HomeDocs changes
Documentation changes
What the docs quietly changed.
Anthropic's published documentation, fetched on a schedule and diffed against the last copy. Every edit dated, including the ones that never made a release note.
Last poll read all 4 cleanly: Claude Code CLI, Claude Developer Platform, Claude Documentation, Model Context Protocol. A date here is when this site noticed, not when Anthropic edited.
Claude Developer Platform
812pages tracked
765 changed this week · last read 23 Oct 2026
Claude Code CLI
228pages tracked
145 changed this week · last read 22 Oct 2026
Claude Documentation
280pages tracked
77 changed this week · last read 23 Oct 2026
Model Context Protocol
357pages tracked
171 changed this week · last read 22 Oct 2026
100 on this page
ReadSourcePageWhatLines
21 Oct 2026
617 pages
+38537 −9704 lines
claude-code
keybindings
±changed
+21−0
claude-codechanged
keybindings
### Cancel a question with one key press
To cancel a question with one key press in vim mode, bind another key to `select:cancel` in the `Select` context. You can cancel this way from the `Other` row of a question where y
Pick a key that you press with Ctrl or Alt. A bare letter or digit [goes to the `Other` row](#text-fields) instead and doesn't reach your bindings. For example, to bind `Ctrl+Q`, a
+21 −0 lines
Upstream edited 21 Oct 2026 20:15 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
interactive-mode
±changed
+15−0
claude-codechanged
interactive-mode
### Cancel a question in vim mode
When you're typing your own answer in the `Other` row of a [question Claude asks](/docs/en/tools-reference#askuserquestion-tool-behavior), press `Esc` twice to cancel the question.
To cancel with a single key press instead, [bind another key to cancel the question](/docs/en/keybindings#cancel-a-question-with-one-key-press).
+15 −0 lines
Upstream edited 21 Oct 2026 20:15 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
fullscreen
±changed
+1−0
claude-codechanged
fullscreen
* A `--chrome` or `--no-chrome` flag you launched with, so [Chrome integration](/docs/en/chrome) stays on or off as you chose for this session
+1 −0 lines
Upstream edited 21 Oct 2026 20:15 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
agent-sdk/typescript
±changed
+152−152
claude-codechanged
agent-sdk/typescript
| `extras` | `{ annotations?: `[`ToolAnnotations`](#toolannotations)`; searchHint?: string; alwaysLoad?: boolean }` | Optional extras. `annotations` provides MCP behavioral hints t
| `extras` | `{ annotations?: `[`ToolAnnotations`](#toolannotations)`; searchHint?: string; alwaysLoad?: boolean }` | Optional extras. `annotations` provides MCP behavioral hints t
| `options.alwaysLoad` | `boolean` | When `true`, this server's tools stay in the initial prompt instead of being deferred behind [tool search](/docs/en/agent-sdk/tool-search). Com
| `options.alwaysLoad` | `boolean` | When `true`, this server's tools stay in the initial prompt instead of being deferred behind [tool search](/docs/en/agent-sdk/tool-search.md).
| `parent_agent_id` | `string \| null` | For messages from a [nested subagent](/docs/en/sub-agents#let-subagents-spawn-their-own-subagents), the `agentId` of the subagent that spaw
+152 −152 lines
Upstream edited 21 Oct 2026 20:15 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-code
memory
±changed
+36−36
claude-codechanged
memory
Claude Code has two complementary memory systems. Both are loaded at the start of every conversation. Claude treats them as context, not enforced configuration. To block an action
Claude Code has two complementary memory systems. Both are loaded at the start of every conversation. Claude treats them as context, not enforced configuration. To block an action
Subagents can also maintain their own auto memory. See [subagent configuration](/docs/en/sub-agents#enable-persistent-memory) for details.
Subagents can also maintain their own auto memory. See [subagent configuration](/docs/en/sub-agents.md#enable-persistent-memory) for details.
Keep it to facts Claude should hold in every session: build commands, conventions, project layout, "always do X" rules. If an entry is a multi-step procedure or only matters for on
+36 −36 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-code
large-codebases
±changed
+41−41
claude-codechanged
large-codebases
> For workflow techniques that keep context small in any repository, such as [running exploration in a subagent](/docs/en/best-practices#use-subagents-for-investigation) so file re
> For workflow techniques that keep context small in any repository, such as [running exploration in a subagent](/docs/en/best-practices.md#use-subagents-for-investigation) so file
Project settings in `.claude/settings.json` aren't inherited from parent directories the way CLAUDE.md files are. For which directory's `.claude/settings.json` a session reads, see
Project settings in `.claude/settings.json` aren't inherited from parent directories the way CLAUDE.md files are. For which directory's `.claude/settings.json` a session reads, see
Claude Code loads every [CLAUDE.md](/docs/en/memory) file from your working directory and every parent directory at launch, then loads each subdirectory's file [on demand](/docs/en
+41 −41 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-code
features-overview
±changed
+76−76
claude-codechanged
features-overview
Claude Code combines a model that reasons about your code with [built-in tools](/docs/en/how-claude-code-works#tools) for file operations, search, execution, and web access. The bu
Claude Code combines a model that reasons about your code with [built-in tools](/docs/en/how-claude-code-works.md#tools) for file operations, search, execution, and web access. The
> For how the core agentic loop works, see [How Claude Code works](/docs/en/how-claude-code-works).
**New to Claude Code?** Start with [CLAUDE.md](/docs/en/memory) for project conventions, then add other extensions [as specific triggers come up](#build-your-setup-over-time).
+76 −76 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-code
agent-teams
±changed
+47−45
claude-codechanged
agent-teams
> Agent teams are experimental and disabled by default. Enable them by setting `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` in your [settings.json](/docs/en/settings) or environment. W
> Agent teams are experimental and disabled by default. Enable them by setting `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` in your [settings.json](/docs/en/settings.md) or environment
Before you set up a team, check whether a lighter option does the job. [Subagents](/docs/en/sub-agents) work within a single session, and with [cross-session messaging](/docs/en/cr
Before you set up a team, check whether a lighter option does the job. [Subagents](/docs/en/sub-agents.md) work within a single session, and with [cross-session messaging](/docs/en
Agent teams add coordination overhead and use significantly more tokens than a single session. They work best when teammates can operate independently. For sequential tasks, same-f
+47 −45 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-code
memory
±changed
+7−0
claude-codechanged
memory
Each pattern is relative to the rule's base directory:
* **Project rules**: the directory that contains the rule's `.claude` folder
* **[User-level rules](#user-level-rules)**: the session's [primary working directory](/docs/en/permissions#working-directories)
+7 −0 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
large-codebases
±changed
+1−1
claude-codechanged
large-codebases
You can also scope a skill by file pattern instead of by placement. The [`paths` frontmatter field](/docs/en/skills#frontmatter-reference) takes glob patterns, and Claude loads the
You can also scope a skill by file pattern instead of by placement. The [`paths` frontmatter field](/docs/en/skills#frontmatter-reference) takes glob patterns, and Claude loads the
+1 −1 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
features-overview
±changed
+1−1
claude-codechanged
features-overview
**Use a subagent** when you need a quick, focused worker: research a question, verify a claim, review a file. The subagent does the work and returns a summary, so your main convers
**Use a subagent** when you need a quick, focused worker: research a question, verify a claim, review a file. The subagent does the work and returns a summary, so your main convers
+1 −1 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-code
agent-teams
±changed
+1−1
claude-codechanged
agent-teams
| **Communication** | Return a result to the caller. Subagents that Claude named when it spawned them can also [message each other](/docs/en/sub-agents#what-loads-at-startup) | Tea
| **Communication** | Return a result to the caller. Subagents that Claude [named](/docs/en/sub-agents#subagent-names) when it spawned them can also [message each other](/docs/en/s
+1 −1 lines
Upstream edited 21 Oct 2026 20:08 UTC Recorded here 21 Oct 2026 20:37 UTC
claude-docs
cowork/monitoring
±changed
+6−6
claude-docschanged
cowork/monitoring
Cowork exports the following events to your OTel collector. The collector settings in [Setup](#setup) apply to every Cowork session that runs on a user's computer in Claude Desktop
Cowork exports the following events to your OTel collector. The collector settings in [Setup](#setup) apply to every Cowork session that runs on a user's computer in Claude Desktop
To set the key, deploy it with your device management tool to the [managed configuration location for each operating system](/docs/third-party/claude-desktop/configuration#how-keys
To set the key, deploy it with your device management tool to the [managed configuration location for each operating system](/docs/third-party/claude-desktop/configuration.md#how-k
| Set to a list of [categories](/docs/third-party/claude-desktop/telemetry#content-capture) | By default, the content in the listed categories, and model response text when `userPr
+6 −6 lines
Upstream edited 21 Oct 2026 19:50 UTC Recorded here 22 Oct 2026 18:07 UTC
claude-docs
cowork/monitoring
±changed
+2−0
claude-docschanged
cowork/monitoring
See [Retrieve session transcripts](https://platform.claude.com/docs/en/manage-claude/compliance-sessions) in the Compliance API documentation to retrieve the transcripts of Cowork
+2 −0 lines
Upstream edited 21 Oct 2026 19:50 UTC Recorded here 21 Oct 2026 20:07 UTC
api
test-and-evaluate/strengthen-guardrails/reduce-prompt-leak
±changed
+5−5
apichanged
test-and-evaluate/strengthen-guardrails/reduce-prompt-leak
Prompt leaks can expose sensitive information that you expect to be "hidden" in your prompt. While no method is foolproof, the strategies below can significantly reduce the risk.
