One read of Claude Developer Platformapi-20260824T210708Z

This capture is too large to show at once. Changes 76-95 of 95 are below, significant first; the rest are on the following screens.

build-with-claude/prompt-engineering/prompting-claude-fable-5 Changed · +2 / -2 lines

from line 4

44description: Behavioral differences and prompting patterns for Claude Fable 5 and Claude Mythos 5, covering effort, instruction following, long runs, memory, and scaffolding changes.

55---

66 

7This guide covers the prompting and scaffolding patterns specific to Claude Fable 5 and Claude Mythos 5. For the model's capabilities, API changes, pricing, and availability, see [Introducing Claude Fable 5 and Claude Mythos 5](https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

7This guide covers the prompting and scaffolding patterns specific to Claude Fable 5 and Claude Mythos 5. For the model's capabilities, API changes, pricing, and availability, see [Introducing Claude Fable 5 and Claude Mythos 5](https://platform.claude.com/docs/en/models/fable-5/introducing-claude-fable-5-and-claude-mythos-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

88 

99Claude Fable 5 takes on problems that were previously too complex, long-running, or ambiguous for prior models, and is particularly effective at end-to-end work that takes a person hours, days, or weeks to complete. The teams seeing the best outcomes apply Claude Fable 5 to their hardest unsolved problems; testing it only on simpler workloads tends to undersell its capability range. It also performs reliably on more straightforward tasks.

1010 

from line 11

1111Claude Fable 5 has several behavioral differences from Claude Opus 4.8 that may require prompt or scaffolding updates. Capability improvements at this level are also a good prompt to re-evaluate which instructions, tools, and guardrails are still needed. The patterns below cover the behaviors that most often require tuning.

1212 

1313<Note>

14 For API parameter changes specific to Claude Fable 5 and Claude Mythos 5 (adaptive thinking only, summarized-only thinking output, no extended thinking budgets, the `refusal` stop reason and fallback handling), see [Introducing Claude Fable 5 and Claude Mythos 5](https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5).

14 For API parameter changes specific to Claude Fable 5 and Claude Mythos 5 (adaptive thinking only, summarized-only thinking output, no extended thinking budgets, the `refusal` stop reason and fallback handling), see [Introducing Claude Fable 5 and Claude Mythos 5](https://platform.claude.com/docs/en/models/fable-5/introducing-claude-fable-5-and-claude-mythos-5).

1515 

1616 Claude Fable 5 runs safety classifiers that target offensive cybersecurity techniques (such as building exploits, malware, or attack tooling), biology and life sciences content (such as lab methods or molecular mechanisms), and extraction of the model's summarized thinking. Benign cybersecurity work and beneficial life sciences tasks may also trigger these safeguards. To re-route declined requests automatically, configure [server-side or client-side fallback](https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback) to Claude Opus 4.8.

1717</Note>

build-with-claude/prompt-engineering/prompting-claude-opus-5 Changed · +1 / -1 lines

from line 4

44description: Behavioral differences and prompting patterns for Claude Opus 5, covering response verbosity, agentic narration, task scoping, subagent delegation, self-correction, and output artifacts when thinking is disabled.

55---

66 

7This guide covers the prompting patterns specific to Claude Opus 5. For the model's capabilities and API changes, see [What's new in Claude Opus 5](https://platform.claude.com/docs/en/about-claude/models/whats-new-opus-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

7This guide covers the prompting patterns specific to Claude Opus 5. For the model's capabilities and API changes, see [What's new in Claude Opus 5](https://platform.claude.com/docs/en/models/opus-5/whats-new-opus-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

88 

99Claude Opus 5 is built for complex agentic coding and enterprise work, with particular strengths in long-horizon agentic tasks. It performs well out of the box on existing Claude Opus 4.8 prompts. The following patterns cover the behaviors that most often require tuning.

1010 

build-with-claude/prompt-engineering/prompting-claude-sonnet-5 Changed · +2 / -2 lines

from line 4

44description: Behavioral differences and prompting patterns for Claude Sonnet 5, covering effort, adaptive thinking defaults, tool use, and migration from Claude Sonnet 4.6.

