Claude Agent SDK Bundles Updated CLI
📡 GitHub · anthropics/claude-agent-sdk-python
🕐 24.06, 19:11
🏷️ #Anthropic #ClaudeCode #Claude #claude_agent_sdk
The `claude-agent-sdk` for Python now ships with an updated Claude CLI, bumping it to version 2.1.190.
💡 This internal update means your Python agent workflows are now running on a more recent version of the underlying Claude CLI.
It's not a direct feature addition to the SDK, but it often brings stability improvements, performance tweaks, or prepares the ground for future integrations with new Claude capabilities.
🔗 Official source
📡 GitHub · anthropics/claude-agent-sdk-python
🕐 24.06, 19:11
🏷️ #Anthropic #ClaudeCode #Claude #claude_agent_sdk
The `claude-agent-sdk` for Python now ships with an updated Claude CLI, bumping it to version 2.1.190.
💡 This internal update means your Python agent workflows are now running on a more recent version of the underlying Claude CLI.
It's not a direct feature addition to the SDK, but it often brings stability improvements, performance tweaks, or prepares the ground for future integrations with new Claude capabilities.
🔗 Official source
Claude Code v2.1.190: Stability & Reliability Boost
📡 GitHub · anthropics/claude-code
🕐 24.06, 18:53
🏷️ #Anthropic #ClaudeCode #Claude #bugfix
• This update focuses on core stability.
• Expect fewer unexpected hiccups and more consistent performance.
💡 When the underlying tool gets a reliability bump, it translates directly into less wasted time chasing phantom bugs or wrestling with inconsistent behavior.
This means more predictable development cycles and fewer "why is this happening now?" moments.
🔗 Official source
📡 GitHub · anthropics/claude-code
🕐 24.06, 18:53
🏷️ #Anthropic #ClaudeCode #Claude #bugfix
• This update focuses on core stability.
• Expect fewer unexpected hiccups and more consistent performance.
💡 When the underlying tool gets a reliability bump, it translates directly into less wasted time chasing phantom bugs or wrestling with inconsistent behavior.
This means more predictable development cycles and fewer "why is this happening now?" moments.
🔗 Official source
Anthropic Economic Report: User Cadences & Impact
📡 anthropic.com
🕐 27.06, 08:00
🏷️ #Anthropic #ClaudeCode #Claude #economic_report
The new Anthropic Economic Index report just dropped.
It dives deep into *when* developers are engaging with Claude.
The report also details *what* they're actually building and producing using the AI.
Crucially, it explores *how* users perceive AI's real-world impact on their day-to-day work.
💡 This report isn't just numbers; it's a window into the evolving developer-AI symbiosis, showing where Claude is truly making a difference in productivity and workflow integration.
If you're building tools around Claude or just curious about real-world adoption, this data is gold.
🔗 Official source
📡 anthropic.com
🕐 27.06, 08:00
🏷️ #Anthropic #ClaudeCode #Claude #economic_report
The new Anthropic Economic Index report just dropped.
It dives deep into *when* developers are engaging with Claude.
The report also details *what* they're actually building and producing using the AI.
Crucially, it explores *how* users perceive AI's real-world impact on their day-to-day work.
💡 This report isn't just numbers; it's a window into the evolving developer-AI symbiosis, showing where Claude is truly making a difference in productivity and workflow integration.
If you're building tools around Claude or just curious about real-world adoption, this data is gold.
🔗 Official source
Fullscreen Mouse Control & Hook Fixes
📡 GitHub · anthropics/claude-code
🕐 27.06, 00:29
🏷️ #Anthropic #ClaudeCode #Claude #fullscreen_mode
• New `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` environment variable added.
• Disable mouse click, drag, and hover in fullscreen mode while keeping wheel scroll.
• Hook matchers now correctly handle hyphenated identifiers (e.g., `code-rev`).
💡 The `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` variable is a game-changer for presentations or focused debugging sessions where you absolutely don't want accidental clicks disrupting your flow.
No more rogue clicks jumping you out of fullscreen or triggering unintended actions mid-demo.
🚀
🔗 Official source
📡 GitHub · anthropics/claude-code
🕐 27.06, 00:29
🏷️ #Anthropic #ClaudeCode #Claude #fullscreen_mode
• New `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` environment variable added.
