch-tool) tools (`web_search_20260209`, `web_fetch_20260209`, and later) require `code_execution_20260120` or later as their code execution version.
4836
37Older tool versions aren't guaranteed to stay compatible with newer models. When you adopt a new model, check [Tool versions](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#tool-versions) and [Compatibility](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#compatibility), and prefer the newest tool version your integration supports.
38
4939<Note>
5040 If you're still using the legacy `code_execution_20250522` (Python only), see [Upgrade to latest tool version](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#upgrade-to-latest-tool-version) to migrate from it.
5141</Note>
5242
53<Warning>
54 Older tool versions are not guaranteed to stay compatible with newer models. When you adopt a new model, check the [model compatibility table](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#model-compatibility) and prefer the newest tool version your integration supports.
55</Warning>
56
57## Platform availability
58
59Code execution is available on:
60
61* **Claude API** (Anthropic)
62* **[Claude Platform on AWS](https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws)**
63* **[Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry)** (requires a [Hosted on Anthropic deployment](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry#additional-features-not-supported-when-hosted-on-azure))
64
65Code execution is not currently available on Amazon Bedrock or Google Cloud.
66
67<Note>
68 For [Claude Mythos Preview](https://anthropic.com/glasswing), code execution is supported on the Claude API and Microsoft Foundry only. It is not available for Mythos Preview on Amazon Bedrock, Claude Platform on AWS, or Google Cloud.
69</Note>
70
7143## Quick start
7244
7345Here's an example that asks Claude to perform a calculation:
from line 1462
14901462
14911463## Upgrade to latest tool version
14921464
1493The latest tool version is `code_execution_20260521`. To move between the three current versions, update the `type` string in your request: all three return the response blocks documented in [Response format](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#response-format). See [Model compatibility](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#model-compatibility) for what each version adds and which models support it.
1465The latest tool version is `code_execution_20260521`. To move between the three current versions, update the `type` string in your request: all three return the response blocks documented in [Response format](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#response-format). See [Tool versions](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#tool-versions) for what each version adds and [Compatibility](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#compatibility) for the models that support them.
14941466
14951467The rest of this section covers migrating from the legacy Python-only `code_execution_20250522` to the current tool versions.
14961468
agents-and-tools/tool-use/programmatic-tool-calling Changed · +9 / -26 lines
## Compatibility ## Model compatibility
from line 4
44description: Let Claude call your tools from code in the code execution container, cutting model round trips and token use in multi-tool workflows.
55---
66
7## Compatibility
8- [ZDR](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention): not eligible
9- Supported models: `claude-fable-5`, `claude-mythos-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-opus-4-5-20251101`, `claude-sonnet-5`, `claude-sonnet-4-6`, `claude-sonnet-4-5-20250929`
10- Platforms: Claude API, Claude Platform on AWS, Microsoft Foundry [1]; not available on Amazon Bedrock, Google Cloud
11- Programmatic tool calling requires the code execution tool with the `code_execution_20260120` or later [tool version](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#tool-versions).
12- Claude Haiku 4.5 accepts the `code_execution_20260120` and later tool versions but doesn't support programmatic tool calling.
131. On [Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry), programmatic tool calling requires a [Hosted on Anthropic deployment](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry#additional-features-not-supported-when-hosted-on-azure).
14
715Programmatic tool calling allows Claude to write code that calls your tools programmatically within a [code execution](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool) container, rather than requiring round trips through the model for each tool invocation. This reduces latency for multi-tool workflows and decreases token consumption by allowing Claude to filter or process data before it reaches the model's context window. On agentic search benchmarks like [BrowseComp](https://arxiv.org/abs/2504.12516) and [DeepSearchQA](https://github.com/google-deepmind/deepsearchqa), which test multistep web research and complex information retrieval, adding programmatic tool calling on top of basic search tools improved performance by an average of 11% while using 24% fewer input tokens (see [Improved web search with dynamic filtering](https://claude.com/blog/improved-web-search-with-dynamic-filtering)).
816
917Consider checking budget compliance across 20 employees: the traditional approach requires 20 separate model round-trips, pulling thousands of expense line items into the context along the way. With programmatic tool calling, a single script runs all 20 lookups, filters the results, and returns only the employees who exceeded their limits, shrinking what Claude needs to reason over from hundreds of kilobytes down to a handful of lines.
from line 20
1220 For a deeper look at the inference and context costs that programmatic tool calling addresses, see [Advanced tool use](https://www.anthropic.com/engineering/advanced-tool-use).
1321</Tip>
1422
15<Note>
16 This feature requires the code execution tool to be enabled.
17</Note>
18
19<Note>
20 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
21</Note>
22
23## Model compatibility
24
25Programmatic tool calling requires `code_execution_20260120` or later, which is supported on the following models:
26
27| Model |
28| ---------------------------------------------- |
29| Claude Fable 5 (claude-fable-5) |
30| Claude Mythos 5 (claude-mythos-5) |
31| Claude Opus 5 (claude-opus-5) |
32| Claude Opus 4.8 (claude-opus-4-8) |
33| Claude Opus 4.7 (claude-opus-4-7) |
34| Claude Opus 4.6 (claude-opus-4-6) |
35| Claude Sonnet 5 (claude-sonnet-5) |
36| Claude Sonnet 4.6 (claude-sonnet-4-6) |
37| Claude Opus 4.5 (claude-opus-4-5-20251101) |
38| Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) |
39
40For the full code execution tool version matrix, see the [code execution tool model compatibility table](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#model-compatibility). Programmatic tool calling is available on the Claude API, [Claude Platform on AWS](https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws), and [Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry). On Microsoft Foundry, programmatic tool calling requires a [Hosted on Anthropic deployment](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry#additional-features-not-supported-when-hosted-on-azure). It is not currently available on Amazon Bedrock or Google Cloud.