Prompt leaks can expose sensitive information that you expect to be "hidden" in your prompt. Although no method is foolproof, the following strategies can significantly reduce the
Consider using leak-resistant prompt engineering strategies only when **absolutely necessary**. Attempts to leak-proof your prompt can add complexity that may degrade performance i
Consider using leak-resistant prompt engineering strategies only when **absolutely necessary**. Attempts to leak-proof your prompt can make the LLM’s overall task more complex, whi
Try monitoring techniques first, like output screening and post-processing, to try to catch instances of prompt leak.
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/strengthen-guardrails/reduce-latency
±changed
+5−5
apichanged
test-and-evaluate/strengthen-guardrails/reduce-latency
description: Reduce Claude's response latency by choosing a faster model like Claude Haiku 5.5, trimming prompt and output tokens, and streaming responses.
description: Reduce Claude's response latency by choosing a faster model such as Claude Haiku 5.5, trimming prompt and output tokens, and streaming responses.
For a more in-depth understanding of these terms, check out the [glossary](https://platform.claude.com/docs/en/about-claude/glossary).
For a more in-depth understanding of these terms, see the [glossary](https://platform.claude.com/docs/en/about-claude/glossary).
One 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/
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/strengthen-guardrails/reduce-hallucinations
±changed
+9−9
apichanged
test-and-evaluate/strengthen-guardrails/reduce-hallucinations
Even the most advanced language models, like Claude, can sometimes generate text that is factually incorrect or inconsistent with the given context. This phenomenon, known as "hall
Even the most advanced language models, such as Claude, can sometimes generate text that is factually incorrect or inconsistent with the given context. This phenomenon, known as "h
* **Allow Claude to say "I don't know":** Explicitly give Claude permission to admit uncertainty. This simple technique can drastically reduce false information.
* **Allow Claude to say "I don't know":** Explicitly give Claude permission to admit uncertainty. This technique can drastically reduce false information.
<Accordion title="Example: Analyzing a merger & acquisition report">
+9 −9 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/strengthen-guardrails/mitigate-jailbreaks
±changed
+6−6
apichanged
test-and-evaluate/strengthen-guardrails/mitigate-jailbreaks
Jailbreaking and prompt injection are attempts to make Claude ignore its guidelines or your instructions. While Claude is inherently resilient to such attacks, the additional steps
Jailbreaking and prompt injection are attempts to make Claude ignore its guidelines or your instructions. Although Claude is inherently resilient to such attacks, the additional st
* **Indirect prompt injection**, where the user is trusted but Claude processes *third-party content* (web pages, emails, documents, tool results) that contains adversarial instruc
* **Indirect prompt injection**, where the user is trusted but Claude processes *third-party content* (webpages, emails, documents, tool results) that contains adversarial instruct
* **Harmlessness screens:** Use a lightweight model like Claude Haiku 5.5 to pre-screen user input before it reaches your main conversation. Use [structured outputs](https://platfo
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/strengthen-guardrails/increase-consistency
±changed
+5−5
apichanged
test-and-evaluate/strengthen-guardrails/increase-consistency
## Specify the desired output format
Precisely define your desired output format using JSON, XML, or custom templates so that Claude follows every output formatting element you require.
## Specify the output format you want
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/strengthen-guardrails/handle-streaming-refusals
±changed
+8−8
apichanged
test-and-evaluate/strengthen-guardrails/handle-streaming-refusals
A `refusal` response from streaming classifiers includes a `stop_details` object with a `category` and a human-readable `explanation` that you can surface to the user. See [Refus
A `refusal` response from streaming classifiers includes a `stop_details` object with a `category` and a human-readable `explanation` that you can surface to the user. See [What
* **Monitor for refusals:** Include **`stop_reason`: `refusal`** checks in your error handling
* **Reset automatically:** Implement automatic context reset when refusals are detected
* **Fall back to another model:** Configure [server-side fallback or the SDK middleware](https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback) so refused req
+8 −8 lines
Recorded here 21 Oct 2026 19:37 UTC
api
test-and-evaluate/develop-tests
±changed
+10−10
apichanged
test-and-evaluate/develop-tests
* User feedback: Implicit measures like task completion rates.
* User feedback: Implicit measures such as task completion rates.
How well does the model need to perform on the task? You may also need to consider edge case handling, such as how well the model needs to perform on rare or challenging inputs
How well does the model need to perform on the task? You might also need to consider edge case handling, such as how well the model needs to perform on rare or challenging inpu
How well does the model directly address the user's questions or instructions? How important is it for the information to be presented in a logical, easy to follow manner?
+10 −10 lines
Recorded here 21 Oct 2026 19:37 UTC
api
resources/overview
±changed
+1−1
apichanged
resources/overview
Detailed documentation of Claude 3 models including latest 3.5 addendum.
Detailed documentation of Claude 3 models including the latest 3.5 addendum.
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
release-notes/overview
±changed
+2−0
apichanged
release-notes/overview
- We've added the `enforcement_status` and `enforcement_actions` fields to [user profiles](https://platform.claude.com/docs/en/manage-claude/user-profiles), so you can see whether
+2 −0 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/sonnet-5/overview
±changed
+2−2
apichanged
models/sonnet-5/overview
Effort defaults to `high` on the Claude API and Claude Code. Choose a level per workload.
Effort defaults to `high` on the Claude API and Claude Code. Choose a level for each workload.
1M tokens by default. How the window is counted and managed.
Holds 1M tokens by default. How the window is counted and managed.
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/sonnet-5-5/whats-new-sonnet-5-5
±changed
+5−5
apichanged
models/sonnet-5-5/whats-new-sonnet-5-5
`between_tools` is accepted at `low`, `medium`, and `high` effort. At `xhigh` or `max` effort, a request with `between_tools` returns a 400 error. To run at `xhigh` or `max`, use a
`between_tools` is accepted at `low`, `medium`, and `high` effort. At `xhigh` or `max` effort, a request with `between_tools` returns a 400 error. To run at `xhigh` or `max`, use a
`between_tools` takes no other field: `display`, `budget_tokens`, or `block_binding` sent with it returns a 400 error. Manual thinking budgets (`thinking: {"type": "enabled", "budg
`between_tools` takes no other field: `display`, `budget_tokens`, or `block_binding` sent with it returns a 400 error. Manual thinking budgets (`thinking: {"type": "enabled", "budg
Keep the conversation append-only so the check never fails: change instructions or tools with [mid-conversation system messages](https://platform.claude.com/docs/en/build-with-clau
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/sonnet-5-5/overview
±changed
+2−2
apichanged
models/sonnet-5-5/overview
description: "Claude Sonnet 5.5 at a glance: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for buil
description: "An overview of Claude Sonnet 5.5: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for b
The control for thinking depth, latency, and cost. Choose a level per workload.
