The PromptFoo.tech blog
Editorial pieces on AI workflows, automation, agents, MCP, and prompt engineering — with a bias toward what actually ships.
What Is MCP? A Practical Guide to the Model Context Protocol
MCP is quickly becoming the USB-C of AI tools. Here's what it actually is, when to use it, and how to get value from it in a week.
n8n vs Make vs Zapier in 2025: Which to Pick
A practical comparison of the three big automation platforms across pricing, AI-readiness, self-hosting, and real-world usability.
Prompt Engineering Fundamentals That Still Matter in 2025
Models got smarter but prompt structure still moves outcomes. Here are the fundamentals that survived every model release.
How to Write an AI Workflow That Doesn't Break in Production
Six lessons from workflows that survived contact with real traffic — retries, schema validation, cost caps, and human gates.
Choosing an LLM in 2025: A Buyer's Guide for Workflow Builders
How to pick between GPT, Claude, Gemini, and open models for real production workflows — not vibes benchmarks.
Building Your First MCP Server: A Practical Walkthrough
A hands-on tutorial for building a working MCP server in TypeScript in under an hour — with the debugging tips nobody puts in the README.
Prompt Engineering Patterns That Actually Work in Production
The prompt patterns we keep reaching for when reliability matters more than cleverness — with the failure modes each one solves.
AI Agents vs Workflows: When to Use Which (and When Not To)
Everyone's shipping agents. Most of them should have been workflows. Here's how to tell the difference before you commit.
n8n vs Make vs Zapier in 2026: Which Automation Platform Fits You?
A pragmatic comparison of the three dominant automation platforms — pricing, AI integrations, self-hosting, and where each one actually wins.
Prompt Caching in Production: The Cheat Code Most Teams Miss
How to use Anthropic and OpenAI prompt caching to cut RAG and agent bills by 60-90% without changing model quality.
Evals That Actually Catch Regressions
Most eval suites give teams false confidence. Here's how to build an eval harness that catches real regressions in production LLM workflows.
Structured Outputs and Tool Use: The Practical Guide
How to get 99%+ valid JSON from OpenAI, Anthropic, and open-weights models — with the failure modes each still has.
RAG That Doesn't Suck in 2026
Hybrid retrieval, reranking, evals, and when to just stuff everything into a long-context model instead.
Self-Hosting LLMs in 2026: When It Actually Makes Sense
The real economics of running Llama 3.3, Mistral, and DeepSeek yourself — and the cases where it beats API providers.
Agent Observability: Debugging LLM Workflows in Production
The tracing, replay, and evaluation patterns that turn opaque agent runs into debuggable systems.
How to Choose an LLM in 2026
A decision tree that maps common workflow requirements to the right model — GPT, Claude, Gemini, Llama, DeepSeek.
Building an AI Workflow That Actually Ships
A 6-week playbook from idea to production for internal AI workflows — the traps that stall most teams and how to avoid them.
MCP Servers Actually Worth Installing in 2026
A curated list of Model Context Protocol servers that earn their place in a Claude Desktop or Cursor config — with the ones to skip.
Cursor vs GitHub Copilot in 2026: The Honest Comparison
Where each tool wins, where each falls short, and how to pick for your team based on how you actually work.
The 8 Cost-Optimization Patterns Every LLM Team Should Know
Practical techniques to cut your OpenAI and Anthropic bill by 60-90% without hurting quality — with the numbers.