23Programmatic tool calling requires the [code execution tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool) with tool version `code_execution_20260120` or later.
4124
4225## Quick start
4326
build-with-claude/thinking Changed · +3 / -3 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111A model that answers in a single pass has to get everything right on the first try: no scratch work, no checking, no changing course halfway through. For a proof, a tricky bug, or a long agentic task, the first approach is often not the best one.
from line 805
805805
806806The `thinking` parameter controls whether Claude thinks in [thinking blocks](https://platform.claude.com/docs/en/build-with-claude/thinking) before answering; the `effort` parameter controls how much work Claude puts into the whole response, which in adaptive mode includes how often and how deeply it thinks. Don't pass `adaptive` as an `effort` value: `adaptive` is a thinking mode, not an effort level.
807807
808For what each effort level does to thinking behavior, see the [per-level thinking behavior table](https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost#effort-levels) on the [Steering thinking](https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost) page. The [Effort](https://platform.claude.com/docs/en/build-with-claude/effort) page documents the parameter itself, including which levels each model supports. On Claude Opus 4.5, the only extended-thinking-only model that supports effort, effort composes with `budget_tokens`. See [Budget rules and tuning](https://platform.claude.com/docs/en/build-with-claude/extended-thinking#budget-rules-and-tuning).
808To learn what each effort level does to thinking behavior, see the [per-level thinking behavior table](https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost#effort-levels) on the [Steering thinking](https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost) page. The [Effort](https://platform.claude.com/docs/en/build-with-claude/effort) page documents the parameter itself, including which levels each model supports. On Claude Opus 4.5, the only extended-thinking-only model that supports effort, effort composes with `budget_tokens`. See [Budget rules and tuning](https://platform.claude.com/docs/en/build-with-claude/extended-thinking#budget-rules-and-tuning).
809809
810810With the two controls separated this way, pick the one that matches your goal:
811811
from line 1004
10041004
10051005When continuing a conversation on the same model, pass each thinking block back to the API exactly as received, including blocks whose `thinking` field is empty. Don't edit or reconstruct them. Reading the summary text for display is fine: the API rejects blocks whose returned content has been modified, not blocks you have read. Text placed in an empty omitted `thinking` field is [ignored rather than rejected](https://platform.claude.com/docs/en/build-with-claude/thinking#controlling-thinking-display).
10061006
1007For how thinking blocks are handled when you switch models mid-conversation, see [Thinking block preservation by model](https://platform.claude.com/docs/en/build-with-claude/thinking#thinking-block-preservation-by-model).
1007To learn how thinking blocks are handled when you switch models mid-conversation, see [Thinking block preservation by model](https://platform.claude.com/docs/en/build-with-claude/thinking#thinking-block-preservation-by-model).
10081008
10091009Two exceptions, covered in [Fallback credit](https://platform.claude.com/docs/en/build-with-claude/fallback-credit):
10101010
build-with-claude/token-counting Changed · +3 / -3 lines
from line 1410
14101410
14111411| Usage tier | Requests per minute (RPM) |
14121412| ---------- | ------------------------- |
1413| Start | 2,000 |
1414| Build | 4,000 |
1415| Scale | 8,000 |
1413| Start | 5,000 |
1414| Build | 10,000 |
1415| Scale | 20,000 |
14161416
14171417<Note>
14181418 Token counting and message creation have separate and independent rate limits. Usage of one does not count against the limits of the other.
manage-claude/admin-api-keys Changed · +14 / -14 lines
from line 61
6161
6262When you create a Claude Enterprise key, select every scope that the APIs you plan to call require. Scopes are fixed at creation; to add a scope later, create a new key.
6363
64| To call... | Select these scopes |
65| --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------- |
66| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): list and look up members and invites; read custom roles and their permissions | `read:members` |
67| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): change member roles, remove members, create and withdraw invites | `write:members` |
68| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): read groups and their members | `read:rbac_groups` |
69| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): create, rename, and delete groups; add and remove group members; assign groups on invite creation | `write:rbac_groups` |
70| [Spend Limits API](https://platform.claude.com/docs/en/manage-claude/spend-limits-api): read members' effective spend limits and increase requests | `read:spend_limits` |
71| [Spend Limits API](https://platform.claude.com/docs/en/manage-claude/spend-limits-api): set or clear per-user spend limits; approve or deny increase requests | `write:spend_limits` |
72| [Claude Enterprise Analytics API](https://platform.claude.com/docs/en/manage-claude/analytics-api): engagement, adoption, cost, and usage reports | `read:analytics` |
73| [Compliance API Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed): organization-wide activity events | `read:compliance_activities` |
74| [Compliance API chat, file, and project endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-content-data) and [Compliance API session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions): read chats, files, projects, Cowork and Claude Code session transcripts, and [organization users](https://platform.claude.com/docs/en/manage-claude/compliance-org-data#list-organization-users) | `read:compliance_user_data` |
75| [Compliance API chat, file, and project endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-content-data): delete chats, files, and projects | `delete:compliance_user_data` |
76| [Compliance API organization endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-org-data): read organization metadata and effective settings | `read:compliance_org_data` |
77| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management) read endpoints and every Compliance API read endpoint, with a single read-only scope (for security-audit integrations; does not include the Spend Limits or Analytics APIs) | `read:org_audit` |
64| To call... | Select these scopes |
65| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------- |
66| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): list and look up members and invites; read custom roles and their permissions | `read:members` |
67| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): change member roles, remove members, create and withdraw invites | `write:members` |
68| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): read groups and their members | `read:rbac_groups` |
69| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management): create, rename, and delete groups; add and remove group members; assign groups on invite creation | `write:rbac_groups` |
70| [Spend Limits API](https://platform.claude.com/docs/en/manage-claude/spend-limits-api): read members' effective spend limits and increase requests | `read:spend_limits` |
71| [Spend Limits API](https://platform.claude.com/docs/en/manage-claude/spend-limits-api): set or clear per-user spend limits; approve or deny increase requests | `write:spend_limits` |
72| [Claude Enterprise Analytics API](https://platform.claude.com/docs/en/manage-claude/analytics-api): engagement, adoption, cost, and usage reports | `read:analytics` |
73| [Compliance API Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed): organization-wide activity events | `read:compliance_activities` |
74| [Compliance API chat, file, and project endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-content-data) and [Compliance API session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions): read chats, files, projects, session transcripts, and [organization users](https://platform.claude.com/docs/en/manage-claude/compliance-org-data#list-organization-users) | `read:compliance_user_data` |
75| [Compliance API chat, file, and project endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-content-data): delete chats, files, and projects | `delete:compliance_user_data` |
76| [Compliance API organization endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-org-data): read organization metadata and effective settings | `read:compliance_org_data` |
77| Admin API [user management](https://platform.claude.com/docs/en/manage-claude/user-management) read endpoints and every Compliance API read endpoint, with a single read-only scope (for security-audit integrations; does not include the Spend Limits or Analytics APIs) | `read:org_audit` |
7878
7979The Compliance and Analytics APIs must be enabled for your organization before keys with those scopes can be used. See [Set up the Compliance API](https://platform.claude.com/docs/en/manage-claude/compliance-api-access#set-up-the-compliance-api) and [Get access to the Claude Enterprise Analytics API](https://platform.claude.com/docs/en/manage-claude/analytics-api#get-access-to-the-claude-enterprise-analytics-api).
8080
manage-claude/cmek-aws-kms Changed · +3 / -3 lines
from line 532
532532On [Claude Platform on AWS](https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws), CMEK uses AWS KMS keys only, and setup differs from the preceding sections in these ways:
533533
534534* **Principal:** Your key policy grants access to the AWS service principal `aws-external-anthropic.amazonaws.com`. Anthropic's IAM role and account ID are not used, so the [ARN for Anthropic](https://platform.claude.com/docs/en/manage-claude/cmek-aws-kms#amazon-resource-name-arn-for-anthropic) does not apply.
535* **Key requirements:** The key must be a symmetric KMS key with encrypt and decrypt usage, single-region, and in the same AWS account and region as the workspace you attach it to. Multi-region keys (key IDs that begin with `mrk-`) and alias ARNs are rejected; use the key ARN.
536* **No separate validation step:** The key is validated when you attach it to a workspace. The attach call performs an encrypt/decrypt round against the key with that workspace's compartment ID as the encryption context, so a key policy problem surfaces at attach time rather than at registration. Unlike the Claude Platform policy earlier on this page, an `EncryptionContext` condition therefore needs no all-zeros entry.
535* **Key requirements:** The key must be a symmetric KMS key with encrypt and decrypt usage, single-region, and in the same AWS account and region as the workspace you attach it to. Cross-account keys are not supported: the key must be in the AWS account that hosts your organization. Multi-region keys (key IDs that begin with `mrk-`) and alias ARNs are rejected when you register the key; use the key ARN.
536* **No separate validation step:** Apart from those checks on the key ARN at registration, the key is validated when you attach it to a workspace. The attach call performs an encrypt/decrypt round against the key with that workspace's compartment ID as the encryption context, so a key policy problem surfaces at attach time rather than at registration. Unlike the Claude Platform policy earlier on this page, an `EncryptionContext` condition therefore needs no all-zeros entry.
537537* **Where you manage keys:** Register and attach keys in the Claude Console, signed in through AWS with the Admin role. The external key endpoints are also available on Claude Platform on AWS, authorized through [IAM actions](https://platform.claude.com/docs/en/api/claude-platform-on-aws-iam-actions#encryption-keys); there, a key is identified by its KMS key ARN rather than an `ekey_` ID.
538538
539539<Warning>
from line 614
614614
615615<Steps>
616616 <Step title="Register the key">
617 In the Claude Console, open **Settings > Encryption keys** and click **Add key**. Enter a display name, then choose the key from the key picker or choose **Enter ARN manually** and paste the key ARN, and click **Add**. The picker lists the enabled, customer-managed, symmetric, single-region keys in your account in one of your organization's regions; for a key the picker doesn't list, enter the ARN. It lists keys only if the principal you signed in with can call `kms:ListKeys` and `kms:DescribeKey`.
617 In the Claude Console, open **Settings > Encryption keys** and click **Add key**. Enter a display name, then choose the key from the key picker or choose **Enter ARN manually** and paste the key ARN, and click **Add**. The key must be in the AWS account that hosts your organization; cross-account keys are not supported. The picker lists the enabled, customer-managed, symmetric, single-region keys in your account in one of your organization's regions; for a key the picker doesn't list, enter the ARN. It lists keys only if the principal you signed in with can call `kms:ListKeys` and `kms:DescribeKey`.
618618 </Step>
619619
620620 <Step title="Attach the key to a workspace">
manage-claude/compliance-api-access Changed · +9 / -9 lines
from line 16
1616
1717## Which key do you need?