The control for thinking depth, latency, and cost. Choose a level for each workload.
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/sonnet-5-5/migration-guide
±changed
+17−17
apichanged
models/sonnet-5-5/migration-guide
Claude Sonnet 5.5 has the same prices as Claude Sonnet 5, except for prompt cache reads, which cost $0.10 USD per million tokens, half the Claude Sonnet 5 rate. See [Claude pricing
Claude Sonnet 5.5 has the same prices as Claude Sonnet 5, except for prompt cache reads, which cost $0.10 USD per million tokens, half the Claude Sonnet 5 rate. See [Claude pricing
This request works on Claude Sonnet 5.5 as written. It sets an effort level, and the SDK tabs read the reply by block type. It leaves out five settings that return a 400 error: [th
This request works on Claude Sonnet 5.5 as written. The example sets an effort level and, in the SDK, reads the reply by block type. It leaves out five settings that return a 400 e
model: "claude-sonnet-5-5",
+17 −17 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/overview
±changed
+4−4
apichanged
models/overview
Once you've picked a model, [learn how to make your first API call](https://platform.claude.com/docs/en/get-started). To understand how model IDs, aliases, and snapshots work, see
Once you've picked a model, [learn how to make your first API call](https://platform.claude.com/docs/en/get-started). To understand how model IDs, aliases, and snapshots work, see
* **Output quality:** When migrating from a previous model generation, you may notice larger improvements in overall performance. If you're on Claude Opus 5 or earlier, see the [Cl
* **Output quality:** When migrating from a previous model generation, you might notice larger improvements in overall performance. If you're on Claude Opus 5 or earlier, see the [
If you're ready to start exploring what Claude can do for you, dive in! Whether you're a developer looking to integrate Claude into your applications or a user wanting to experienc
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/opus-5-5/whats-new-opus-5-5
±changed
+7−7
apichanged
models/opus-5-5/whats-new-opus-5-5
The [effort parameter](https://platform.claude.com/docs/en/build-with-claude/effort) is the control for thinking depth, latency, and cost: lower it where you previously disabled th
The [effort parameter](https://platform.claude.com/docs/en/build-with-claude/effort) is the control for thinking depth, latency, and cost: lower it where you previously disabled th
`tool_choice: {"type": "auto"}` (the default) and `{"type": "none"}` are supported, and the same validation applies to the [token counting](https://platform.claude.com/docs/en/buil
`tool_choice: {"type": "auto"}` (the default) and `{"type": "none"}` are supported, and the same validation applies to the [token counting](https://platform.claude.com/docs/en/buil
The API also checks whether anything before a Claude Opus 5.5 thinking block (the `system` prompt, the `tools`, or an earlier message) has changed since the block was produced. As
+7 −7 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/opus-5-5/overview
±changed
+2−2
apichanged
models/opus-5-5/overview
description: "Claude Opus 5.5 at a glance: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for buildi
description: "An overview of Claude Opus 5.5: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for bui
The control for thinking depth, latency, and cost. Choose a level per workload.
The control for thinking depth, latency, and cost. Choose a level for each workload.
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/opus-5-5/migration-guide
±changed
+15−15
apichanged
models/opus-5-5/migration-guide
Claude Opus 5.5 costs less than Claude Opus 5 ($4 / $20 USD per million input / output tokens, compared with $5 / $25; see [Claude pricing](https://platform.claude.com/docs/en/abou
Claude Opus 5.5 costs less than Claude Opus 5 ($4 / $20 USD per million input / output tokens, compared with $5 / $25 USD; see [Claude pricing](https://platform.claude.com/docs/en/
The following request satisfies every item in the list: effort is set, and there is no `thinking` field. The SDK tabs that print text select it by block type, because `thinking` bl
The following request satisfies every item in the list: effort is set, and there is no `thinking` field. Where the example prints text, it finds the text block by its type, because
model: "claude-opus-5-5",
+15 −15 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/mythos-5-1/overview
±changed
+1−1
apichanged
models/mythos-5-1/overview
description: "Claude Mythos 5.1 at a glance: the same model as Claude Fable 5.1, available only to organizations verified through Anthropic's verification programs. Model IDs, spec
description: Claude Mythos 5.1 is the same model as Claude Fable 5.1, available only to organizations verified through Anthropic's verification programs. Model IDs, specifications,
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/haiku-5-5/whats-new-haiku-5-5
±changed
+6−6
apichanged
models/haiku-5-5/whats-new-haiku-5-5
Claude Haiku 5.5 is built for high-volume, latency-sensitive work such as classification, routing, extraction, and subagent tasks. It supports adaptive thinking with the effort par
Claude Haiku 5.5 is built for high-volume, latency-sensitive work such as classification, routing, extraction, and subagent tasks. It supports adaptive thinking with the effort par
| [Responses can begin with thinking blocks](https://platform.claude.com/docs/en/models/haiku-5-5/whats-new-haiku-5-5#responses-can-begin-with-thinking-blocks)
| [Responses can start with thinking blocks](https://platform.claude.com/docs/en/models/haiku-5-5/whats-new-haiku-5-5#responses-can-begin-with-thinking-blocks)
With [adaptive thinking](https://platform.claude.com/docs/en/build-with-claude/thinking), Claude Haiku 5.5 determines when and how much to think. Adaptive thinking is on by default
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/haiku-5-5/overview
±changed
+2−2
apichanged
models/haiku-5-5/overview
description: "Claude Haiku 5.5 at a glance: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for build
description: "An overview of Claude Haiku 5.5: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and the guides and resources for bu
Claude Haiku 5.5 is built for high-volume, latency-sensitive work such as classification, routing, extraction, and subagent tasks. It supports adaptive thinking with the effort par
Claude Haiku 5.5 is built for high-volume, latency-sensitive work such as classification, routing, extraction, and subagent tasks. It supports adaptive thinking with the effort par
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/haiku-5-5/migration-guide
±changed
+6−6
apichanged
models/haiku-5-5/migration-guide
Claude Haiku 5.5 uses the same newer tokenizer as Claude 4.7 and later models. As with all models that use this tokenizer, the same input text produces approximately 30% more token
Claude Haiku 5.5 uses the same newer tokenizer as Claude 4.7 and later models. As with all models that use this tokenizer, the same input text produces approximately 30 percent mor
* A `max_tokens` limit tuned for Claude Haiku 4.5 may cut off equivalent output.
* A `max_tokens` limit tuned for Claude Haiku 4.5 might cut off equivalent output.
Adaptive thinking is on by default, so a response can begin with one or more `thinking` blocks even when the request doesn't set `thinking`. Leave `thinking` unset or set it to `{"
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/fable-5/overview
±changed
+1−1
apichanged
models/fable-5/overview
## Fable vs. Mythos
## Fable versus Mythos
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/fable-5-1/whats-new-fable-5-1
±changed
+24−19
apichanged
models/fable-5-1/whats-new-fable-5-1
Claude Fable 5.1 extends Claude Fable 5 at the same input and output prices, with cache reads at a quarter of the cost, and brings stronger long-running agentic coding, multistep r
If you already call Claude Fable 5, three changes are breaking: [forced tool use returns an error](https://platform.claude.com/docs/en/models/fable-5-1/whats-new-fable-5-1#forced-t
Claude Fable 5.1 extends Claude Fable 5 at the same input and output prices, with cache reads at a quarter of the cost, and brings stronger long-running agentic coding, multistep r
+24 −19 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/fable-5-1/overview
±changed
+3−3
apichanged
models/fable-5-1/overview
description: "Claude Fable 5.1 at a glance: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and resources for building with it."
description: "An overview of Claude Fable 5.1: what it's for, model IDs on every platform, context window, output limits, pricing, availability, and resources for building with it.