• Disable mouse click, drag, and hover in fullscreen mode while keeping wheel scroll.
• Hook matchers now correctly handle hyphenated identifiers (e.g., `code-rev`).
💡 The `CLAUDE_CODE_DISABLE_MOUSE_CLICKS` variable is a game-changer for presentations or focused debugging sessions where you absolutely don't want accidental clicks disrupting your flow.
No more rogue clicks jumping you out of fullscreen or triggering unintended actions mid-demo.
🚀
🔗 Official source
Stream Tool JSON Input Gets Lazy Parsing
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #typescript_sdk
• The SDK now lazily parses partial tool JSON input when streaming.
• This enhances robustness and efficiency for tool calls in streaming contexts.
💡 Lazy parsing of partial tool JSON input in streams is a game-changer for long-running or complex tool interactions.
It means your application can start processing tool arguments even before the full JSON object is transmitted, potentially reducing perceived latency and making your streaming tool calls more resilient to network hiccups or chunked responses.
This is crucial for building highly interactive agents where prompt responses and tool execution need to feel instantaneous.
⚡
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #typescript_sdk
• The SDK now lazily parses partial tool JSON input when streaming.
• This enhances robustness and efficiency for tool calls in streaming contexts.
💡 Lazy parsing of partial tool JSON input in streams is a game-changer for long-running or complex tool interactions.
It means your application can start processing tool arguments even before the full JSON object is transmitted, potentially reducing perceived latency and making your streaming tool calls more resilient to network hiccups or chunked responses.
This is crucial for building highly interactive agents where prompt responses and tool execution need to feel instantaneous.
⚡
🔗 Official source
Bedrock SDK v0.31.0: Lazy Tool JSON Parsing for Streams
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #bedrock_sdk
• Introduces lazy parsing for partial tool JSON input within streaming operations.
• This means the SDK can start processing tool calls even before the complete JSON payload arrives.
• Internal tests have been updated to no longer use deprecated models.
💡 Lazy parsing of partial tool JSON input is a huge win for real-time applications leveraging streaming tool calls.
Instead of buffering the entire tool input, which can introduce noticeable delays, the SDK now processes chunks as they stream in.
This dramatically reduces perceived latency, making your tool-orchestrated applications feel much snappier and more responsive, especially for complex, multi-step operations.
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #bedrock_sdk
• Introduces lazy parsing for partial tool JSON input within streaming operations.
• This means the SDK can start processing tool calls even before the complete JSON payload arrives.
• Internal tests have been updated to no longer use deprecated models.
💡 Lazy parsing of partial tool JSON input is a huge win for real-time applications leveraging streaming tool calls.
Instead of buffering the entire tool input, which can introduce noticeable delays, the SDK now processes chunks as they stream in.
This dramatically reduces perceived latency, making your tool-orchestrated applications feel much snappier and more responsive, especially for complex, multi-step operations.
🔗 Official source
Vertex SDK: Stream Tool JSON Input Lazily
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #vertex_sdk
• Streams now lazily parse partial tool JSON input.
This means the SDK can process tool calls incrementally as data arrives, improving responsiveness.
💡 Lazily parsing partial tool JSON in streams is a game-changer for applications relying on real-time tool execution.
It significantly reduces the perceived latency, allowing your application to react to tool calls much faster without waiting for the entire JSON payload to complete.
This is crucial for highly interactive UIs or rapid backend task orchestration.
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #vertex_sdk
• Streams now lazily parse partial tool JSON input.
This means the SDK can process tool calls incrementally as data arrives, improving responsiveness.
💡 Lazily parsing partial tool JSON in streams is a game-changer for applications relying on real-time tool execution.
It significantly reduces the perceived latency, allowing your application to react to tool calls much faster without waiting for the entire JSON payload to complete.
This is crucial for highly interactive UIs or rapid backend task orchestration.
🔗 Official source
Code Execution Tool Lands in TypeScript SDK
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
• Added support for the powerful new `code_execution_20260120` tool.
• Improved streaming with lazy parsing of partial tool JSON inputs, making stream processing more robust.
💡 The `code_execution_20260120` tool is a game-changer for building more autonomous agents, allowing Claude to execute code and integrate external logic directly within its reasoning loop.