1818
19| Key type | Created in | Used for | Works with the Compliance API? |
20| ---------------------------------------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------ |
21| **Compliance Access Key** (`sk-ant-api01-...`) | [claude.ai > Organization settings > API](https://claude.ai/admin-settings/api-access) | Activity Feed, chats, files, projects, Cowork and Claude Code sessions, users, organization metadata, and organization settings | Yes (all endpoints) |
22| **Admin API key** (`sk-ant-admin01-...`) | [Claude Console > Settings > Admin keys](https://platform.claude.com/settings/admin-keys) | The [Admin API](https://platform.claude.com/docs/en/manage-claude/admin-api) and the Compliance API Activity Feed | Activity Feed only |
23| **Analytics API key** | [claude.ai > Organization settings > API](https://claude.ai/admin-settings/api-access) | The Claude Enterprise Analytics API (see [Analytics APIs](https://platform.claude.com/docs/en/manage-claude/analytics-api)) | No |
24| **Claude API key** (`sk-ant-api03-...`) | [Claude Console > Settings > API keys](https://platform.claude.com/settings/keys) | Calling Claude models through the [Claude API](https://platform.claude.com/docs/en/api/overview) | No |
19| Key type | Created in | Used for | Works with the Compliance API? |
20| ---------------------------------------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------ |
21| **Compliance Access Key** (`sk-ant-api01-...`) | [claude.ai > Organization settings > API](https://claude.ai/admin-settings/api-access) | Activity Feed, chats, files, projects, sessions (in apps such as Cowork and Claude Code), users, organization metadata, and organization settings | Yes (all endpoints) |
22| **Admin API key** (`sk-ant-admin01-...`) | [Claude Console > Settings > Admin keys](https://platform.claude.com/settings/admin-keys) | The [Admin API](https://platform.claude.com/docs/en/manage-claude/admin-api) and the Compliance API Activity Feed | Activity Feed only |
23| **Analytics API key** | [claude.ai > Organization settings > API](https://claude.ai/admin-settings/api-access) | The Claude Enterprise Analytics API (see [Analytics APIs](https://platform.claude.com/docs/en/manage-claude/analytics-api)) | No |
24| **Claude API key** (`sk-ant-api03-...`) | [Claude Console > Settings > API keys](https://platform.claude.com/settings/keys) | Calling Claude models through the [Claude API](https://platform.claude.com/docs/en/api/overview) | No |
2525
2626A Claude Enterprise tenant has one **parent organization** that centralizes identity, SSO, and SCIM for every workload organization beneath it. These workload organizations are the parent's **linked organizations**.
2727
from line 34
3434Setup is one flow: enable the Compliance API for your organization, then create a Compliance Access Key in claude.ai. A Claude Console organization instead [creates an Admin API key](https://platform.claude.com/docs/en/manage-claude/compliance-api-access#create-an-admin-api-key) after enablement; Admin API keys reach the [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed) only.
3535
3636<Warning>
37 A Compliance Access Key with `read:compliance_user_data` can read every chat, file, project, and Cowork and Claude Code session transcript in every linked organization, including content the primary owner has not seen. A key with `delete:compliance_user_data` can permanently delete chats, files, and projects. Treat Compliance Access Keys like production database credentials: store them in a secrets manager, never in source control or SIEM forwarder configuration.
37 A Compliance Access Key with `read:compliance_user_data` can read every chat, file, project, and session transcript in every linked organization, including content the primary owner has not seen. A key with `delete:compliance_user_data` can permanently delete chats, files, and projects. Treat Compliance Access Keys like production database credentials: store them in a secrets manager, never in source control or SIEM forwarder configuration.
3838</Warning>
3939
4040<Steps>
from line 46
4646 * **Claude Console organizations linked to a parent organization:** There is nothing to turn on in Claude Console. Ask the primary owner of your parent organization to enable the Compliance API in claude.ai, or contact your account team.
4747
4848 <Warning>
49 **Turning the Compliance API off stops activity recording.** An organization admin can turn the Compliance API off at any time with the same **Compliance API** toggle that turns it on. While the Compliance API is off, no activity events are recorded for your organization, so the [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed) receives no new events. If your organization is enrolled in [Access Transparency](https://platform.claude.com/docs/en/manage-claude/access-transparency), turning the Compliance API off also stops Access Transparency event delivery. Activity that is not recorded while the Compliance API is off cannot be recovered later. Turning the Compliance API back on resumes recording from that point forward; activity that was already recorded is not deleted. For Claude Enterprise organizations, the Compliance API setting in claude.ai also governs transcript capture for Cowork and Claude Code sessions on users' machines (local sessions): capture starts when the Compliance API is enabled and stops while it is off, and transcript content from sessions that run while it is off is not captured and cannot be recovered later.
49 **Turning the Compliance API off stops activity recording.** An organization admin can turn the Compliance API off at any time with the same **Compliance API** toggle that turns it on. While the Compliance API is off, no activity events are recorded for your organization, so the [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed) receives no new events. If your organization is enrolled in [Access Transparency](https://platform.claude.com/docs/en/manage-claude/access-transparency), turning the Compliance API off also stops Access Transparency event delivery. Activity that is not recorded while the Compliance API is off cannot be recovered later. Turning the Compliance API back on resumes recording from that point forward; activity that was already recorded is not deleted. For Claude Enterprise organizations, the Compliance API setting in claude.ai also governs transcript capture for [local sessions](https://platform.claude.com/docs/en/manage-claude/compliance-sessions#retrieve-local-sessions) (sessions on users' machines): capture starts when the Compliance API is enabled and stops while it is off, and transcript content from sessions that run while it is off is not captured and cannot be recovered later.
