Testing and evals for Claude agents: gate every release
Measure Claude agent quality and gate releases with an eval loop — golden sets, deterministic checks, LLM-judges, and trajectory evaluation that catch regressions.
7 min read
Agentic AI & LLMs
Agentic AI & LLMs archive page 157 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
Security hardening for Claude Code agents in production
A practical security playbook for Claude Code agents: sandboxing, least privilege, secrets handling, and layered prompt-injection defense for production.
7 min read
Cutting Claude agent token costs: caching and batching
Slash Claude Code agent costs with prompt caching, model tiering, batching, and lean context. A practical guide to keeping agentic runs cheap and fast.
7 min read
Debugging Claude Code agents: loops, bad tool calls (Non Technical PM Ships App)
Fix the three Claude Code agent failure modes — runaway loops, wrong tool calls, and hallucinated arguments — with trace reading, stop conditions, and grounding.
8 min read
Prompt and Context Design for Claude Code: What to Include
What to put in Claude Code's context, what to leave out, and why — context-design lessons from a non-technical PM who shipped an app in six weeks.
7 min read
Wiring MCP Servers into Claude Code: Auth, Schemas, Errors
How to wire tools and MCP servers into Claude Code: keep auth in the server, write typed schemas, return structured errors, and make every tool idempotent.
7 min read
Claude Code Patterns: Structuring Prompts, Tools, and Context
Reusable code-level patterns for Claude Code: outcome-plus-check prompts, narrow typed tools, layered context, and Skills that made a six-week build reliable.
7 min read
Ship an App with Claude Code: A Step-by-Step Walkthrough
A concrete week-by-week walkthrough of how a non-technical PM built, tested, and deployed a real app with Claude Code and MCP in six weeks.
7 min read
How a PM Ships an App with Claude Code: The Architecture
The end-to-end architecture that let a non-technical PM ship a real app with Claude Code in six weeks: agent loop, tools, MCP, and context grounding.
8 min read