55---

66 

7This guide covers the prompting patterns specific to Claude Sonnet 5. For the model's capabilities and API changes, see [What's new in Claude Sonnet 5](https://platform.claude.com/docs/en/about-claude/models/whats-new-sonnet-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

7This guide covers the prompting patterns specific to Claude Sonnet 5. For the model's capabilities and API changes, see [What's new in Claude Sonnet 5](https://platform.claude.com/docs/en/models/sonnet-5/whats-new-sonnet-5). For techniques that apply across all current Claude models, see [Prompting best practices](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices).

88 

99Claude Sonnet 5 has particular strengths in coding and agentic tasks. It performs well out of the box on existing Claude Sonnet 4.6 prompts. The patterns in this guide cover the behaviors that most often require tuning.

1010 

from line 57

5757Manual extended thinking (`thinking: {type: "enabled", budget_tokens: N}`) is not supported on Claude Sonnet 5 and returns a 400 error. It was deprecated on Claude Sonnet 4.6 and is now removed. Use adaptive thinking with the effort parameter instead.

5858 

5959<Note>

60 If you are running Claude Sonnet 5 at `high`, `xhigh`, or `max` effort, leave headroom in `max_tokens` so the model has room for thinking and tool calls. On long tasks, adaptive thinking can use a large share of the budget; if the budget is tight, you may see a response that is almost entirely thinking followed by a truncated answer and `stop_reason: "max_tokens"`. Raising `max_tokens` or dropping to `medium` effort resolves this. Because Claude Sonnet 5 uses a [new tokenizer](https://platform.claude.com/docs/en/about-claude/models/whats-new-sonnet-5#new-tokenizer) that produces approximately 30% more tokens for the same text, `max_tokens` limits tuned for Claude Sonnet 4.6 may truncate equivalent output. The exact increase depends on the content and workload shape.

60 If you are running Claude Sonnet 5 at `high`, `xhigh`, or `max` effort, leave headroom in `max_tokens` so the model has room for thinking and tool calls. On long tasks, adaptive thinking can use a large share of the budget; if the budget is tight, you may see a response that is almost entirely thinking followed by a truncated answer and `stop_reason: "max_tokens"`. Raising `max_tokens` or dropping to `medium` effort resolves this. Because Claude Sonnet 5 uses a [new tokenizer](https://platform.claude.com/docs/en/models/sonnet-5/whats-new-sonnet-5#new-tokenizer) that produces approximately 30% more tokens for the same text, `max_tokens` limits tuned for Claude Sonnet 4.6 may truncate equivalent output. The exact increase depends on the content and workload shape.

6161</Note>

6262 

6363## Tool use triggering

build-with-claude/search-results Changed · +1 / -1 lines

from line 10

1010 

1111Search result content blocks let Claude cite your own content the same way it cites web search results: each citation carries the source and title you provided. Use them in RAG (Retrieval-Augmented Generation) applications where Claude needs to attribute answers to your documents.

1212 

13All [active models](https://platform.claude.com/docs/en/about-claude/models/overview) support search results with citations, with the exception of Claude Haiku 3. No beta header is required: search results are part of the standard Messages API.

13All [active models](https://platform.claude.com/docs/en/models/overview) support search results with citations, with the exception of Claude Haiku 3. No beta header is required: search results are part of the standard Messages API.

1414 

1515## How it works

1616 

build-with-claude/thinking Changed · +2 / -2 lines

from line 46

4646 

4747On current models, thinking is on by default or one parameter away. Which configuration each model accepts, and what it defaults to, is listed in the [per-model configuration table](https://platform.claude.com/docs/en/build-with-claude/thinking-troubleshooting#supported-models) on the Troubleshooting page.

4848 

49On Claude Opus 5, Claude Sonnet 5, Claude Fable 5, Claude Mythos 5, and Claude Mythos Preview, thinking is already on: no configuration needed. The first thing most developers need on these models is to see the thinking text, because `display` defaults to `"omitted"` there. Opt in with `thinking: {"type": "adaptive", "display": "summarized"}`, which is exactly the following request with the [model string](https://platform.claude.com/docs/en/about-claude/models/overview) swapped.

49On Claude Opus 5, Claude Sonnet 5, Claude Fable 5, Claude Mythos 5, and Claude Mythos Preview, thinking is already on: no configuration needed. The first thing most developers need on these models is to see the thinking text, because `display` defaults to `"omitted"` there. Opt in with `thinking: {"type": "adaptive", "display": "summarized"}`, which is exactly the following request with the [model string](https://platform.claude.com/docs/en/models/overview) swapped.