This means you can now design complex workflows where Claude can dynamically generate, run, and debug code, opening up possibilities for advanced data analysis, system automation, and more robust problem-solving.
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 18.06, 20:18
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
• Added support for the powerful new `code_execution_20260120` tool.
• Improved streaming with lazy parsing of partial tool JSON inputs, making stream processing more robust.
💡 The `code_execution_20260120` tool is a game-changer for building more autonomous agents, allowing Claude to execute code and integrate external logic directly within its reasoning loop.
This means you can now design complex workflows where Claude can dynamically generate, run, and debug code, opening up possibilities for advanced data analysis, system automation, and more robust problem-solving.
🔗 Official source
Claude on Mars
📡 youtube.com
🕐 17.06, 19:39
🏷️ #Anthropic #ClaudeCode #Claude #teaser
This item consists solely of the title 'Claude on Mars'.
No further content, release notes, or technical details are provided to describe any new features, updates, or changes.
💡 When an announcement is this sparse, it's a strong signal that there's no immediate, actionable technical update to integrate.
Keep an eye out for follow-up communications, as this is likely a placeholder or a very early, high-level teaser.
🔗 Official source
📡 youtube.com
🕐 17.06, 19:39
🏷️ #Anthropic #ClaudeCode #Claude #teaser
This item consists solely of the title 'Claude on Mars'.
No further content, release notes, or technical details are provided to describe any new features, updates, or changes.
💡 When an announcement is this sparse, it's a strong signal that there's no immediate, actionable technical update to integrate.
Keep an eye out for follow-up communications, as this is likely a placeholder or a very early, high-level teaser.
🔗 Official source
Claude Opus 4.6 Released
📡 youtube.com
🕐 17.06, 19:35
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Opus
Claude Opus 4.6 has been officially introduced.
💡 While the specific enhancements are not detailed in this announcement, a new version number for a model like Opus typically signifies advancements in areas such as reasoning, context window management, or overall performance.
Look for the official release notes to dive into the concrete improvements that will impact your applications.
🔗 Official source
📡 youtube.com
🕐 17.06, 19:35
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Opus
Claude Opus 4.6 has been officially introduced.
💡 While the specific enhancements are not detailed in this announcement, a new version number for a model like Opus typically signifies advancements in areas such as reasoning, context window management, or overall performance.
Look for the official release notes to dive into the concrete improvements that will impact your applications.
🔗 Official source
Claude Fable 5 Is Here
📡 youtube.com
🕐 16.06, 23:28
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Fable
A new major version, Claude Fable 5, has been officially introduced.
Details regarding specific new features and enhancements are anticipated.
💡 The arrival of a new major version like Fable 5, even without immediate specifics, often signals foundational architectural shifts or significant leaps in core capabilities.
This is where developers typically find the biggest performance gains or entirely new interaction paradigms, so get ready for potential new APIs or even breaking changes to adapt to.
🔗 Official source
📡 youtube.com
🕐 16.06, 23:28
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Fable
A new major version, Claude Fable 5, has been officially introduced.
Details regarding specific new features and enhancements are anticipated.
💡 The arrival of a new major version like Fable 5, even without immediate specifics, often signals foundational architectural shifts or significant leaps in core capabilities.
This is where developers typically find the biggest performance gains or entirely new interaction paradigms, so get ready for potential new APIs or even breaking changes to adapt to.
🔗 Official source
Org Default Models for Claude
📡 GitHub · anthropics/claude-code
🕐 30.06, 02:27
🏷️ #Anthropic #ClaudeCode #Claude #organization_settings
• Admins can now set a default Claude model for their organization ⚙️.
• This default is configured directly in the organization console.
• Users will see "Org default" (or "Role default") in the /model selector if they haven't explicitly chosen one themselves.
💡 This is a game-changer for consistency and cost control in larger teams.
Admins can now enforce a specific model, ensuring everyone is on the same page for projects, or even steer users towards more cost-effective models without sacrificing core functionality.
🔗 Official source
📡 GitHub · anthropics/claude-code
🕐 30.06, 02:27
🏷️ #Anthropic #ClaudeCode #Claude #organization_settings
• Admins can now set a default Claude model for their organization ⚙️.