Claude Fable 5.1 extends Claude Fable 5 at the same input and output prices, with cache reads at a quarter of the cost, and brings stronger long-running agentic coding, multistep r
Claude Fable 5.1 extends Claude Fable 5 at the same input and output prices, with cache reads at a quarter of the cost, and brings stronger long-running agentic coding, multistep r
If you already call Claude Fable 5, three changes are breaking: [forced tool use returns an error](https://platform.claude.com/docs/en/models/fable-5-1/whats-new-fable-5-1#forced-t
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
models/fable-5-1/migration-guide
±changed
+23−17
apichanged
models/fable-5-1/migration-guide
* **Priority Tier:** Neither model is supported on [Priority Tier](https://platform.claude.com/docs/en/api/service-tiers#supported-models). Claude Fable 5 is.
* **Priority Tier:** Neither model is supported on [Priority Tier](https://platform.claude.com/docs/en/api/service-tiers#supported-models). Claude Fable 5 is.
* **Conversation check on thinking blocks:** [Editing earlier turns](https://platform.claude.com/docs/en/models/fable-5-1/migration-guide#fable-5-1-preserved-thinking) invalidates
Migration is mostly drop-in. The API surface, limits, per-token pricing, tokenizer, always-on adaptive thinking, refusal handling, and `stop_details` categories all match Claude Fa
Migration is mostly drop-in. The API surface, limits, per-token pricing, tokenizer, always-on adaptive thinking, refusal handling, and `stop_details` categories all match Claude Fa
+23 −17 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/workflow-runs
±changed
+9−9
apichanged
managed-agents/workflow-runs
* **Phases:** A workflow can divide its work into phases. A phase is a named stage of the run, such as "Read the contracts". You follow a run's progress by its phase events.
* **Agent threads:** In a phase, the program runs agents. Each agent works in its own [session thread](https://platform.claude.com/docs/en/managed-agents/session-threads), on a pro
* **Phases:** A workflow can divide its work into phases. A phase is a named stage of the run, such as "Read the contracts." You follow a run's progress by its phase events.
* **Agent threads:** In a phase, the program runs agents. Each agent works in its own [session thread](https://platform.claude.com/docs/en/managed-agents/session-threads), on a pro
There's no approval step before a run starts: the agent starts a run in the same turn that it writes the workflow. To review a plan first, tell the agent to describe the run and wa
+9 −9 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/webhooks
±changed
+13−13
apichanged
managed-agents/webhooks
description: Get notified when major events happen without polling.
description: Get notified when major events occur without polling.
Sessions are long-running interactions. While most real-time interactions happen through the [SSE event stream](https://platform.claude.com/docs/en/managed-agents/events-and-stream
Sessions are long-running interactions. Although most real-time interactions occur through the [SSE event stream](https://platform.claude.com/docs/en/managed-agents/events-and-stre
| `session.status_run_started` | Agent execution started. This triggers at every session status transition to `running`.
+13 −13 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/vaults
±changed
+10−10
apichanged
managed-agents/vaults
--data '{
-d '{
--data @- <<'EOF'
-d @- <<'EOF'
--data @- <<'EOF'
+10 −10 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/tools-web-restrictions
±changed
+11−11
apichanged
managed-agents/tools-web-restrictions
For how the SDKs type these entries, see [Config entry types in the SDKs](https://platform.claude.com/docs/en/managed-agents/tools#config-entry-types-in-the-sdks).
## When a domain is not permitted
`web_search` omits results that its domain list does not permit. A `web_fetch` call for a URL that its domain list does not permit returns an error result to the agent. The `agent.
+11 −11 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/tools
±changed
+14−14
apichanged
managed-agents/tools
Claude Managed Agents also supports custom, user-defined tools. Your application executes these tools separately and returns the results to Claude, which uses them to continue the
Claude Managed Agents also supports custom, user-defined tools. Your application runs these tools separately and returns the results to Claude, which uses them to continue the task
| Tool | Name | Description |
| ---------- | ------------ | ---------------------------------------------- |
| Bash | `bash` | Execute bash commands in a shell session |
+14 −14 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/skills
±changed
+3−3
apichanged
managed-agents/skills
Skills reach your agent in two ways: attach them through the agent's `skills` (csharp, go: `Skills`) array, or [load them from a GitHub repository](https://platform.claude.com/docs
Skills reach your agent in two ways: attach them through the agent's `skills` (csharp, go: `Skills`) array, or [load them from a GitHub repository](https://platform.claude.com/docs
To list, retrieve, delete, and version custom skills, see [Managing custom skills](https://platform.claude.com/docs/en/build-with-claude/skills-guide#managing-custom-skills). For t
To list, retrieve, delete, and version custom skills, see [Managing custom skills](https://platform.claude.com/docs/en/build-with-claude/skills-guide#managing-custom-skills). For t
Skills can also live in your codebase. When a session mounts a repository through the [`github_repository` resource](https://platform.claude.com/docs/en/managed-agents/github), the
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/sessions
±changed
+5−5
apichanged
managed-agents/sessions
description: Create a session to run your agent and begin executing tasks.
description: Create a session to run your agent and start running tasks.
To cap what a session can spend, pass the optional `budget` (csharp, go: `Budget`) object when you create it. A budget is a hard ceiling on the session's list cost: the platform pr
The following example creates a session with a $1.25 budget:
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/session-threads
±changed
+1−1
apichanged
managed-agents/session-threads
**Session threads** are where you drill into a specific agent's activity.
**Session threads** are where you examine a specific agent's activity.
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/session-operations
±changed
+2−2
apichanged
managed-agents/session-operations
| `terminated` | Session has ended, either because of an unrecoverable error or because it was archived. A session that finishes its work goes `idle`, not `terminated`.
| `terminated` | Session has ended, either because of an unrecoverable error or because it was archived. A session that completes its work goes `idle`, not `terminated`.
You can update a session's `agent.tools` (csharp, go: `Agent.Tools`) and `agent.mcp_servers` (java, php: `agent.mcpServers`; csharp: `Agent.McpServers`; go: `Agent.MCPServers`), in
You can update a session's `agent.tools` (csharp, go: `Agent.Tools`) and `agent.mcp_servers` (java, php: `agent.mcpServers`; csharp: `Agent.McpServers`; go: `Agent.MCPServers`), in
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/session-observability
±changed
+3−3
apichanged
managed-agents/session-observability
* **Timeline minimap:** A zoomable overview of the session's activity over time, with one lane per thread in [multiagent](https://platform.claude.com/docs/en/managed-agents/multiag
* **Timeline minimap:** A zoomable overview of the session's activity over time, with one lane for each thread in [multiagent](https://platform.claude.com/docs/en/managed-agents/mu
* **Review tool results:** Tool execution failures often explain unexpected agent behavior. The Inspector's **Tools** tab shows each tool's failures.
* **Review tool results:** Tool failures often explain unexpected agent behavior. The Inspector's **Tools** tab shows each tool's failures.
Send events, stream responses, and interrupt or redirect your session mid-execution.