5050 

5151On Claude Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6, thinking is off until you set `thinking: {type: "adaptive"}`, which lets Claude decide when and how deeply to think based on the request. The following examples do that, set `display: "summarized"` so the thinking text is visible, and use a roomy `max_tokens`:

5252 

from line 1019

10191019 

10201020**Response prefill and forced tool use.** You can't pre-fill the assistant response while thinking is on. Forced tool use (`tool_choice: {"type": "any"}` or `{"type": "tool", ...}`) is incompatible with manual extended thinking but works with adaptive thinking. See [Thinking with tool use](https://platform.claude.com/docs/en/build-with-claude/thinking#thinking-with-tool-use).

10211021 

1022**Output limits.** Claude Fable 5, Claude Mythos 5, Claude Mythos Preview, Claude Opus 5, Claude Opus 4.8, Claude Opus 4.7, Claude Sonnet 5, Claude Opus 4.6, and Claude Sonnet 4.6 support up to 128k output tokens per request. Claude Haiku 4.5, Claude Sonnet 4.5, and Claude Opus 4.5 support up to 64k. On the [Message Batches API](https://platform.claude.com/docs/en/build-with-claude/batch-processing#extended-output-beta), the `output-300k-2026-03-24` [beta header](https://platform.claude.com/docs/en/api/beta-headers) raises the limit to 300k for Claude Opus 5, Claude Opus 4.8, Claude Opus 4.7, Claude Sonnet 5, Claude Opus 4.6, and Claude Sonnet 4.6. See the [models overview](https://platform.claude.com/docs/en/about-claude/models/overview) for limits on legacy models.

1022**Output limits.** Claude Fable 5, Claude Mythos 5, Claude Mythos Preview, Claude Opus 5, Claude Opus 4.8, Claude Opus 4.7, Claude Sonnet 5, Claude Opus 4.6, and Claude Sonnet 4.6 support up to 128k output tokens per request. Claude Haiku 4.5, Claude Sonnet 4.5, and Claude Opus 4.5 support up to 64k. On the [Message Batches API](https://platform.claude.com/docs/en/build-with-claude/batch-processing#extended-output-beta), the `output-300k-2026-03-24` [beta header](https://platform.claude.com/docs/en/api/beta-headers) raises the limit to 300k for Claude Opus 5, Claude Opus 4.8, Claude Opus 4.7, Claude Sonnet 5, Claude Opus 4.6, and Claude Sonnet 4.6. See each model's page under [Models](https://platform.claude.com/docs/en/models/overview) for its limits.

10231023 

10241024**Long requests.** The SDKs require streaming when `max_tokens` is greater than 21,333, to avoid HTTP timeouts on long-running requests. This is a client-side validation, not an API restriction. If you don't need to process events incrementally, use `.stream()` with `.get_final_message()` (Python) or `.finalMessage()` (TypeScript) to get the complete `Message` object without handling individual events. See [Streaming Messages](https://platform.claude.com/docs/en/build-with-claude/streaming#get-the-final-message-without-handling-events). Expect longer response times when thinking is active, because generating thinking blocks adds processing time. For workloads that push thinking above roughly 32k tokens per request, use [batch processing](https://platform.claude.com/docs/en/build-with-claude/batch-processing) to avoid networking issues: such requests can run long enough to hit system timeouts and open connection limits.

10251025 

build-with-claude/token-counting Changed · +1 / -1 lines

from line 28

2828 

2929### Supported models

3030 

31All [active models](https://platform.claude.com/docs/en/about-claude/models/overview) support token counting, including Claude Opus 5 and Claude Sonnet 5.

31All [active models](https://platform.claude.com/docs/en/models/overview) support token counting, including Claude Opus 5 and Claude Sonnet 5.

3232 

3333<Note>

3434 Claude 4.7 and later models and Claude Mythos Preview use a newer tokenizer. The same input text produces approximately 30 percent more tokens than on earlier models. The exact increase depends on the content and workload shape. Recount prompts against the model you plan to use rather than reusing counts measured against earlier models.

cli-sdks-libraries/libraries/apple-foundation-models Changed · +1 / -1 lines

from line 71

7171ClaudeLanguageModel(name: .opus5, auth: auth)

7272```

7373 

74Constants mirror API model IDs (`.opus5` is `claude-opus-5`) and carry each model's capabilities. New models ship as new constants in package releases; check `ClaudeModel` in Xcode for the current list, and the [Models overview](https://platform.claude.com/docs/en/about-claude/models/overview) to compare models.

