Writing

Notes from the work

Practical essays on shipping Laravel, building AI agents and MCP servers, and running a one-person consulting business — drawn from what I actually ship.

Can I Give My Claude Agent Access to My Production Database Without Building an API?

Yes — an MCP server gives Claude live, read-only SQL on prod without a REST API. The part that keeps it safe is a database grant, not a read-only flag. Tested.

How to Make Your Site Usable by AI Agents: llms.txt, MCP, and GEO

My llms.txt documented an endpoint that never existed. What actually makes a site usable by AI agents — llms.txt, MCP, GEO, and the checks that fail silently.

Reliable Structured Output from LLMs in Laravel (and the One Line That Saves You)

Typed JSON from Claude in a Laravel app using the AI SDK's schema() method — enums, nested objects, arrays — and the gateway edge case that hands you a JSON string where you expected an array.

Who Pays for the Tokens? Metering & Billing AI Agents in Laravel

The part nobody writes about: per-customer token metering and billing for AI features in a Laravel app. Real production patterns with the Laravel AI SDK + a Stripe LLM gateway — middleware cost attribution, provider override, graceful degradation.

How Much Does a Fractional CTO Cost in 2026? (Real Rate Ranges by Stage)

What fractional CTOs actually charge in 2026 — hourly ranges, the retainer floor nobody advertises, and how to tell a fair rate from an agency markup before you sign.

Integrating Claude into Laravel in Production (Laravel AI SDK + MCP, No Python)

How I run tool-calling Claude agents and MCP servers inside a Laravel app — real production code, no separate Python stack. Laravel AI SDK + laravel/mcp.