5050 </Warning>
5151
5252 A standalone Claude Console organization uses Admin API keys rather than Compliance Access Keys: after enablement, skip the remaining steps and [create a new Admin API key](https://platform.claude.com/docs/en/manage-claude/compliance-api-access#create-an-admin-api-key) instead. The remaining steps provision Compliance Access Keys, which are available only to organizations that are part of a Claude Enterprise tenant.
from line 75
7575 | Scope | Grants |
7676 | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
7777 | `read:compliance_activities` | Read the Activity Feed. A key covering the parent organization reads events for the parent organization and all linked organizations. |
78 | `read:compliance_user_data` | Read user chats, messages, files, projects, Cowork and Claude Code sessions and their transcripts, organization users, and group members |
78 | `read:compliance_user_data` | Read user chats, messages, files, projects, session metadata and transcripts, organization users, and group members |
7979 | `delete:compliance_user_data` | Delete user chats, files, and projects |
8080 | `read:compliance_org_data` | Read organization metadata (names, types, roles, and groups) and the effective settings in force for organizations under the parent organization. User listings and group membership require `read:compliance_user_data`. |
8181
manage-claude/compliance-faq Changed · +3 / -3 lines
from line 66
6666 <Accordion title="Does the Activity Feed include prompt or message content?">
6767 No. The Activity Feed records who did what and when (authentication, chat creation, file uploads, project changes, administrative actions, and similar resource events), but it does not capture the prompt text or model responses inside chats or messages.
6868
69 To retrieve message bodies and file contents, use the chat, message, and file endpoints with a Compliance Access Key carrying `read:compliance_user_data`. The same key and scope retrieve transcripts of Cowork and Claude Code sessions on users' machines through the [local session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions#retrieve-local-sessions), and transcripts of Cowork sessions in the cloud through the [remote session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions#retrieve-remote-sessions). These endpoints serve Claude Enterprise content only; Claude Console workloads, and Claude API workloads authenticated with an API key, expose administrative and resource events through the Activity Feed but do not expose prompt text or model responses through the Compliance API.
69 To retrieve message bodies and file contents, use the chat, message, and file endpoints with a Compliance Access Key carrying `read:compliance_user_data`. The same key and scope retrieve transcripts of sessions on users' machines (such as Cowork and Claude Code sessions) through the [local session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions#retrieve-local-sessions), and transcripts of Cowork sessions in the cloud through the [remote session endpoints](https://platform.claude.com/docs/en/manage-claude/compliance-sessions#retrieve-remote-sessions). These endpoints serve Claude Enterprise content only; Claude Console workloads, and Claude API workloads authenticated with an API key, expose administrative and resource events through the Activity Feed but do not expose prompt text or model responses through the Compliance API.
7070 </Accordion>
7171
7272 <Accordion title="Do Cowork, Claude Code, Claude Science, and Claude for Microsoft 365 sessions appear in the Compliance API?">
from line 77
7777 The local and remote session endpoints are stable for Cowork and Claude Code sessions; coverage of Claude Science and Claude for Microsoft 365 sessions is in beta.
7878 </Accordion>
7979
80 <Accordion title="What do Cowork, Claude Code, Claude Science, and Claude for Microsoft 365 session transcripts include?">
80 <Accordion title="What do session transcripts include?">
8181 Local and remote session transcripts both carry user prompts, assistant responses, and tool calls and results. For local sessions (on users' machines), that is what Claude was asked to do and what it returned, not what happened on the device.
8282
8383 | Data | Local sessions (on users' machines) | Remote sessions (in the cloud) |
from line 109
109109 <Accordion title="What does the Compliance API not capture?">
110110 The Compliance API has known coverage boundaries: the Activity Feed records resource events but not prompt or response text, Claude Console and Claude API workloads authenticated with an API key expose no message content at all, and content removed by your retention policy, deleted by a user in claude.ai, or hard-deleted through the Compliance API is not recoverable. For the full coverage boundaries and delivery contract, see [Delivery guarantees and completeness](https://platform.claude.com/docs/en/manage-claude/compliance-integration-patterns#delivery-guarantees-and-completeness).
111111
112 Cowork and Claude Code session transcripts have boundaries of their own. Local sessions are captured only as their requests reach the Claude API, so on-device activity that never reaches the API is not captured. Claude Code sessions authenticated with a Claude Console API key, Claude Code sessions run through a third-party cloud platform (Amazon Bedrock, Google Cloud, or Microsoft Foundry), and Claude Code on the web are not captured either; organizations with HIPAA readiness enabled get no local session data; and sessions for which zero data retention is in effect are excluded. No session transcript, local or remote, includes thinking blocks or tool definitions. Organizations that use [customer-managed encryption keys](https://platform.claude.com/docs/en/manage-claude/cmek) receive local session transcripts as usual. While the key cannot be used, the messages endpoint returns [503 Service Unavailable](https://platform.claude.com/docs/en/manage-claude/compliance-errors#local-sessions-temporarily-unavailable) instead of transcript content, and session metadata is still listed.
112 Session transcripts have boundaries of their own. Local sessions are captured only as their requests reach the Claude API, so on-device activity that never reaches the API is not captured. Claude Code sessions authenticated with a Claude Console API key, Claude Code sessions run through a third-party cloud platform (Amazon Bedrock, Google Cloud, or Microsoft Foundry), and Claude Code on the web are not captured either; organizations with HIPAA readiness enabled get no local session data; and sessions for which zero data retention is in effect are excluded. No session transcript, local or remote, includes thinking blocks or tool definitions. Organizations that use [customer-managed encryption keys](https://platform.claude.com/docs/en/manage-claude/cmek) receive local session transcripts as usual. While the key cannot be used, the messages endpoint returns [503 Service Unavailable](https://platform.claude.com/docs/en/manage-claude/compliance-errors#local-sessions-temporarily-unavailable) instead of transcript content, and session metadata is still listed.
113113 </Accordion>
114114</AccordionGroup>
115115
manage-claude/inference-hooks Changed · +5 / -5 lines
from line 70
7070
7171Both features serve security, legal, and compliance teams at Claude Enterprise organizations.
