Migrating a Workflow to Claude Code Agents Safely
A phased playbook for moving an existing workflow onto Claude Code agents — shadow mode, human-in-the-loop, scoped autonomy, and fast rollback at every stage.
6 min read
Agentic AI & LLMs
Agentic AI & LLMs archive page 54 of 631
Agentic AI & LLMs
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
Latest articles
Evals for Claude Cowork: Gating a Sales-Book Release
Build an eval loop for Claude Cowork: golden sets, scored runs on correctness and cost, and release gates so a 4,000-account workflow never regresses.
7 min read
Testing and Evals for Claude Code Agents That Ship
Build an eval loop for Claude Code agents — real test sets, exact and LLM-judge scoring, trajectory metrics, and a CI gate that blocks regressions before release.
7 min read
Securing Claude Code Agents: Sandboxing & Least Privilege
Harden Claude Code agents with sandboxing, least-privilege tools, secret handling, and prompt-injection defense in depth — contain the blast radius safely.
7 min read
Securing Claude Cowork: Sandboxing & Least Privilege
Harden Claude Cowork on a real sales book: sandboxing, least-privilege connectors, secrets handling, and prompt-injection defense for agentic automation.
7 min read
Cut Claude Code Token Costs: Caching and Batching
Cut Claude Code agent costs with prompt caching, batching, context hygiene, and model routing — keep multi-turn runs cheap and fast without losing quality.
7 min read
Cutting Claude Cowork Token Cost on a 4,000-Account Book
Performance tuning for Claude Cowork: prompt caching, batching, model routing, and context discipline to run a 4,000-account book cheap and fast.
8 min read
Debugging Claude Code: Loops, Bad Tool Calls, Fixes
Why Claude Code agents loop, pick wrong tools, or hallucinate arguments — and the concrete guardrails, schemas, and observability that fix each failure mode.
8 min read
Debugging Claude Cowork: Loops, Bad Tool Calls, Fixes
Diagnose and fix the common Claude Cowork failure modes at scale: runaway loops, wrong tool calls, and hallucinated arguments across a 4,000-account book.
8 min read