▸ Blog
Notes from the trenches.
Laravel internals, AI engineering with Claude + MCP, RAG pipelines, Shopify Plus, DevOps. Production lessons from shipped work — every post earns its slot. Featured = hand-picked or most recent.
▸ 100 posts · last updated
Nov 17, 2026
▸ New here? Start with these
Browse series▸ Search the blog
Browse by topic.
▸ Pick a lane · click a card to filter
Latest drop.
▸ Most recent · or pinned by hand
Migrate Storefront MCP carts to UCP Cart MCP
Storefront MCP get_cart and update_cart on /api/mcp are deprecated. Move agents to UCP Cart MCP at /api/ucp/mcp with create, get, update, and cancel.
Read post →
More posts.
▸ 9 of 99 · paginated below
-
ArchitectureMCP first is the new mobile first
Mobile first made every screen work on a phone. MCP first makes every product capability work for an AI agent: tools, skills, and permissions before the human UI.
Read post →
-
LaravelMCP auth and audit logging in a Laravel SaaS
An MCP server on Laravel is only as safe as its tenancy boundary. Scoped tool auth, request-bound actor context, and an audit log you can actually read at 3am.
Read post →
-
AITrust is not a QA strategy: test AI code too
AI code is 95% syntactically clean and fails security tests 45% of the time. Half of agentic PRs ship no tests. The verification pipeline I run instead.
Read post →
-
ShopifyBuilding secure agentic commerce on Shopify in 2026
How agentic commerce on Shopify stays safe: UCP negotiation, an MCP governance layer, scoped credentials, signed requests, and human approval at checkout.
Read post →
-
ArchitectureMCP vs A2A vs ACP: the agent protocol stack
MCP connects agents to tools, A2A connects agents to each other, and ACP has folded into A2A. How the two-layer agent protocol stack fits together in 2026.
Read post →
-
ShopifyThe last Storefront MCP tool is deprecated
ask_business returns a written answer, not the policy sources. Shopify will remove search_shop_policies_and_faqs on 15 January 2027, so stop parsing the old sources array and read result.structuredContent.answer.plain.
Read post →
-
LaravelA throwing Laravel AI tool no longer ends the run. The exception message now goes to the model.
Laravel AI SDK 1.4.0 keeps the agent running when a tool throws, and sends the exception message to the model. Review that text before you upgrade, or rethrow from ToolFailed to keep a hard stop.
Read post →
-
AICloudflare will convert your pages to Markdown for agents. The default signal says they can train on them.
Markdown for Agents is generally available. If your origin sends no Content-Signal, Cloudflare adds ai-train=yes. I ship my own Markdown and state ai-train=no in robots.txt and on the response.
Read post →
-
LaravelYour Laravel MCP server now checks Origin. APP_URL just became a security setting.
Laravel MCP 1.2.0 returns 403 to a browser request whose Origin is not your APP_URL. That makes APP_URL part of your MCP security, and the allow list is a config key that is not in the config file.
Read post →
Tags.
▸ Search by specific tech