7272
73| | Inference hooks | Compliance API |
74| ------------ | --------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------- |
75| When it acts | Inline, before inference runs | After the fact |
76| What it does | Allows or denies each governed request in real time | Retrieves activity, chats, files, projects, Cowork and Claude Code session transcripts, and users for audit and export |
77| Direction | Anthropic calls your AI security server | You call Anthropic's API |
73| | Inference hooks | Compliance API |
74| ------------ | --------------------------------------------------- | ----------------------------------------------------------------------------------------------- |
75| When it acts | Inline, before inference runs | After the fact |
76| What it does | Allows or denies each governed request in real time | Retrieves activity, chats, files, projects, session transcripts, and users for audit and export |
77| Direction | Anthropic calls your AI security server | You call Anthropic's API |
7878
7979Use Inference hooks to stop a request before it reaches the model, and the [Compliance API](https://platform.claude.com/docs/en/manage-claude/compliance-api) to audit what happened afterward.
8080
manage-claude/inference-hooks-configuration Changed · +7 / -3 lines
from line 8
88 Inference hooks are in beta and available to Claude Enterprise organizations. Configuring them requires the `organization:manage` permission, which the built-in Admin, Owner, and Primary owner roles hold, as does any custom role granted it.
99</Note>
1010
11Inference hooks send prompts from your organization to an AI security server you choose, and hold each request for an allow or deny verdict before Claude processes it. This page walks through turning the feature on, connecting your server, and controlling enforcement. For what Inference hooks are and when to use them, see the [Inference hooks overview](https://platform.claude.com/docs/en/manage-claude/inference-hooks). For building the AI security server itself, see [Develop an Inference hooks integration](https://platform.claude.com/docs/en/manage-claude/inference-hooks-endpoint).
11Inference hooks send prompts from your organization to an AI security server you choose, and hold each request for an allow or deny verdict before Claude processes it. This page walks through turning the feature on, connecting your server, and controlling enforcement. To learn what Inference hooks are and when to use them, see the [Inference hooks overview](https://platform.claude.com/docs/en/manage-claude/inference-hooks). To build the AI security server itself, see [Develop an Inference hooks integration](https://platform.claude.com/docs/en/manage-claude/inference-hooks-endpoint).
1212
1313## Before you begin
1414
from line 118
118118
119119## Circuit breaker
120120
121Sustained webhook failures attributable to your AI security server trip the circuit breaker, which stops enforcement: your server is no longer contacted, and your **Failure handling** choice applies to every inspected request. With **Block the request** selected, users in your organization are blocked until you act. When the breaker trips, administrators are also notified in the claude.ai notification center.
121Sustained webhook failures attributable to your AI security server trip the circuit breaker, which stops enforcement: your server is no longer contacted, and your **Failure handling** choice applies to every inspected request. With **Block the request** selected, users in your organization are blocked until the breaker resets. When the breaker trips, administrators are also notified in the claude.ai notification center.
122122
123123Each trip is also recorded in your organization's [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed) as an `inference_hooks_circuit_breaker_tripped` activity, so your security team or vendor can alert on trips from monitoring they already run, such as a SIEM that ingests the feed. One activity is recorded per trip, not one per affected request. Recording requires the Compliance API to be enabled for your organization; see [Set up the Compliance API](https://platform.claude.com/docs/en/manage-claude/compliance-api-access).
124124
125125To recover, fix the server, then turn **Enforce verdicts** back on to reset the breaker.
126126
127The breaker can also reset on its own. Starting 10 minutes after the trip, Anthropic tests whether your server has recovered: at most about once per minute, one request from your organization's normal traffic is sent to your server for inspection, and that request proceeds for its user whether or not your server answers. If your server responds with a valid verdict, allow or deny, the breaker resets and enforcement resumes. Any other outcome is a webhook failure: the breaker stays tripped and testing continues.
128
129Automatic recovery runs only while your Inference hooks settings are unchanged since the trip. If you change any Inference hooks setting after a trip, including rotating the signing secret, testing stops and the breaker no longer resets on its own; turn **Enforce verdicts** back on when your server is fixed. Automatic recovery applies only to trips: if you turn **Enforce verdicts** off yourself, enforcement stays off until you turn it back on.
130
127131## Rotate your signing secret
128132
129133Click **Rotate secret** under **Request signing** to replace your signing secret. Rotation is an immediate cutover: the new secret is generated and revealed once, the old secret can no longer be retrieved, and no request is ever signed with both secrets, so there is no overlap period to rely on.
from line 143
139143There are two levels of off:
140144
141145* **Enforce verdicts** off, on the Inference hooks settings page: within about a minute, prompts from your organization stop being sent to your AI security server; requests already in flight finish under the old setting. The settings page stays available, so use this to pause enforcement while you work on your AI security server.
142* **Allow for your organization** off, in **Data and privacy** settings: prompts are no longer inspected, and the Inference hooks settings become unavailable until you turn it back on. Your endpoint configuration, custom headers, and signing secret are kept either way; turning it back on forces **Enforce verdicts** off, so turn enforcement on again when you are ready.
146* **Allow for your organization** off, in **Data and privacy** settings: prompts are no longer inspected, and the Inference hooks settings become unavailable until you turn it back on. Your endpoint configuration, custom headers, and signing secret are kept either way; turning it back on forces **Enforce verdicts** off and clears a tripped circuit breaker, so turn enforcement on again when you are ready.
143147
144148## Next steps
145149
manage-claude/inference-hooks-endpoint Changed · +4 / -2 lines
from line 10
1010
1111An Inference hooks integration is an AI security server: an HTTPS service that Anthropic calls. For each governed request, your server receives a signed `POST` carrying the conversation transcript and responds with an allow or deny verdict. This page documents the protocol for building that server: the request and verdict schemas, signature verification, and the operational contract.
1212
13For turning Inference hooks on and pointing them at your endpoint, see [Configure Inference hooks](https://platform.claude.com/docs/en/manage-claude/inference-hooks-configuration). For what Inference hooks are and when to use them, see the [Inference hooks overview](https://platform.claude.com/docs/en/manage-claude/inference-hooks).