74Constants mirror API model IDs (`.opus5` is `claude-opus-5`) and carry each model's capabilities. New models ship as new constants in package releases; check `ClaudeModel` in Xcode for the current list, and the [Models overview](https://platform.claude.com/docs/en/models/overview) to compare models.

7575 

7676### Capabilities

7777 

cli-sdks-libraries/libraries/openai-sdk Changed · +1 / -1 lines

from line 26

2626 

2727 * Update your base URL to point to the Claude API

2828 * Replace your API key with a [Claude API key](https://platform.claude.com/settings/keys)

29 * Update your model name to use a [Claude model](https://platform.claude.com/docs/en/about-claude/models/overview)

29 * Update your model name to use a [Claude model](https://platform.claude.com/docs/en/models/overview)

3030 

31313. Review the following sections for what features are supported

3232 

get-started Changed · +1 / -1 lines

from line 661

661661Once you're comfortable with the basics, explore further:

662662 

663663<CardGroup cols={2}>

664 <Card title="Models overview" icon="brain" href="https://platform.claude.com/docs/en/about-claude/models/overview">

664 <Card title="Models overview" icon="brain" href="https://platform.claude.com/docs/en/models/overview">

665665 Compare Claude models by capability and cost.

666666 </Card>

667667 

intro Changed · +1 / -1 lines

from line 57

5757 <Step title="Choose the right model">

5858 Compare Claude models by capability and cost to pick the best fit for your use case.

5959 

60 [See the models overview](https://platform.claude.com/docs/en/about-claude/models/overview)

60 [See the models overview](https://platform.claude.com/docs/en/models/overview)

6161 </Step>

6262 

6363 <Step title="Explore features and tools">

manage-claude/cmek Changed · +1 / -1 lines

from line 86

8686**Claude Enterprise**

8787 

8888* Conversation history search is disabled. Conversation titles are encrypted, so searching by title or content returns no results.

89* Search across large numbers of files is slower.

89* [Project knowledge search](https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projects) (retrieval-augmented generation, or RAG) is disabled. Project knowledge loads directly into each conversation's context instead of being indexed and searched. As a result, a project can use substantially less knowledge than it could without CMEK. Knowledge beyond what can be loaded is left out of the conversation.

9090* Certain analytics are degraded: admin analytics for claude.ai skills and connectors (under claude.ai/analytics/usage and through the [Claude Enterprise Analytics API](https://platform.claude.com/docs/en/manage-claude/analytics-api)), Claude smart reports (under claude.ai/analytics/insights), and Claude Code contribution metrics (under claude.ai/analytics/claude-code).

9191* Audit log exports are disabled.

9292* Signed URLs for temporary file exchanges are disabled. These back organization data exports in claude.ai and Claude Code Remote file flows such as screenshot updates.

manage-claude/compliance-activity-feed Changed · +1 / -1 lines

from line 14

1414 Both Compliance Access Keys (`sk-ant-api01-...`) carrying this scope and Admin API keys (`sk-ant-admin01-...`) can call the Activity Feed. See [Set up the Compliance API](https://platform.claude.com/docs/en/manage-claude/compliance-api-access) for the conditions under which each key type carries the scope.

1515</Check>

1616 

17The Activity Feed records authentication, chat, file, project, administrative, and platform activity across your organization and returns it in reverse chronological order. Activities are queryable within 1 minute of occurring and are retained for 6 years.

17The Activity Feed records authentication, chat, file, project, administrative, and platform activity across your organization and returns it in reverse chronological order. Activities are queryable within 1 minute of occurring and are retained for 6 years. Recording is not retroactive: it begins when the Compliance API is first enabled for your organization, and activity from before enablement is not backfilled.

1818 

1919```bash cURL

2020curl --fail-with-body -sS \

manage-claude/compliance-api Changed · +1 / -1 lines

from line 74

7474 

7575### OpenTelemetry logging

7676 

77[Cowork's OpenTelemetry logging](https://support.claude.com/en/articles/14477985-monitor-claude-cowork-activity-with-opentelemetry) and [Claude Code monitoring](https://code.claude.com/docs/en/monitoring-usage) stream per-event telemetry, including token, cost, and host metadata, to a collector you run as activity happens, whereas the Compliance API returns retained per-session transcripts from Anthropic on request and works with your existing Compliance Access Key. For a side-by-side comparison, see the [Compliance API FAQ](https://platform.claude.com/docs/en/manage-claude/compliance-faq#data-coverage-and-retention).