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-workers
±changed
+6−6
apichanged
managed-agents/self-hosted-sandboxes-workers
The [quickstart](https://platform.claude.com/docs/en/managed-agents/self-hosted-sandboxes#quickstart) runs one `ant` CLI worker that polls continuously and runs every session in on
The [quickstart](https://platform.claude.com/docs/en/managed-agents/self-hosted-sandboxes#quickstart) runs one `ant` CLI worker that polls continuously and runs every session in on
Along with the environment ID and key from the [quickstart](https://platform.claude.com/docs/en/managed-agents/self-hosted-sandboxes#run-your-first-session), export the webhook
Along with the environment ID and key from [Run your first session](https://platform.claude.com/docs/en/managed-agents/self-hosted-sandboxes#run-your-first-session), export the
Invoke the worker when `session.status_run_started` fires. The handler drains the queue and hands each claimed work item to `handle_item()` (typescript: `handleItem()`; go: `Ha
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-security
±changed
+2−2
apichanged
managed-agents/self-hosted-sandboxes-security
* **Tool-execution blast radius.** Tools run inside your sandbox with whatever permissions your process has. Apply least privilege to the process user and mount only the directorie
* **Tool permissions.** Tools run inside your sandbox with whatever permissions your process has. Apply least privilege to the process user and mount only the directories your tool
* **Memory store contents.** [Memory stores](https://platform.claude.com/docs/en/managed-agents/memory) remain hosted by Anthropic, including their version history. When a session
* **Memory store contents.** [Memory stores](https://platform.claude.com/docs/en/managed-agents/memory) remain hosted by Anthropic, including their version history. When a session
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-reference
±changed
+1−1
apichanged
managed-agents/self-hosted-sandboxes-reference
| `block_ms` (typescript: `blockMs`; go: `BlockMs`) | How long each poll waits for work to arrive before returning, in milliseconds. Must be betw
| `block_ms` (typescript: `blockMs`; go: `BlockMs`) | How long each poll waits for work to arrive before returning, in milliseconds. Must be betw
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-operations
±changed
+7−7
apichanged
managed-agents/self-hosted-sandboxes-operations
By default the work item moves to `stopping`. The worker notices on its next lease heartbeat, cancels the session's in-flight tool call, and confirms the shutdown. The work item th
By default the work item moves to `stopping`. The worker detects this on its next lease heartbeat, cancels the session's in-flight tool call, and confirms the shutdown. The work it
A worker that is cancelled while a session runs stops its in-flight work before it exits. If the session has [memory stores](https://platform.claude.com/docs/en/managed-agents/self
A worker that is canceled while a session runs stops its in-flight work before it exits. If the session has [memory stores](https://platform.claude.com/docs/en/managed-agents/self-
| SDK | Backstop | Fires after |
+7 −7 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-memory
±changed
+2−2
apichanged
managed-agents/self-hosted-sandboxes-memory
Two sessions cannot mount the same store on one host at the same time, because both need the same path. If your sessions attach the same store, run one session per filesystem. Givi
Two sessions cannot mount the same store on one host at the same time, because both need the same path. If your sessions attach the same store, run one session for each filesystem.
The memory store on Anthropic's side remains the source of truth. [Memory versions](https://platform.claude.com/docs/en/managed-agents/memory#audit-memory-changes), redaction, and
The memory store on Anthropic's side remains the source of truth. [Memory versions](https://platform.claude.com/docs/en/managed-agents/memory#audit-memory-changes), redaction, and
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/self-hosted-sandboxes-custom-tools
±changed
+4−4
apichanged
managed-agents/self-hosted-sandboxes-custom-tools
[Custom tools](https://platform.claude.com/docs/en/managed-agents/tools#custom-tools) are tools your own code executes: the agent emits an `agent.custom_tool_use` event and waits f
[Custom tools](https://platform.claude.com/docs/en/managed-agents/tools#custom-tools) are tools your own code runs: the agent emits an `agent.custom_tool_use` event and waits for a
* **A generator such as pydantic factors schemas into `$defs`:** Inline those schemas before you declare the tool.
* **A generator such as Pydantic factors schemas into `$defs`:** Inline those schemas before you declare the tool.
Set a timeout on the MCP client for a faster and clearer failure, as the Python worker example does with `read_timeout_seconds`. See [A wrapped MCP tool call hangs](https://platfor
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/secure-deployment-risks
±changed
+1−1
apichanged
managed-agents/secure-deployment-risks
Claude can also take a step you did not expect while it keeps working to finish the task you gave it. For example, it might use a tool or credential it can reach to get past an obs
Claude can also take a step you did not expect while it keeps working to complete the task you gave it. For example, it might use a tool or credential it can reach to get past an o
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/secure-deployment
±changed
+62−62
apichanged
managed-agents/secure-deployment
Where the agent's tools run decides who secures the sandbox: Anthropic, with a cloud environment, or you, with a self-hosted sandbox. To choose between them and learn what reaches
Where the agent's tools run determines who secures the sandbox: Anthropic, with a cloud environment, or you, with a self-hosted sandbox. To choose between them and learn what reach
Set `networking.type` to `limited` and list only the hosts the agent needs in `allowed_hosts`. An environment create request that omits `networking` might get `unrestricted`, so se
Files, memory stores, and GitHub repositories that you attach stay available to the agent regardless of `allowed_hosts`.
+62 −62 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/scheduled-deployments
±changed
+16−16
apichanged
managed-agents/scheduled-deployments
A **scheduled deployment** allows an [agent](https://platform.claude.com/docs/en/managed-agents/agent-setup) to start [sessions](https://platform.claude.com/docs/en/managed-agents/
A **scheduled deployment** allows an [agent](https://platform.claude.com/docs/en/managed-agents/agent-setup) to start [sessions](https://platform.claude.com/docs/en/managed-agents/
* Deployments require [agent configuration](https://platform.claude.com/docs/en/managed-agents/agent-setup) and [environment configuration](https://platform.claude.com/docs/en/mana
* Deployments require [agent configuration](https://platform.claude.com/docs/en/managed-agents/agent-setup) and [environment configuration](https://platform.claude.com/docs/en/mana
The upcoming run timestamps reflect the exact schedule configured. However, to distribute load, actual execution applies jitter of up to 15% of the interval between runs, with a mi
+16 −16 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/reference
±changed
+17−17
apichanged
managed-agents/reference
Persisted event type strings follow a `{domain}.{action}` naming convention; the stream-only event deltas (see the Event deltas tab) are the exception. See [Session event stream](h
Persisted event type strings follow a `{domain}.{action}` naming convention; the stream-only event deltas (see the **Event deltas** tab) are the exception. See [Session event strea
| `user.interrupt` | Stop the agent mid-execution. It ends no [workflow run](https://platform.claude.com/docs/en/managed-agents/workflow-runs#interrupt-a-session-with-
| `user.interrupt` | Stop the agent while it runs. It ends no [workflow run](https://platform.claude.com/docs/en/managed-agents/workflow-runs#interrupt-a-session-with-
| Type | Description
+17 −17 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/quickstart
±changed
+1−1
apichanged
managed-agents/quickstart
Handle events and steer the agent mid-execution
Handle events and steer the agent while it runs
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/permission-policies
±changed
+7−7
apichanged
managed-agents/permission-policies
description: Control when agent and MCP tools execute.
description: Control when agent and MCP tools run.
Permission policies control whether server-executed tools (the pre-built agent toolset and MCP toolset) run automatically, wait for your approval, or have each call evaluated by th
Permission policies control whether server-executed tools (the pre-built agent toolset and MCP toolset) run automatically, wait for your approval, or have each call evaluated by th
MCP toolsets default to `always_ask`. This ensures that new tools added to an MCP server do not execute in your application without approval. To auto-approve tools from a trusted M
+7 −7 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/overview
±changed
+1−1
apichanged
managed-agents/overview
Send additional user events to guide the agent mid-execution, or interrupt it to change direction.
Send additional user events to guide the agent while it runs, or interrupt it to change direction.
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/onboarding
±changed
+3−3
apichanged
managed-agents/onboarding
description: Create, test, and iterate on agents visually in Console, then run them from your code with the API.
description: Create, test, and iterate on agents visually in the Claude Console, then run them from your code with the API.