13To turn Inference hooks on and point them at your endpoint, see [Configure Inference hooks](https://platform.claude.com/docs/en/manage-claude/inference-hooks-configuration). To learn what Inference hooks are and when to use them, see the [Inference hooks overview](https://platform.claude.com/docs/en/manage-claude/inference-hooks).
1414
1515## Get a first verdict round trip
1616
from line 687
687687
688688### Circuit breaker
689689
690Sustained webhook failures attributable to your AI security server trip a circuit breaker that stops enforcement: Anthropic stops contacting your server, and failure handling applies to every request. Recovery happens on the admin side: fix the server, then have your administrator turn **Enforce verdicts** back on. See [Circuit breaker](https://platform.claude.com/docs/en/manage-claude/inference-hooks-configuration#circuit-breaker).
690Sustained webhook failures attributable to your AI security server trip a circuit breaker that stops enforcement: Anthropic stops contacting your server, and failure handling applies to every request.
691
692Starting 10 minutes after the trip, Anthropic tests whether your server has recovered: at most about once per minute, one request, carried by your organization's own traffic, is delivered to your server for inspection, signed and shaped like any other. Respond to it normally. A valid verdict, allow or deny, resets the breaker and enforcement resumes. A webhook failure leaves the breaker tripped, and testing continues. Either way, the test request itself proceeds for its user: its verdict is not enforced, and a failed test does not block it, even under **Block the request**. An administrator can also reset the breaker at any time, and administrator configuration changes stop the automatic testing; see [Circuit breaker](https://platform.claude.com/docs/en/manage-claude/inference-hooks-configuration#circuit-breaker).
691693
692694Each trip is recorded as an `inference_hooks_circuit_breaker_tripped` activity in the [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed), one activity per trip. While the breaker is tripped, no per-request Inference hooks activities are recorded, so the trip activity is the feed's only record of the tripped window.
693695
agents-and-tools/agent-skills/enterprise Changed · +1 / -1 lines
from line 46
4646
4747### Skill content scanning
4848
49Claude Enterprise organizations can turn on automated security scanning for custom Skills in claude.ai and Claude Cowork. The feature is in beta. After you turn on **Skill and plugin security scanning** at [claude.ai > Organization settings > Skills](https://claude.ai/admin-settings/skills), Skills that members then upload or edit in claude.ai or Cowork are scanned for signs of malicious behavior, such as hidden code execution, sending your data to an outside service, or instructions that tamper with Claude's safeguards. A Skill that fails the scan, or whose scan hasn't finished, is blocked from use. A Skill that passes with a warning stays usable behind a caution notice. If scanning is available to your organization, turn it on. It complements, but doesn't replace, the [review checklist](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/enterprise#review-checklist).
49Claude Enterprise organizations can turn on automated security scanning for custom Skills in claude.ai and Claude Cowork. After you turn on **Skill and plugin security scanning** at [claude.ai > Organization settings > Skills](https://claude.ai/admin-settings/skills), Skills that members then upload or edit in claude.ai or Cowork are scanned for signs of malicious behavior, such as hidden code execution, sending your data to an outside service, or instructions that tamper with Claude's safeguards. A Skill that fails the scan, or whose scan hasn't finished, is blocked from use. A Skill that passes with a warning stays usable behind a caution notice. If scanning is available to your organization, turn it on. It complements, but doesn't replace, the [review checklist](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/enterprise#review-checklist).
5050
5151Scanning doesn't cover the Claude API. Skills you upload through the Skills API (`/v1/skills`), including from the Claude Console, aren't scanned, so for API deployments, rely on the review checklist and [version pinning](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/enterprise#versioning-strategy). Scanning also doesn't apply to Skills that were already in your organization when you turned it on, or to organizations with certain data handling configurations, such as customer-managed encryption keys (CMEK), zero data retention (ZDR), or HIPAA readiness. For setup steps, exclusions, and result types, see [Get started with skill and plugin scanning](https://support.claude.com/en/articles/15927065-get-started-with-skill-and-plugin-scanning) in the Claude Help Center.
5252
agents-and-tools/agent-skills/overview Changed · +1 / -1 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111## Why use Skills
agents-and-tools/agent-skills/quickstart Changed · +1 / -1 lines
from line 342
342342
343343The request includes the following parts:
344344
345* **`model`:** A [model that supports the code execution tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#model-compatibility)
345* **`model`:** A [model that supports the code execution tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool#compatibility)
346346* **`container.skills`:** Specifies which Skills Claude can use
347347* **`type: "anthropic"`:** Indicates this is an Anthropic-managed Skill
348348* **`skill_id: "pptx"`:** The PowerPoint Skill identifier
agents-and-tools/mcp-connector Changed · +1 / -1 lines
from line 363
363363| `enabled` | boolean | `true` | Whether this tool is enabled. |
364364| `defer_loading` | boolean | `false` | If true, tool description is not sent to the model initially. Used with [Tool search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool). |
365365
366For the full directory of Anthropic-provided tools and optional properties such as `defer_loading`, see the [Tool reference](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-reference). For searching across large tool sets, see [Tool search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool).
366For the full directory of Anthropic-provided tools and optional properties such as `defer_loading`, see the [Tool reference](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-reference). To search across large tool sets, see [Tool search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool).