77[Cowork's OpenTelemetry logging](https://support.claude.com/en/articles/14477985-monitor-claude-cowork-activity-with-opentelemetry) and [Claude Code monitoring](https://code.claude.com/docs/en/monitoring-usage) stream per-event telemetry, including token, cost, and host metadata, to a collector you run as activity happens, whereas the Compliance API returns retained per-session transcripts from Anthropic on request and works with your existing Compliance Access Key. OpenTelemetry logging can also capture prompts and responses, but Anthropic recommends the Compliance API for retrieving the content of Cowork and Claude Code sessions. For a table comparing local sessions, remote sessions, and OpenTelemetry logging, see the introduction to [Retrieve session transcripts](https://platform.claude.com/docs/en/manage-claude/compliance-sessions).

7878 

7979### Inference hooks

8080 

manage-claude/compliance-content-data Changed · +1 / -1 lines

from line 66

6666 

6767A few constraints apply to these organization-wide queries. Cursors are opaque and bound to the sort key, so an `after_id` issued under one `order_by` value is rejected with a 400 error under the other. Time-filter bounds must match the sort key too: pair `updated_at.*` bounds with `order_by=updated_at`, and `created_at.*` bounds with the default `order_by=created_at`. Backward pagination with `before_id` is not supported, and the `project_ids[]` filter is not available. See [List chats](https://platform.claude.com/docs/en/api/compliance/apps/chats/list) for the full filter reference.

6868 

69To scope the list to specific users instead (for example, a legal hold on named custodians), pass 1–10 `user_ids[]` values. Obtain the IDs from [List organization users](https://platform.claude.com/docs/en/manage-claude/compliance-org-data#list-organization-users). User-filtered queries always sort by `created_at` (passing `order_by=updated_at` returns a 400 error) and support both `after_id` and `before_id`. Filtering by `project_ids[]` is only available in this user-filtered form.

69To scope the list to specific users instead (for example, a legal hold on named custodians), pass 1–10 `user_ids[]` values. Obtain the IDs from [List organization users](https://platform.claude.com/docs/en/manage-claude/compliance-org-data#list-organization-users). User-filtered queries always sort by `created_at` (passing `order_by=updated_at` returns a 400 error) and support both `after_id` and `before_id`. Filtering by `project_ids[]` is only available in this user-filtered form. Combining `user_ids[]` with any `updated_at.*` bound is deprecated and will be rejected with a 400 error after 2026-09-22; to keep a custodian set current by update time, run the org-wide `order_by=updated_at` walk without `user_ids[]` and select the custodians' chats from its results, and keep the user-filtered listing for `created_at`-ordered exports.

7070 

7171```bash cURL

7272curl --fail-with-body -sS -G \

manage-claude/compliance-errors Changed · +1 / -1 lines

from line 122

122122 

123123## 403 Forbidden

124124 

125The key in `x-api-key` is valid but does not carry the scope the endpoint requires. The verbatim message lists the scopes the key carries (`Got:`) and the scopes the endpoint requires (`Needed:`), so you can confirm what the key carries without rechecking Claude Console or claude.ai. Compliance Access Key scopes are immutable after creation, so each insufficient-scope fix directs you to create a new key rather than edit the existing one.

125The key in `x-api-key` is valid but does not carry the scope the endpoint requires. The verbatim message lists the scopes the key carries (`Got:`) and the scopes the endpoint requires (`Needed:`), so you can confirm what the key carries without rechecking Claude Console or claude.ai. Compliance Access Key scopes are immutable after creation, so each insufficient-scope fix directs you to create a new key rather than edit the existing one. A standalone Claude Console organization (one with no parent organization) cannot create a Compliance Access Key, so fixes that require one do not apply to it; it can query the Activity Feed only.

126126 

127127### Insufficient scope: Activity Feed

128128 

manage-claude/compliance-integration-patterns Changed · +1 / -1 lines

from line 26

2626 

2727Both patterns share these constraints:

2828 

29* Activities are queryable within 1 minute of occurring and retained for 6 years.