• This default is configured directly in the organization console.
• Users will see "Org default" (or "Role default") in the /model selector if they haven't explicitly chosen one themselves.
💡 This is a game-changer for consistency and cost control in larger teams.
Admins can now enforce a specific model, ensuring everyone is on the same page for projects, or even steer users towards more cost-effective models without sacrificing core functionality.
🔗 Official source
SDK v0.107.0: Web Fetch & Tooling Arrive
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 29.06, 18:01
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
• The SDK now fully supports the `20260318` API version.
• This brings `web fetch` capabilities directly into your prompts.
• Crucially, it adds robust `tools` support for more complex agentic workflows.
• Includes a fix for `agent-toolset` symlink canonicalization.
💡 The addition of `tools` support with the `20260318` API is a game-changer for building sophisticated agents.
You can now define external functions and have the model intelligently decide when and how to call them, moving beyond simple text generation to dynamic, action-oriented applications.
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 29.06, 18:01
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
• The SDK now fully supports the `20260318` API version.
• This brings `web fetch` capabilities directly into your prompts.
• Crucially, it adds robust `tools` support for more complex agentic workflows.
• Includes a fix for `agent-toolset` symlink canonicalization.
💡 The addition of `tools` support with the `20260318` API is a game-changer for building sophisticated agents.
You can now define external functions and have the model intelligently decide when and how to call them, moving beyond simple text generation to dynamic, action-oriented applications.
🔗 Official source
💡 Tip: Master Tool Use: Force or Block Calls with `tool_choice`
📡 Dobby-ai · Tip · Claude Code
🕐 30.06, 17:00
🏷️ #Anthropic #ClaudeCode #Claude #API
Ever need Claude to *always* use a specific tool, or *never* touch one? The `tool_choice` parameter in the `messages` API is your secret weapon.
Set `tool_choice: {"type": "tool", "name": "your_tool_name"}` to force a specific tool call, or use `tool_choice: {"type": "none"}` to prevent *any* tool calls, even if Claude thinks it's a good idea.
It's all about precision in your AI workflows.
🛠️
💡 This helps immensely when building robust applications where you need deterministic behavior for critical actions, like ensuring a payment gateway is always invoked or preventing accidental external API calls during a sensitive phase.
🔗 Official docs
📡 Dobby-ai · Tip · Claude Code
🕐 30.06, 17:00
🏷️ #Anthropic #ClaudeCode #Claude #API
Ever need Claude to *always* use a specific tool, or *never* touch one? The `tool_choice` parameter in the `messages` API is your secret weapon.
Set `tool_choice: {"type": "tool", "name": "your_tool_name"}` to force a specific tool call, or use `tool_choice: {"type": "none"}` to prevent *any* tool calls, even if Claude thinks it's a good idea.
It's all about precision in your AI workflows.
🛠️
💡 This helps immensely when building robust applications where you need deterministic behavior for critical actions, like ensuring a payment gateway is always invoked or preventing accidental external API calls during a sensitive phase.
🔗 Official docs
Claude Sonnet 5: Agentic Boost for Devs
📡 anthropic.com
🕐 01.07, 08:00
🏷️ #Anthropic #ClaudeCode #Claude #agentic_ai
• Sonnet 5 is now live, featuring a significant leap in agentic capabilities.
• Experience top-tier intelligence specifically tuned for complex coding and demanding professional workflows.
• This release is engineered to handle multi-step operations and intricate problem-solving with greater autonomy.
🤖
💡 The 'most agentic yet' claim means Sonnet 5 should be significantly better at tackling complex, multi-step coding problems without constant hand-holding.
Developers can potentially offload more intricate refactoring, debugging, or even small feature implementations to Claude, trusting it to manage the intermediate steps and self-correct.
Try feeding it a multi-stage problem that previously required breaking down into smaller prompts.
🔗 Official source
📡 anthropic.com
🕐 01.07, 08:00
🏷️ #Anthropic #ClaudeCode #Claude #agentic_ai
• Sonnet 5 is now live, featuring a significant leap in agentic capabilities.
• Experience top-tier intelligence specifically tuned for complex coding and demanding professional workflows.
• This release is engineered to handle multi-step operations and intricate problem-solving with greater autonomy.
