Anthropic AI Updates: October 6, 2026

1. Claude Code 2.1.290 Adds Managed Agents Onboarding and More Hook Data for Mods

Anthropic. Claude Code 2.1.290, tagged on October 5, adds /claude-api managed-agents-onboard, which takes either a URL describing a Managed Agents pattern and turns it into ant apply files, or a Console quickstart name such as deep-researcher and builds it with the ant CLI. For mod and plugin authors, a mod’s turn.step hook now receives serverToolUses (tool calls the API ran itself, such as the advisor), plugin tool.check events carry an agentId to distinguish subagent permission checks, and a new ceiling field names the approval level an organization requires for a tool. The release also adds claude attach <name> and claude logs <name> with partial session-name matching, and claude plugin validate now flags gating hooks that lack a .catch. Source

2. Cresta Built Its Conductor Agent Builder on the Claude Agent SDK

Anthropic. In a case study published October 5, Anthropic described how contact-center AI company Cresta ($100M+ ARR) built Conductor, a meta-agent that walks teams through designing, implementing, evaluating, and optimizing customer experience agents. Conductor uses the Claude Agent SDK as its execution harness for gathering context, calling tools, and writing code, with Cresta’s CX workflows layered on top. Cresta reports that Conductor cut initial deployment time roughly in half in early use cases, and notes the initial build is about 20 percent of the work, with testing and optimization making up the rest. Source