The original title is: "Claude Agents + Chat SDK, Gemini cuts tokens 17%"

Short summary

Anthropic's managed agent runtime now integrates with Vercel's AI Chat SDK, letting you deploy Claude agents to Slack, WhatsApp, Discord, or web without writing webhook handlers. Gemini 2.5 Flash cuts output tokens 17% versus 2.0 Flash at $7.50/M tokens, while Flash-Lite hits 350 tokens/sec for latency-critical workloads. OpenAI's codex-security npm package adds AI-assisted vulnerability scanning to CI pipelines.

• •Claude agent runtime + Vercel Chat SDK eliminates custom webhook and session management code for multi-platform agent deployment
• •Gemini 2.5 Flash reduces output tokens 17% vs 2.0 Flash; Flash-Lite targets 350 tokens/sec throughput for batch and parallel agent workloads
• •OpenAI codex-security npm package adds AI SAST scanning to CLI and CI with minimal setup

Generated with AI, which can make mistakes.