🤖
💡 The 'most agentic yet' claim means Sonnet 5 should be significantly better at tackling complex, multi-step coding problems without constant hand-holding.
Developers can potentially offload more intricate refactoring, debugging, or even small feature implementations to Claude, trusting it to manage the intermediate steps and self-correct.
Try feeding it a multi-stage problem that previously required breaking down into smaller prompts.
🔗 Official source
Claude Science: AI Workbench for Researchers
📡 anthropic.com
🕐 01.07, 09:00
🏷️ #Anthropic #ClaudeCode #Claude #AI_workbench
• Introducing Claude Science, a new customizable application for scientists.
• It integrates tools and packages researchers frequently use.
• The platform produces auditable artifacts, enhancing research integrity.
• Provides flexible access to computing resources for demanding tasks.
💡 The inclusion of auditable artifacts is a massive win for scientific rigor, enabling researchers to trace and verify every step of their AI-assisted workflows, which is absolutely critical for reproducibility and building trust in experimental outcomes.
🔬
🔗 Official source
📡 anthropic.com
🕐 01.07, 09:00
🏷️ #Anthropic #ClaudeCode #Claude #AI_workbench
• Introducing Claude Science, a new customizable application for scientists.
• It integrates tools and packages researchers frequently use.
• The platform produces auditable artifacts, enhancing research integrity.
• Provides flexible access to computing resources for demanding tasks.
💡 The inclusion of auditable artifacts is a massive win for scientific rigor, enabling researchers to trace and verify every step of their AI-assisted workflows, which is absolutely critical for reproducibility and building trust in experimental outcomes.
🔬
🔗 Official source
Claude Fable 5 Returns with Beefed-Up Security
📡 anthropic.com
🕐 01.07, 18:00
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Fable_5
• Claude Fable 5 is back in action, with redeployment starting July 1.
• This follows the lifting of export controls.
• It now includes updated cybersecurity safeguards.
• A brand-new industry jailbreak framework has been integrated.
💡 The introduction of a new industry jailbreak framework is a game-changer for building more resilient and secure AI applications.
This isn't just about preventing malicious prompts; it's about establishing a standardized, robust defense mechanism that developers can trust to keep their deployments safe from adversarial attacks, allowing for more predictable and safer model behavior in production.
🔗 Official source
📡 anthropic.com
🕐 01.07, 18:00
🏷️ #Anthropic #ClaudeCode #Claude #Claude_Fable_5
• Claude Fable 5 is back in action, with redeployment starting July 1.
• This follows the lifting of export controls.
• It now includes updated cybersecurity safeguards.
• A brand-new industry jailbreak framework has been integrated.
💡 The introduction of a new industry jailbreak framework is a game-changer for building more resilient and secure AI applications.
This isn't just about preventing malicious prompts; it's about establishing a standardized, robust defense mechanism that developers can trust to keep their deployments safe from adversarial attacks, allowing for more predictable and safer model behavior in production.
🔗 Official source
Claude Fable 5 is back in Claude Code 🎉
📡 Anthropic
🏷️ #ClaudeCode #Fable5 #Anthropic
Anthropic's most capable model ever is live again globally as of July 1 — in Claude Code, claude.ai, the Claude Platform and Cowork.
It was suspended on June 12 under a US export-control directive, after researchers showed a jailbreak that pushed it into finding software vulnerabilities.
Controls are now lifted: Fable 5 returns with a new safety classifier that blocks that technique in over 99% of cases, falling back to Opus 4.8 when triggered.
💡 Run /model and pick Fable 5 for your hardest tasks — but budget it: it draws down weekly usage much faster than Opus 4.8, and through July 7 it's capped at 50% of weekly limits on Pro/Max/Team plans.
Which tasks get the Mythos-class brain — and where is Opus 4.8 still enough?
🔗 anthropic.com/news/redeploying-fable-5
📡 Anthropic
🏷️ #ClaudeCode #Fable5 #Anthropic
Anthropic's most capable model ever is live again globally as of July 1 — in Claude Code, claude.ai, the Claude Platform and Cowork.
It was suspended on June 12 under a US export-control directive, after researchers showed a jailbreak that pushed it into finding software vulnerabilities.
Controls are now lifted: Fable 5 returns with a new safety classifier that blocks that technique in over 99% of cases, falling back to Opus 4.8 when triggered.