[Console](https://platform.claude.com/workspaces/default/agent-quickstart/) provides a visual interface for creating and configuring agents. It lets you iterate on configuration in
The [Claude Console](https://platform.claude.com/workspaces/default/agent-quickstart/) provides a visual interface for creating and configuring agents. It lets you iterate on confi
## From Console to your codebase
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/multiagent-orchestration
±changed
+11−11
apichanged
managed-agents/multiagent-orchestration
| Approach | What happens
| -------------------------------------------------------------------------------------------------------------------------------------- | -----------------------------------------
| [Subagents](https://platform.claude.com/docs/en/managed-agents/multiagent-orchestration#delegate-to-subagents) | The agent delegates tasks to its subagent
| [Dynamic workflows](https://platform.claude.com/docs/en/managed-agents/multiagent-orchestration#dynamic-workflows) | The agent writes a workflow: a program th
| [Give the session an advisor](https://platform.claude.com/docs/en/managed-agents/multiagent-orchestration#give-the-session-an-advisor) | The session's [primary thread](https://pl
+11 −11 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/migration
±changed
+3−3
apichanged
managed-agents/migration
| You maintain the conversation history array and pass it back on every turn. | The session stores history server-side. Send events, receive events.
| You maintain the conversation history array and pass it back on every turn. | The session stores history server-side. You send events and receive events.
If you built with the [Claude Agent SDK](https://code.claude.com/docs/en/agent-sdk/overview), you're already working with agents, tools, and sessions as concepts. The difference is
If you built with the [Claude Agent SDK](https://code.claude.com/docs/en/agent-sdk/overview), you're already working with agents, tools, and sessions as concepts. The difference is
When a new Claude model is released, migrating a Claude Managed Agents integration is typically a one-field change: update `model` (csharp, go: `Model`) on your [agent definition](
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/memory
±changed
+29−29
apichanged
managed-agents/memory
Individual memories within the store are capped at 100 kB (\~25k tokens). A store holds a maximum of 10,000 memories. Structure memory as many small focused files, not a few larg
Individual memories within the store are capped at 100 kB (\~25k tokens). A store holds a maximum of 10,000 memories. Structure memory as many small focused files, not a few larg
--data '{
-d '{
A maximum of **8 memory stores** are supported per session. Attach multiple stores when different parts of memory have different owners or access rules. Common reasons:
+29 −29 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/mcp-connector
±changed
+6−6
apichanged
managed-agents/mcp-connector
The MCP toolset defaults to a permission policy of `always_ask`, which requires user approval before each tool call. See [permission policies](https://platform.claude.com/docs/en
The MCP toolset defaults to a permission policy of `always_ask`, which requires user approval before each tool call. See [Permission policies](https://platform.claude.com/docs/en
| Error type | Meaning
| --------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------
| `mcp_connection_failed_error` | The MCP server could not be reached (network error, timeout, or non-authentication HTTP failure).
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/github
±changed
+1−1
apichanged
managed-agents/github
Mounting a repository also loads any skills stored in its root `.claude/skills` directory. Skills are discovered once per session, from the repository state checked out at session
Mounting a repository also loads any skills stored in its root `.claude/skills` directory. Skills are discovered once for each session, from the repository state checked out at ses
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/files
±changed
+6−6
apichanged
managed-agents/files
A maximum of 500 files is supported per session.
A maximum of 500 files is supported for each session.
* The path you specify is rooted under the session's uploads directory: a `mount_path` (java, php: `mountPath`; csharp, go: `MountPath`) of `/data.csv` places the file at `/mnt/ses
* If you omit `mount_path` (java, php: `mountPath`; csharp, go: `MountPath`), the file is placed at `/mnt/session/uploads/<file_id>`
* Parent directories are created automatically
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/events-and-streaming
±changed
+8−8
apichanged
managed-agents/events-and-streaming
description: Send events, stream responses, and interrupt or redirect your session mid-execution.
description: Send events, stream responses, and interrupt or redirect your session while it runs.
The agent's response text arrives as buffered `agent.message` events, each emitted only after the model request that produced it finishes. To render the text while the model is sti
The agent's response text arrives as buffered `agent.message` events, each emitted only after the model request that produced it completes. To render the text while the model is st
# Tail live events, skipping anything already seen — Set#add? returns nil for duplicates
+8 −8 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/event-deltas
±changed
+10−10
apichanged
managed-agents/event-deltas
By default, the agent's response text reaches the [session event stream](https://platform.claude.com/docs/en/managed-agents/events-and-streaming) as buffered `agent.message` events
By default, the agent's response text reaches the [session event stream](https://platform.claude.com/docs/en/managed-agents/events-and-streaming) as buffered `agent.message` events
Previews are opt-in per stream connection. Add the `event_deltas[]` query parameter to the stream you're reading, and repeat it once for each event type you want previewed. The acc
Previews are opt-in for each stream connection. Add the `event_deltas[]` query parameter to the stream you're reading, and repeat it once for each event type you want previewed. Th
When a previewed event begins, the stream emits an `event_start` carrying the upcoming event's type and `id`:
+10 −10 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/environments
±changed
+5−5
apichanged
managed-agents/environments
--data '{
-d '{
[`ant apply`](https://platform.claude.com/docs/en/cli-sdks-libraries/cli/apply) creates the environment from `environment.yaml`, prints its ID, and records it in `claude-lock.j
[`ant apply`](https://platform.claude.com/docs/en/cli-sdks-libraries/cli/apply) creates the environment from `environment.yaml`, prints its ID, and records it in `claude-lock.j
--data '{
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/dreams
±changed
+7−7
apichanged
managed-agents/dreams
description: Let Claude reflect on past sessions to curate an agent's memory and surface new insights.
description: Let Claude review past sessions to curate an agent's memory and surface new insights.
--data '{
-d '{
Use `instructions` (csharp, go: `Instructions`) for high-level synthesis guidance such as focus areas ("focus on coding-style preferences"), content to preserve unchanged, or outpu
+7 −7 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/define-outcomes
±changed
+9−9
apichanged
managed-agents/define-outcomes
If you don't have a rubric on hand, try giving Claude an example of a known-good artifact and asking it to analyze what makes that content good, then turn that analysis into a ru
If you don't have a rubric, try giving Claude an example of a known-good artifact and asking it to analyze what makes that content good, then turn that analysis into a rubric. Th
The following examples create a [session](https://platform.claude.com/docs/en/managed-agents/sessions) for an existing [agent](https://platform.claude.com/docs/en/managed-agents/ag
The following examples create a [session](https://platform.claude.com/docs/en/managed-agents/sessions) for an existing [agent](https://platform.claude.com/docs/en/managed-agents/ag
# Define the outcome — agent starts working on receipt
+9 −9 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/cloud-sandboxes-reference
±changed
+1−1
apichanged
managed-agents/cloud-sandboxes-reference
| Go | 1.24 (default) and 1.25 | go modules |
| Go | 1.24 (default) and 1.25 | Go modules |
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/choosing-an-environment
±changed
+1−1
apichanged
managed-agents/choosing-an-environment
* **Self-hosted sandbox:** You run the sandbox on your own infrastructure. Claude still runs on Anthropic's side and decides which tool to call next. A program that you run, called
* **Self-hosted sandbox:** You run the sandbox on your own infrastructure. Claude still runs on Anthropic's side and determines which tool to call next. A program that you run, cal
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/budgets
±changed
+5−5
apichanged
managed-agents/budgets
A session budget is an optional hard spend ceiling you set when you [create a session](https://platform.claude.com/docs/en/managed-agents/sessions). The platform continuously price
A session budget is an optional hard spend ceiling you set when you [create a session](https://platform.claude.com/docs/en/managed-agents/sessions). The platform continuously price
* `max_list_cost` (java, php: `maxListCost`; csharp, go: `MaxListCost`) is the cap itself: `amount` (csharp, go: `Amount`) is a whole number of US cents written as a string with no
* `max_list_cost` (java, php: `maxListCost`; csharp, go: `MaxListCost`) is the cap itself: `amount` (csharp, go: `Amount`) is a whole number of US cents written as a string with no
If the session has open [workflow runs](https://platform.claude.com/docs/en/managed-agents/workflow-runs#interrupt-a-session-with-runs-open), the interrupt can pause them, as an in
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
managed-agents/agent-setup
±changed
+2−2
apichanged
managed-agents/agent-setup
The examples use curl, the `ant` CLI, or one of the SDKs. If you haven't set one up, the [quickstart](https://platform.claude.com/docs/en/managed-agents/quickstart#install-the-cli)
The examples use curl, the `ant` CLI, or the SDK. If you haven't set one up, the [quickstart](https://platform.claude.com/docs/en/managed-agents/quickstart#install-the-cli) covers
Create a session to run your agent and begin executing tasks.