367367
368368### Configuration merging
369369
agents-and-tools/tool-use/advisor-tool Changed · +1 / -1 lines
from line 24
2424```
2525
2626<Note>
27 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
27 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
2828</Note>
2929
3030## When to use it
agents-and-tools/tool-use/bash-tool Changed · +1 / -1 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111The bash tool is a [client tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/how-tool-use-works): Claude doesn't run commands itself. When you include the tool in a request, Claude replies with a `tool_use` block that names the command to run. Your application runs that command in a bash session it owns and returns the output in a `tool_result` block.
agents-and-tools/tool-use/computer-use-tool Changed · +1 / -1 lines
from line 865
865865
866866### Combining with thinking
867867
868For combining computer use with thinking, see [Thinking](https://platform.claude.com/docs/en/build-with-claude/thinking).
868To combine computer use with thinking, see [Thinking](https://platform.claude.com/docs/en/build-with-claude/thinking).
869869
870870<Tip>
871871 For the earlier `computer_20251124` tool, internal benchmarking on the models that use it suggests these `effort` settings:
agents-and-tools/tool-use/fine-grained-tool-streaming Changed · +1 / -1 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111Fine-grained tool streaming delivers a tool's input to your client as Claude generates it, without server-side buffering or JSON validation. Skipping the buffering step reduces the time to the first fragment of a large parameter, such as a document or a block of code, and the fragments arrive through the same [Streaming messages](https://platform.claude.com/docs/en/build-with-claude/streaming) events as standard tool use.
agents-and-tools/tool-use/memory-tool Changed · +1 / -1 lines
from line 15
1515</Note>
1616
1717<Note>
18 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
18 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
1919</Note>
2020
2121## Use cases
agents-and-tools/tool-use/overview Changed · +1 / -1 lines
from line 733
733733
734734For the full conceptual model including the agentic loop and when to choose each approach, see [How tool use works](https://platform.claude.com/docs/en/agents-and-tools/tool-use/how-tool-use-works).
735735
736For connecting to Model Context Protocol (MCP) servers, see the [MCP connector](https://platform.claude.com/docs/en/agents-and-tools/mcp-connector). For building your own MCP client, see the Model Context Protocol guide to [building an MCP client](https://modelcontextprotocol.io/docs/develop/build-client).
736To connect to Model Context Protocol (MCP) servers, see the [MCP connector](https://platform.claude.com/docs/en/agents-and-tools/mcp-connector). To build your own MCP client, see the Model Context Protocol guide to [building an MCP client](https://modelcontextprotocol.io/docs/develop/build-client).
737737
738738## When Claude uses tools
739739
agents-and-tools/tool-use/text-editor-tool Changed · +1 / -1 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111Claude can use an Anthropic-schema text editor tool to view and modify text files, helping you debug, fix, and improve your code or other text documents. This allows Claude to directly interact with your files, providing hands-on assistance rather than just suggesting changes.
agents-and-tools/tool-use/tool-reference Changed · +2 / -2 lines
from line 69
6969
7070Set `defer_loading` per member, never on the entry, and give every enabled member the same value: under [tool search](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool#deferred-tool-loading) the toolset loads and expands as one definition. When every enabled member defers, only a [tool search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool) that isn't itself deferred can surface the toolset, so declare one in the same request. Don't put `cache_control` on a toolset entry whose members defer; set the breakpoint on a non-deferred tool instead, because deferred definitions are not part of the cached prefix.
7171
72`cache_control` goes on the entry only; for where the breakpoint lands, including markers inside a batch action, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching#cache-control-on-tool-definitions).
72`cache_control` goes on the entry only; to learn where the breakpoint lands, including markers inside a batch action, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching#cache-control-on-tool-definitions).
7373
7474**Handle member tool calls.** Claude calls a member with a `tool_use` block whose `name` is the member name and whose `toolset_name` is `computer` or `browser`; `input` holds that member's parameters and no `action` field. Dispatch on the `toolset_name` and `name` pair, because a custom tool may share a member's name and the two toolsets share names such as `screenshot`. Only member results echo `toolset_name`. Several member calls in one turn form a batch action that you run in order ([computer use](https://platform.claude.com/docs/en/agents-and-tools/tool-use/computer-use-tool#batch-actions), [browser use](https://platform.claude.com/docs/en/agents-and-tools/tool-use/browser-use-tool#batch-actions)). New members arrive only with a new dated `type`.
7575
from line 114
114114
115115This means `defer_loading: true` preserves your prompt cache. You can add deferred tools to a request without invalidating an existing cache entry, and the cache remains valid across the turn where the tool is discovered and the turn where it's called.
116116
117For how to combine `defer_loading` with `cache_control` breakpoints, see the [Tool search tool prompt caching guidance](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool#prompt-caching).
117To learn how to combine `defer_loading` with `cache_control` breakpoints, see the [Tool search tool prompt caching guidance](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool#prompt-caching).
118118
agents-and-tools/tool-use/tool-search-tool Changed · +2 / -2 lines
from line 24
2424</Note>
2525
2626<Note>
27 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
27 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
2828</Note>
2929
3030<Warning>
from line 784
784784
785785## Prompt caching
786786
787For how `defer_loading` preserves prompt caching, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching).
787To learn how `defer_loading` preserves prompt caching, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching).
788788
789789A tool with `defer_loading: true` can't also carry `cache_control`: the API returns a 400. Put the cache breakpoint on a non-deferred tool.
790790
agents-and-tools/tool-use/web-fetch-tool Changed · +2 / -2 lines
from line 5
55---
66
77<Note>
8 For how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
8 To learn how zero data retention (ZDR) applies to this feature, see [API and data retention](https://platform.claude.com/docs/en/manage-claude/api-and-data-retention).
99</Note>
1010
1111The web fetch tool allows Claude to retrieve full content from specified web pages and PDF documents.
from line 892
892892
893893## Prompt caching
894894
895For caching tool definitions across turns, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching).
895To cache tool definitions across turns, see [Tool use with prompt caching](https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-use-with-prompt-caching).
896896
897897## Streaming
898898