💡 Run /model and pick Fable 5 for your hardest tasks — but budget it: it draws down weekly usage much faster than Opus 4.8, and through July 7 it's capped at 50% of weekly limits on Pro/Max/Team plans.
Which tasks get the Mythos-class brain — and where is Opus 4.8 still enough?
🔗 anthropic.com/news/redeploying-fable-5
TypeScript SDK: Non-Functional Types Removed
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 02.07, 00:55
🏷️ #Anthropic #ClaudeCode #Claude #typescript_sdk
• The SDK received a minor cleanup, specifically removing non-functional types from the API definitions.
This is a maintenance release focused on tidying up the codebase.
💡 While not a new feature, pruning non-functional types is a solid step for maintenance.
It means less noise in your IDE's autocompletion and potentially smaller type declaration files, which can subtly speed up your build processes and improve overall developer experience.
This kind of cleanup is a win for long-term project health.
🔗 Official source
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 02.07, 00:55
🏷️ #Anthropic #ClaudeCode #Claude #typescript_sdk
• The SDK received a minor cleanup, specifically removing non-functional types from the API definitions.
This is a maintenance release focused on tidying up the codebase.
💡 While not a new feature, pruning non-functional types is a solid step for maintenance.
It means less noise in your IDE's autocompletion and potentially smaller type declaration files, which can subtly speed up your build processes and improve overall developer experience.
This kind of cleanup is a win for long-term project health.
🔗 Official source
Claude in Chrome Hits GA, Agent Notifications Land
📡 GitHub · anthropics/claude-code
🕐 01.07, 23:45
🏷️ #Anthropic #ClaudeCode #Claude #chrome_extension
• The Claude in Chrome extension is now generally available! No more beta access hurdles.
• Background agent notifications are live for `claude-agents`.
Get pinged when sessions need input or complete their tasks.
🔔
💡 The new `Notification` hook for `agent_needs_input` and session completion is a game-changer for long-running or interactive agent workflows.
You can finally step away from the terminal and trust your agents to signal when they need attention or have finished their work, making complex multi-step tasks far more manageable.
🔗 Official source
📡 GitHub · anthropics/claude-code
🕐 01.07, 23:45
🏷️ #Anthropic #ClaudeCode #Claude #chrome_extension
• The Claude in Chrome extension is now generally available! No more beta access hurdles.
• Background agent notifications are live for `claude-agents`.
Get pinged when sessions need input or complete their tasks.
🔔
💡 The new `Notification` hook for `agent_needs_input` and session completion is a game-changer for long-running or interactive agent workflows.
You can finally step away from the terminal and trust your agents to signal when they need attention or have finished their work, making complex multi-step tasks far more manageable.
🔗 Official source
SDK v0.109.0: Managed Agents Get Event Streaming & Overrides
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 30.06, 22:47
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
This SDK update brings some serious firepower to your agent workflows:
• Managed Agents now support event delta streaming, enabling real-time updates.
• You can now set agent overrides directly, giving you more control.
• API calls offer reverse pagination for more flexible data retrieval.
• Vault credential injection is officially supported for enhanced security.
💡 Managed Agents event delta streaming is a game-changer for building highly interactive and responsive agent experiences.
Instead of waiting for a full response, you can now process partial updates in real-time, enabling richer UI feedback or more dynamic agent orchestrations.
This is huge for complex, multi-turn interactions where immediate feedback is crucial.
🚀
📡 GitHub · anthropics/anthropic-sdk-typescript
🕐 30.06, 22:47
🏷️ #Anthropic #ClaudeCode #Claude #SDK_update
This SDK update brings some serious firepower to your agent workflows:
• Managed Agents now support event delta streaming, enabling real-time updates.
• You can now set agent overrides directly, giving you more control.
• API calls offer reverse pagination for more flexible data retrieval.
• Vault credential injection is officially supported for enhanced security.
💡 Managed Agents event delta streaming is a game-changer for building highly interactive and responsive agent experiences.
Instead of waiting for a full response, you can now process partial updates in real-time, enabling richer UI feedback or more dynamic agent orchestrations.
This is huge for complex, multi-turn interactions where immediate feedback is crucial.
🚀