Cut Claude agent token cost: caching, batching, speed (Extending Claude Skills MCP)
Keep Claude agents fast and cheap on Skills and MCP with prompt caching, batching, context discipline, model routing, and cost-per-run metrics.
Agentic AI, LLM engineering, and the models behind modern automation — multi-agent systems, LLM evaluation and comparisons, RAG, fine-tuning, AI infrastructure, security, and production AI engineering.
From the blog
Keep Claude agents fast and cheap on Skills and MCP with prompt caching, batching, context discipline, model routing, and cost-per-run metrics.
Diagnose and fix the common failure modes of Claude agents on Skills and MCP: tool-call loops, wrong tool selection, and hallucinated arguments.
What to put in Claude's context and what to leave out — prompt and context engineering for agents with Skills and MCP tools, with concrete heuristics.
Wire MCP servers into Claude safely — authentication, typed schemas, structured errors, retries, and idempotency for production agentic systems.
Code-level patterns for Claude Skills and MCP tools — thin tools, fat skills, disclosure ladders, idempotency, and composition that scales.
Step-by-step: scaffold an MCP server, pair it with an Agent Skill, and wire both into Claude Code — with real TypeScript code and testing tips.
Trace the full architecture connecting Claude, Agent Skills, and MCP servers — discovery, loading, tool calls, and context flow, end to end.
Sedgwick's Sidekick Agent improves claims processing efficiency by 30%. How agentic AI transforms insurance from intake to settlement.
Agentic AI and AMRs are redefining warehouse operations in 2026. Learn how adaptive agent orchestration drives the smart warehouse revolution.