Daily AI Builder Brief - May 1, 2026

Executive Summary

Anthropic launched Claude Security public beta on April 30 for Enterprise, powered by Opus 4.7.
Claude Security adds scheduled scans, documented dismissals, CSV and Markdown exports for audit pipelines.
Six security vendors (CrowdStrike, Microsoft Security, Palo Alto, SentinelOne, TrendAI, Wiz) are integrating Opus 4.7 into their platforms.
Claude Code 2.1.126 shipped today with /model gateway discovery and a new claude project purge command.
Claude Code 2.1.123 (April 29) fixed an OAuth 401 retry loop tied to the experimental betas flag.
MCP roadmap names four production priorities: Streamable HTTP scaling, Server Cards, Tasks primitive, enterprise readiness.

Claude Code

2.1.126 adds gateway model discovery and project purge

/model picker now reads from a gateway's /v1/models endpoint when ANTHROPIC_BASE_URL is set. New claude project purge [path] wipes transcripts, tasks, file history, and config with --dry-run and --interactive flags. OAuth login accepts pasted codes for WSL2, SSH, and container setups. Windows PowerShell 7 detection improved. (changelog)

2.1.123 fixes OAuth 401 retry loop

Resolved an authentication failure when CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 was set. (changelog)

Anthropic

Claude Security public beta opens to Enterprise

Announced April 30. Runs on Opus 4.7. Traces data flows across the codebase rather than pattern matching, validates findings to cut false positives, and proposes reviewable patches. Team and Max tiers coming, no date yet. (SecurityWeek, SiliconANGLE)

Security platform integrations expand

CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz are embedding Opus 4.7 into existing enterprise security stacks. Accenture, BCG, Deloitte, Infosys, and PwC are rolling out Claude-integrated services for vulnerability management and incident response. (SecurityWeek)

MCP Ecosystem

2026 roadmap names four production priorities

Maintainers ranked Streamable HTTP scaling (stateful sessions vs. load balancers), Server Cards (.well-known metadata for discovery), Tasks primitive maturity (retry semantics, expiry), and enterprise readiness (audit, SSO, gateways) as the focus where SEPs get expedited review. (MCP Blog, The New Stack)

Broader AI

Nothing today.

What This Means For Builders

If you ship code, Claude Security plus Opus 4.7 is now a viable codebase scanner inside the Anthropic stack. Worth piloting against your current Snyk or Semgrep workflow.
The claude project purge command means you can finally hand a clean Claude Code state to a new contributor or wipe a repo's local history without surgery.
If you're building MCP servers, plan for Server Cards and stateful Streamable HTTP. Both will affect how your server gets discovered and scaled.
Gateway model discovery in /model lets teams run a private model proxy and surface only approved models to Claude Code users.