29* Activities are queryable within 1 minute of occurring and retained for 6 years. Recording is not retroactive: it begins when the Compliance API is first enabled for your organization, and activity from before enablement is not backfilled.

3030* The maximum `limit` for each page is 5,000.

3131* Cursor values are opaque strings that you must not parse.

3232* Requests are limited to 600 per minute per [parent organization](https://platform.claude.com/docs/en/manage-claude/compliance-api#how-the-compliance-api-works), shared across every key, every linked organization, and every `/v1/compliance/*` endpoint; unlike the local session endpoints, the remote session endpoints carry a second request budget on top. See [429 Too Many Requests](https://platform.claude.com/docs/en/manage-claude/compliance-errors#429-too-many-requests) for the response headers and retry contract.

manage-claude/inference-hooks Changed · +1 / -1 lines

from line 29

2929 

3030A verdict is a small JSON object: `{"action": "allow"}` lets the request proceed, and a deny carries the user-facing reason. For the full verdict schema, see [Return a verdict](https://platform.claude.com/docs/en/manage-claude/inference-hooks-endpoint#return-a-verdict).

3131 

32Your AI security server sees what the user sees: transcript text, tool calls and their results, and text extracted from attachments. It never receives raw file or image bytes, system prompts, or Anthropic-internal context.

32Your AI security server sees what the user sees: transcript text, tool calls and their results, and text extracted from attachments. It never receives raw file or image bytes, system prompts, or Anthropic-internal context. Anthropic doesn't store prompt or response content as part of Inference hooks; it records only metadata about hook activity, such as verdicts, timestamps, and request identifiers.

3333 

3434If your AI security server is unreachable, returns an error, or doesn't respond within the timeout, your organization's failure handling setting decides the outcome: block the request, or allow it to proceed without inspection.

3535 

manage-claude/inference-hooks-configuration Changed · +3 / -1 lines

from line 120

120120 

121121Sustained 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.

122122 

123Each 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).

124 

123125To recover, fix the server, then turn **Enforce verdicts** back on to reset the breaker.

124126 

125127## Rotate your signing secret

from line 132

130132 

131133## Audit trail

132134 

133Inference hooks activity is recorded in your organization's [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed): configuration changes, denials, and requests that proceeded without inspection under your failure handling setting. Denial records carry identifiers that let you join each denial to the matching record in your own system.

135Inference hooks activity is recorded in your organization's [Activity Feed](https://platform.claude.com/docs/en/manage-claude/compliance-activity-feed): configuration changes, denials, circuit breaker trips, and requests that proceeded without inspection under your failure handling setting. While the circuit breaker is tripped, no per-request Inference hooks activities are recorded; the trip activity is the feed's record of that window. Denial records carry identifiers that let you join each denial to the matching record in your own system.

134136 

135137## Turn Inference hooks off

136138 

manage-claude/inference-hooks-endpoint Changed · +2 / -0 lines

from line 689

689689 

690690Sustained 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).

691691 

692Each 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.

693 

692694### Latency

693695 

694696Enforcement adds your AI security server's round trip to the latency of every governed request in your organization. Keep the verdict fast, and load-test your server before rolling it out to a large organization.

test-and-evaluate/strengthen-guardrails/reduce-latency Changed · +2 / -2 lines

from line 27

2727 

2828### 1. Choose the right model

2929 

30One of the most direct ways to reduce latency is to select the appropriate model for your use case. Anthropic offers a [range of models](https://platform.claude.com/docs/en/about-claude/models/overview) with different capabilities and performance characteristics. Consider your specific requirements and choose the model that best fits your needs in terms of speed and output quality.

30One of the most direct ways to reduce latency is to select the appropriate model for your use case. Anthropic offers a [range of models](https://platform.claude.com/docs/en/models/overview) with different capabilities and performance characteristics. Consider your specific requirements and choose the model that best fits your needs in terms of speed and output quality.

3131 

3232For speed-critical applications, **Claude Haiku 4.5** offers the fastest response times while maintaining high intelligence:

3333 

from line 164

164164 ```

165165</CodeGroup>

166166 

167For more details about model metrics, see the [models overview](https://platform.claude.com/docs/en/about-claude/models/overview) page.

167For more details about model metrics, see the [models overview](https://platform.claude.com/docs/en/models/overview) page.

168168 

169169### 2. Optimize prompt and output length

170170