Create a session to run your agent and start running tasks.
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/workspaces
±changed
+13−13
apichanged
manage-claude/workspaces
* **Maximum 100 workspaces** per organization by default (archived workspaces don't count); contact your account team if you need more
* **Maximum 100 workspaces** for each organization by default (archived workspaces don't count); contact your account team if you need more
* **API keys** can be scoped to a single workspace. In this case, they can only access resources within that workspace. Some API keys can be granted permissions across multiple wor
* **API keys** can be scoped to a single workspace. In this case, they can only access resources within that workspace. Other API keys aren't scoped to a workspace. With one of the
* **Organization admins** automatically receive Workspace Admin access to all workspaces
+13 −13 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/workload-identity-federation
±changed
+6−6
apichanged
manage-claude/workload-identity-federation
Workload Identity Federation strengthens your security posture by replacing static API keys with tokens that expire in minutes rather than never. It is not a complete security stor
Workload Identity Federation strengthens your security posture by replacing static API keys with tokens that expire in minutes rather than never. It is not a complete security solu
The key distinction versus a workspace API key: a workspace API key *is* a credential, while a service account *has* credentials. You can more easily audit which workloads acted as
The key distinction versus a workspace API key: a workspace API key *is* a credential, whereas a service account *has* credentials. You can more easily audit which workloads acted
You typically register one issuer per environment: your production EKS cluster, your staging cluster, and GitHub Actions are three separate issuers.
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-reference
±changed
+15−15
apichanged
manage-claude/wif-reference
| `org:admin` | Full access to the [Admin API](https://platform.claude.com/docs/en/manage-claude/admin-api) (organization members, invites, workspaces, API keys, and
| `org:admin` | Full access to the [Admin API](https://platform.claude.com/docs/en/manage-claude/admin-api) (organization members, invites, workspaces, API keys, and
| Rule `oauth_scope` | Service account `organization_role` | Effective permissions
| --------------------- | ----------------------------------- | --------------------------------------------------------------------------------------------------------------------
| `workspace:developer` | `admin` | Claude API access in the rule's workspace only. The scope caps the token below the role.
+15 −15 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/spiffe
±changed
+5−5
apichanged
manage-claude/wif-providers/spiffe
--data @- <<JSON | jq -r .access_token
-d @- <<JSON | jq -r .access_token
Except in Java, workloads that link a SPIFFE Workload API client directly can skip spiffe-helper and pass the SDK a token provider that fetches a fresh JWT-SVID from the agent
Except in Java, workloads that link a SPIFFE Workload API client directly can skip spiffe-helper and pass the SDK a token provider that fetches a fresh JWT-SVID from the agent
Before wiring the SDK in, fetch a JWT-SVID directly from SPIRE Agent and confirm the claims match what your federation rule expects. If you use a different SPIFFE implementation, f
+5 −5 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/okta
±changed
+3−3
apichanged
manage-claude/wif-providers/okta
2. **Configure client authentication.** For a keyless setup, choose **Public key / Private key** (`private_key_jwt`) and register your workload's public JWK. Alternatively, use a c
2. **Configure client authentication.** For a keyless setup, select **Public key / Private key** (`private_key_jwt`) and register your workload's public JWK. Alternatively, use a c
* **Use one rule per service app:** Create a separate federation rule for each service app rather than sharing one rule across apps.
* **Use one rule for each service app:** Create a separate federation rule for each service app rather than sharing one rule across apps.
* See the [WIF reference](https://platform.claude.com/docs/en/manage-claude/wif-reference#rule-matching-semantics) to match on custom Okta claims with CEL expressions.
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/kubernetes
±changed
+2−2
apichanged
manage-claude/wif-providers/kubernetes
--data @- <<JSON | jq -r .access_token
-d @- <<JSON | jq -r .access_token
* **Use a separate rule per namespace:** Create a distinct rule and Anthropic service account for each namespace rather than widening one rule.
* **Use a separate rule for each namespace:** Create a distinct rule and Anthropic service account for each namespace rather than widening one rule.
+2 −2 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/github-actions
±changed
+4−4
apichanged
manage-claude/wif-providers/github-actions
The token's `sub` claim encodes the repository and trigger context. For a push to a branch it has the form `repo:<owner>/<repo>:ref:refs/heads/<branch>`. Pull-request runs use `rep
The token's `sub` claim encodes the repository and trigger context. For a push to a branch it has the form `repo:<owner>/<repo>:ref:refs/heads/<branch>`. Pull-request runs use `rep
Be as specific as the workload allows. Loosen `subject_prefix` to `repo:your-org/your-repo:*` (paired with a `claims.ref` constraint) only if the rule must match multiple event typ
Be as specific as the workload allows. Loosen `subject_prefix` to `repo:your-org/your-repo:*` (paired with a `claims.ref` constraint) only if the rule must match multiple event typ
--data @- <<JSON | jq -r .access_token
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/gcp
±changed
+6−6
apichanged
manage-claude/wif-providers/gcp
Enable [Workload Identity](https://cloud.google.com/kubernetes-engine/docs/how-to/workload-identity) on your cluster and bind your Kubernetes service account to a Google servic
Enable [GKE Workload Identity](https://cloud.google.com/kubernetes-engine/docs/how-to/workload-identity) on your cluster and bind your Kubernetes service account to a Google se
A `format=full` token from GKE additionally includes `google.compute_engine.project_id`, `google.compute_engine.zone`, and `google.compute_engine.instance_name` claims, which y
A `format=full` token from GKE additionally includes `google.compute_engine.project_id`, `google.compute_engine.zone`, and `google.compute_engine.instance_name` claims, which y
--data @- <<JSON
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/azure
±changed
+53−53
apichanged
manage-claude/wif-providers/azure
2. **Set up the identity for your platform:** A managed identity on VMs, VM Scale Sets, App Service, Functions, and Container Apps, or Entra Workload Identity on AKS.
2. **Set up the identity for your platform:** Use a managed identity on VMs, VM Scale Sets, App Service, Functions, and Container Apps, or Entra Workload Identity on AKS.
On both paths the token you present to Anthropic carries your tenant-specific Entra issuer and the managed identity's object ID in the `sub` and `oid` claims; only how the workload
On both paths the token you present to Anthropic carries your tenant-specific Entra issuer and the managed identity's object ID in the `sub` and `oid` claims; only how the workload
**Federation issuer:** Choose **v2.0 (login.microsoftonline.com)** in the wizard's **Token issuer** selector. (The selector defaults to v1; that default exists for tenants reusing
+53 −53 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-providers/aws
±changed
+11−11
apichanged
manage-claude/wif-providers/aws
This is an account-level flag, off by default. In the AWS console, open **IAM**, choose **Account settings**, and enable **Outbound web identity federation**. To enable it prog
This is an account-level flag, off by default. In the AWS console, open **IAM**, select **Account settings**, and enable **Outbound web identity federation**. To enable it prog
After enabling outbound federation, the **IAM > Account settings** page shows a **Get Token Issuer URL** field with a value of the form `https://<uuid>.tokens.sts.global.api.aw
After you enable outbound federation, the **IAM > Account settings** page shows a **Get Token Issuer URL** field with a value of the form `https://<uuid>.tokens.sts.global.api.
