🤖 Artificial Intelligence Sep 25, 2026 · Chew Loong Nian - AI ENGINEER

Claude’s New addTools() Can Reuse 98.7% of Your Next Request. Editing tools[] Reuses None.

Towards AI
View Channel →
Claude’s New addTools() Can Reuse 98.7% of Your Next Request. Editing tools[] Reuses None.
Source ↗ 👁 6 💬 0
Author(s): Chew Loong Nian – AI ENGINEER Originally published on Towards AI. Anthropic’s SDK 0.128.0, released with Claude Opus 5.5, can hand the model a new tool mid-run without touching tools[]. It needs one beta flag the runner won’t add for you. If your Claude agent picks up a new tool halfway through a long conversation, the way you add it decides how much of the next request can still come from the prompt cache: 98.7% of it with the new runner.addTools(), none if you edit tools[]. After in

Comments (0)

Sign in to join the discussion

More Like This

The 3 Files That Make Claude Code Much Smarter
Towards AI · 5d ago
System 1 (Jev) Models: Faster and Cheaper Proxy to Frontier Models — With a Working Router
Towards AI · 5d ago
The Number That Matters in Cloudflare’s Clef System-One Model Isn’t 38.8 ms — Jev and Laya compared
Towards AI · 5d ago
What Constrained Decoding Does That Prompting Never Can
Towards AI · 5d ago
Why Is Microsoft Foundry’s Content Filter Blocking Legitimate Medical Questions?
Towards AI · 5d ago
Production RBAC, Cost Optimization, and Deployment Patterns for Cortex Agents
Towards AI · 5d ago