LatestMCP Server Architecture: Hosts, Clients, Servers, and Transports
MCP server architecture in one picture: the host opens one client per server. Stdio for a local process, Streamable HTTP for a remote one, and three things a server can expose.
Services
- Vibe-Built MVPValidate an idea · 3 weeks$15–30K
- AI Product SprintBuild a real product · 4–6 weeks$30–60K
- Production Audit + RescueAgent live but failingFrom $3.5K
- Agentic Product RetainerFractional CPO · 2–4 days/week$15–30K/mo
- AI Agent Development
- LangChain & LangGraph
- LLM Cost Optimization
- AI Automation Consulting
- Forward Deployed Engineer
29 articles · tutorials, comparisons, case studies
The AI agent engineering blog. Field notes from production.
Tutorials and field notes on LangChain, LangGraph, multi-agent systems, RAG, evals and cutting LLM costs, written from real client builds.
All 29 articles, newest first
Latest
File System MCP Server: What It Is and How to Scope It
The file system MCP server is @modelcontextprotocol/server-filesystem 1.0.0. Pass one directory. It can read and write inside that directory, and it denies paths outside it.
October 19, 2026
Postgres MCP Server: Give an Agent Read-Only Database Access
A Postgres MCP server an agent can query but not change: a SELECT-only role, one statement at a time, and the calls I ran on Postgres 16.
October 19, 2026
MCP Server List: The Seven Official Reference Servers
The official MCP server list is seven reference servers. Package versions checked October 19, 2026. The old Postgres server is archived. Pin the version before you add one to a host.
October 19, 2026
MCP Server Examples: A Tool, a Resource, and a Prompt
One MCP server example with a tool, a resource, and a prompt. On mcp 2.3.0, unit_price(250, 3) returned 750, and a call missing quantity never reached the function.
October 19, 2026
How to Build an MCP Server in TypeScript and Python
How to build an MCP server: a minimal Node.js MCP server and a Python MCP server, MCP server examples you can run, and how to test them in MCP Inspector before you add auth.
October 12, 2026
How to Add Authentication to an MCP Server
How to add authentication to an MCP server: your HTTP server is an OAuth 2.1 resource server. It verifies a bearer token, checks the audience, and answers 401. Stdio does not use this flow.
October 12, 2026
MCP vs Function Calling: When to Use Which
MCP vs function calling: use a function call when one process owns the tool. Use MCP when a second host must call it, or the tool needs its own auth and deploy.
October 12, 2026
How to Deploy an MCP Server to Production
How to deploy an MCP server: one HTTPS POST endpoint on Streamable HTTP, a localhost bind until TLS and bearer auth are in front, and a host that can hold the process.
October 12, 2026
LLM API Pricing Comparison: Claude, Gemini, Grok, and OpenAI
An LLM API pricing comparison of Claude API pricing, Gemini API pricing, and Grok API pricing, plus OpenAI, with a cost-per-conversation example. List prices as of October 12, 2026.
October 12, 2026
How Much Does It Cost to Build an AI Agent? (2026 Pricing)
Real 2026 price ranges for building an AI agent, from a $15K MVP to a $100K+ multi-agent system, plus the monthly running costs most budgets forget and how to keep both under control.
September 25, 2026
Agentic AI vs Generative AI: What's the Difference?
Generative AI creates content when you ask. Agentic AI pursues a goal: it plans, uses tools, checks its work and acts. Here's the difference in plain terms, a side-by-side comparison, and how to tell which one your problem needs.
September 25, 2026
How to Reduce LLM Costs: 9 Proven Tactics for 2026
Cut your LLM bill without losing quality. Nine proven tactics — model routing, semantic caching, prompt compression, output caps, and more — ranked by effort versus payoff, with the order I actually use on client projects.
June 22, 2026
AI Agent Examples: 26 Real Use Cases by Industry (2026)
26 real AI agent examples grouped by industry: support, sales, finance, legal, healthcare, engineering, operations and HR. What each agent does, the tools it uses, agentic AI examples with several agents, and when an agent beats a plain workflow.
June 22, 2026
AI Consulting Services: What to Expect & What It Costs (2026)
What AI consulting services actually include, the typical engagement phases, the four pricing models, real 2026 cost ranges, and how to choose between a firm, an agency, and an independent engineer.