Call `GetWebIdentityToken` with `https://api.anthropic.com` as the audience, then pass the result to the SDK's federation credentials through `identity_token_provider` (typescript:
+11 −11 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/wif-admin-api
±changed
+4−4
apichanged
manage-claude/wif-admin-api
The SDKs and the `ant` CLI read `ANTHROPIC_AUTH_TOKEN` automatically; leave `ANTHROPIC_API_KEY` unset in the same shell, because these endpoints reject API keys and every client us
The SDK and the `ant` CLI read `ANTHROPIC_AUTH_TOKEN` automatically; leave `ANTHROPIC_API_KEY` unset in the same shell, because these endpoints reject API keys and every client use
A workload that uses one of the SDKs or the `ant` CLI does not perform the exchange itself. Point the client at the rule with the federation environment variables and construct
A workload that uses the SDK or the `ant` CLI does not perform the exchange itself. Point the client at the rule with the federation environment variables and construct it with
The `ant` CLI reads the same variables, or takes `--federation-rule`, `--organization-id`, `--service-account-id`, and `--identity-token-file` flags. For a workload that runs m
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/user-profiles
±changed
+3−3
apichanged
manage-claude/user-profiles
## Before you begin
## Before you start
`enforcement_status` sums up the open actions in one value:
`enforcement_status` summarizes the open actions in one value:
The following example marks the account as suspended and removes a `trial_cohort` metadata key.
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/user-management
±changed
+4−4
apichanged
manage-claude/user-management
The Admin API is a single set of endpoints under `https://api.anthropic.com/v1/organizations/`. Claude Console and Claude Enterprise organizations authenticate with [different keys
The Admin API is a single set of endpoints under `https://api.anthropic.com/v1/organizations/`. Claude Console and Claude Enterprise organizations authenticate with different [Admi
Groups connect members to custom roles (role-based access control, the `rbac` in the endpoint paths and scope names). Groups are owned by your enterprise as a whole (the parent org
Groups connect members to custom roles (role-based access control, the `rbac` in the endpoint paths and scope names). Groups are owned by your enterprise as a whole (the parent org
Member and invite lists use ID-based pagination: pass `limit` (default 20, max 1000) plus at most one of `before_id` or `after_id`, and page using the `first_id` and `last_id` fiel
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/usage-cost-api
±changed
+6−6
apichanged
manage-claude/usage-cost-api
This API enables you to better monitor, analyze, and optimize your Claude implementations:
This API allows you to better monitor, analyze, and optimize your Claude implementations:
* **[Rate limit](https://platform.claude.com/docs/en/api/rate-limits) optimization:** Optimize features like [prompt caching](https://platform.claude.com/docs/en/build-with-claude/
* **[Rate limit](https://platform.claude.com/docs/en/api/rate-limits) optimization:** Optimize features such as [prompt caching](https://platform.claude.com/docs/en/build-with-clau
Explore detailed implementations in [Claude Cookbook](https://platform.claude.com/cookbook):
+6 −6 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/spend-limits-api
±changed
+17−17
apichanged
manage-claude/spend-limits-api
All monetary values are strings in **minor units of the organization's billing currency** (cents, for USD). For example, `"50000"` represents 500.00 USD. Parse as a decimal and div
All monetary values are strings in **minor units of the organization's billing currency** (cents, for USD). For example, `"50000"` represents $500.00 USD. Parse as a decimal and di
`period_to_date_spend` is the member's spend accrued since the start of the current `period`, in the same minor-unit format; it may include a fractional part (for example, `"41280.
`period_to_date_spend` is the member's spend accrued since the start of the current `period`, in the same minor-unit format; it might include a fractional part (for example, `"4128
| Status | Meaning
+17 −17 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/rate-limits-api
±changed
+3−3
apichanged
manage-claude/rate-limits-api
* **The `id` inside `group`:** A group has the same `id` in every organization and on every workspace override, and it never changes. Use it to match entries across organizations o
* **The `id` inside `group`:** A group has the same `id` in every organization and on every workspace override, and it never changes. Use it to match entries across organizations o
To retrieve your organization's workspace IDs, use the [List Workspaces](https://platform.claude.com/docs/en/api/organization/workspaces/list) endpoint, or find them in the [Clau
To retrieve your organization's workspace IDs, use the [List Workspaces](https://platform.claude.com/docs/en/api/organization/workspaces/list) endpoint, or find them on the [Work
The list covers the models your organization has access to, so it can differ between organizations.
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/plugins-api
±changed
+69−69
apichanged
manage-claude/plugins-api
Every request must include the [beta header](https://platform.claude.com/docs/en/api/beta-headers) `anthropic-beta: ce-plugins-2026-09-01` (the SDKs and the `ant` CLI send it for
Every request must include the [beta header](https://platform.claude.com/docs/en/api/beta-headers) `anthropic-beta: ce-plugins-2026-09-01` (the SDK and the `ant` CLI send it for
The Python, TypeScript, C#, Go, Java, PHP, and Ruby SDKs expose these endpoints under `client.beta.organization` (csharp, go: `client.Beta.Organization`; java: `client.beta().organ
The SDK exposes these endpoints under `client.beta.organization` (csharp, go: `client.Beta.Organization`; java: `client.beta().organization()`; php: `$client->beta->organization`),
| Component | What it is |
+69 −69 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/organization-webhooks
±changed
+1−1
apichanged
manage-claude/organization-webhooks
## Before you begin
## Before you start
+1 −1 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/opentelemetry/sdk
±changed
+4−4
apichanged
manage-claude/opentelemetry/sdk
includes. Otherwise, a
[bug in that package](https://github.com/open-telemetry/opentelemetry-js-contrib/issues/3711)
includes. Otherwise, an
+4 −4 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/opentelemetry/api
±changed
+8−8
apichanged
manage-claude/opentelemetry/api
| **Provider** | Your observability platform, or **Custom** if it isn't listed. If you choose a listed platform, the Console asks for its details, such as an A
| **Provider** | Your observability platform, or **Custom** if it isn't listed. If you select a listed platform, the Console asks for its details, such as an A
| **Maximum content size in MB** | Optional, under **Advanced settings** when the content mode is **Full content**. The most content that the API sends in each span. The Console
| **Maximum content size in MB** | Optional, under **Advanced settings** when the content mode is **Full content**. The most content that the API sends in each span. The Console
| Problem | What to check
+8 −8 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/opentelemetry
±changed
+3−3
apichanged
manage-claude/opentelemetry
| [Token usage](https://platform.claude.com/docs/en/manage-claude/opentelemetry#count-tokens) | ?
| [Token counts](https://platform.claude.com/docs/en/manage-claude/opentelemetry#count-tokens) | ?
With Full content, each span holds up to 0.75 MB of content by default. That fits a 200k-token context window of English text and stays below the 1 MB request limit of many observa
With Full content, each span holds up to 0.75 MB of content by default. That fits a 200k-token context window of English text and stays below the 1 MB request limit of many observa
![Example trace of a Managed Agents session: your POST /chat span, the event stream and event list requests of the session tool runner, the client and server spans for the request
+3 −3 lines
Recorded here 21 Oct 2026 19:37 UTC
api
manage-claude/inference-hooks-endpoint
±changed
+9−9
apichanged
manage-claude/inference-hooks-endpoint
There are two hook events, told apart by the top-level `type` field. The prompt frame is sent once per governed inference request, before inference begins. The tool call frame is s
There are two hook events, told apart by the top-level `type` field. The prompt frame is sent once for each governed inference request, before inference starts. The tool call frame
Transcripts are sent untruncated, so a long conversation with large attachments produces a large request body. In practice the model's context window keeps bodies under about 10 MB
Transcripts are sent untruncated, so a long conversation with large attachments produces a large request body. In practice the model's context window keeps bodies under about 10 MB
* `messages` holds only the latest message, the `assistant` message Claude just produced: any `text` blocks and one `tool_use` block per tool call the frame lists, in the order the
+9 −9 lines
Recorded here 21 Oct 2026 19:37 UTC
Nothing on this page matches that. Press Enter to search every page of the ledger.