Anthropic AI Updates: October 22, 2026

1. Anthropic Previews Claude Duet, a Proactive Agent App That Works in the Background

Anthropic. Anthropic released Claude Duet as a research preview, a standalone app for iOS, Android, and claude.ai in which Claude sends a daily brief, checks connected accounts every few hours, and completes assigned tasks on its own virtual machine with a browser so the user’s device can stay off. It can draw on more than 6,000 connectors and plugins from the Claude Marketplace (Gmail, Slack, QuickBooks, Stripe, Shopify and others) and can make purchases through Stripe’s Link wallet with a virtual card or through Shop Pay. Sending messages, making payments, signing documents, and deleting anything require explicit approval, trading securities and moving money between accounts are blocked outright, and a dedicated classifier rejects requests to share credit card numbers or government IDs. Duet is built around Claude Opus 5.5 and is rolling out by waitlist to US Pro and Max subscribers, sharing their existing usage limits. Source

2. Claude API Version 2026-10-21 Ships With Major SDK Releases Relicensed to Apache 2.0

Anthropic. Anthropic released API version 2026-10-21 along with new major SDK versions that send it by default: Python 2.0.0, TypeScript 1.0.0, C# 13.0.0, Go 2.0.0, Java 3.0.0, PHP 1.0.0, Ruby 2.0.0, and the ant CLI 2.0.0. The releases also include breaking changes independent of the API version, covered in a migration guide, and the SDKs and CLI move from the MIT license to Apache 2.0. A day earlier Anthropic launched user profiles (/v1/user_profiles), which let API customers tag requests by end customer so a safety enforcement action can apply to one end customer rather than the whole organization; profiles now expose enforcement_status and enforcement_actions fields, and beta organization webhooks report restrictions and bans. Source

3. Claude Security Adds Continuous Repository Scanning, Fix Tracking, and an API

Anthropic. Claude Security, in public beta since April, now runs a baseline scan of connected GitHub (including Enterprise Server), GitLab, or other Git repositories and then rescans changes based on their risk and complexity, with projects able to span multiple repositories to catch cross-service issues. Findings are automatically marked fixed when a fix lands and reopen if the issue recurs, and each project gets a knowledge base plus an editable, Claude-generated threat model used to rate severity and explain attack paths. A new API, webhooks, and an MCP connector for Claude.ai and Claude Code let teams manage projects and triage findings programmatically. The beta is open to Claude Enterprise customers, with $25,000 in free monthly usage credits for two months through December 31, 2026. Source

4. HCLTech Becomes a Global Premier Partner and Embeds Claude in HCLSoftware Products

Anthropic. HCLTech, with more than 230,000 employees in 60 countries, joined the Claude Partner Network as a Global Premier Partner and is setting up an Anthropic business unit and center of excellence where forward-deployed engineers build sandboxes for clients to test Claude against their own systems. HCLSoftware becomes an ISV partner, with HCL BigFix, HCL AppScan, and the Actian AI analyst as the first products to embed Claude. Claude and Claude Code are also being added to HCLTech’s AI Force platform for IT support and code remediation, with HCLTech engineers approving any change that touches client production systems. Source

5. Claude Code 2.1.307 Adds an API Version Migration Command and Safety Event Telemetry

Anthropic. Claude Code 2.1.307 adds /claude-api api-version, which moves Claude API code in a project to the new 2026-10-21 API version. A new claude_code.safety_event OpenTelemetry event fires for each tool call that a permission rule, auto mode, safety check, hook, or the sandbox stopped, prompted, or sent back, giving admins a single signal for auditing blocked agent actions. Code Review gains @claude review stop (also off and pause) to end an @claude review always subscription on a pull request, and Claude Code now retries unparseable tool calls up to twice before failing the turn. Source

6. Physicist Christopher Stubbs Joins Anthropic’s Long-Term Benefit Trust

Anthropic. Harvard physicist Christopher Stubbs, a former Harvard Dean of Science, member of the High-Z Supernova Search Team, and member of the JASON defense advisory group, has joined Anthropic’s Long-Term Benefit Trust as its first physicist. The Trust can appoint a majority of Anthropic’s board and advises leadership on decisions involving AI risks and societal impacts. He joins trustees Neil Buddy Shah (Chair), Richard Fontaine, and Ben Bernanke. Source