June 15, 2026
How to Build an AI Chatbot Agent (Step-by-Step Guide)
A hands-on, code-backed guide to building an AI chatbot agent in 2026: the conversation loop, tools via function calling, short-term and persistent memory, RAG for knowledge, and the guardrails that keep it reliable.
June 15, 2026
How to Build an AI Agent: A Practical 2026 Guide (with Code)
A hands-on guide to building an AI agent in 2026 — the five parts every agent needs, a working code example, the difference between a workflow and an agent, and the mistakes that break agents in production.
June 09, 2026
LLM Evaluation for AI Agents: The Production Stack for 2026 (LangSmith vs Langfuse vs Braintrust)
A neutral, framework-first guide to AI agent evaluation in 2026. Decision tree, comparison matrix for LangSmith, Langfuse, Braintrust, and the four eval layers every production agent needs.
April 27, 2026
Model Context Protocol (MCP): The 2026 Production Implementation Guide
MCP is not a framework, it is a protocol. This is the production implementation guide nobody is publishing yet: server taxonomy, auth patterns, MCP vs function calling, and the failure modes you only learn the hard way.
April 27, 2026
Agentic RAG: 5 Patterns for Self-Correcting Retrieval (2026 Guide)
Naive RAG plateaus around 70% accuracy. Agentic RAG patterns - query rewriting, retrieval routing, self-correction, multi-hop, hybrid scoring - get you the next 25%. Decision tree, code patterns, and real benchmarks inside.
April 27, 2026
LLM Observability for AI Agents: The 2026 Stack Founders Should Actually Buy
Your agent made 14 LLM calls, used 3 tools, and gave the wrong answer. Without observability you spend a day reconstructing it. This is the 2026 observability stack and the four things to instrument first.
April 27, 2026
Voice AI Agents in Production: Vapi vs Retell vs LiveKit vs DIY (2026 Guide)
Sub-700ms voice latency or your agent sounds robotic. Vendor-neutral comparison of Vapi, Retell, LiveKit Agents, and DIY (Twilio + Deepgram + GPT-4o + Cartesia). The latency budget that makes or breaks voice.
April 27, 2026
LLM Routing: How to Cut Costs 60% Without Losing Quality (2026 Strategy)
An LLM router sends easy queries to cheap models and hard queries to expensive ones. Done right, it cuts your token bill 60% with zero accuracy loss. Decision matrix, comparison of OpenRouter, Portkey, and custom builds, plus a real $14K-to-$5K case study.
April 27, 2026
LangChain vs LangGraph: Which to Use in 2026
LangChain vs LangGraph, in one decision: chains for a straight line, LangGraph when the agent must loop, branch, or wait for a person. Includes CrewAI vs LangGraph and LangGraph vs n8n.
April 15, 2026
How I Cut a Customer Support Agent's LLM Cost by 60% (Full Playbook)
A step-by-step LLM cost optimization playbook from a real production engagement. Semantic caching, prompt compression, and intent-aware model routing - with the exact numbers, code patterns, and pitfalls.
April 02, 2026
LangGraph Tutorial: Production Human-in-the-Loop Pattern (2026)
A complete production tutorial for building a stateful LangGraph agent that pauses on high-risk actions, waits for human approval, and resumes from the exact checkpoint - with code, persistence, and battle-tested edge cases.
March 20, 2026
How to Hire an AI Engineer (or AI Developer): The Founder's 2026 Guide
Stop hiring 'prompt engineers'. A practical, founder-friendly guide to hiring AI engineers who ship production systems - with screening questions, red flags, and 2026 pricing benchmarks.
March 05, 2026
RAG vs Fine-Tuning: A Founder's Decision Framework for 2026
When should you invest in fine-tuning versus a RAG pipeline? A practical decision framework with cost benchmarks, real-world examples, latency trade-offs, and a flowchart that gets to the right answer in under a minute.
February 18, 2026
Multi-Agent Systems: The Supervisor Pattern for Production-Safe Agent Swarms
How to architect a multi-agent system where specialized agents collaborate without going off the rails. The supervisor-worker pattern, explained with a real e-commerce example, working code, and production guardrails.
January 28, 2026
Rather have it built? Start with one call.
Book a free scoping callWhat happens after you book
- A 30-minute call about your workflow
- A one-page scope with price and timeline, free
- You decide: build, or keep the scope and walk away