Blog
Engineering notes.
Cases, decisions, and what we learned shipping AI agents and data platforms to production.
- financial
AI inside the walls: building agents when the data cannot leave
Fintechs and financial services can't send customer data to a SaaS AI tool. What actually changes in the architecture, and the three problems that survive the constraint.
Read post - agencies
Reporting eats your agency's month. Here is what we build instead.
Five business days a month building decks nobody reads, briefings in five different formats, and a CPA nobody can explain. Three agency bottlenecks and the software that removes them.
Read post - retail
Three reasons your retail data never gives you the right number
Scattered SaaS, rigid BI, and manual operations. The three failure modes we hear from every Brazilian retail operator, and what we actually build for each one.
Read post - whatsapp
The WhatsApp agent that becomes your operational spine
A configurable WhatsApp agent that integrates with your data, runs 24/7, works for customers or for internal teams. The foundation your company should have had in 2020.
Read post - financial-services
Software Zechim builds for financial services
Five AI agents for financial operations: reconciliation, conversational anti-fraud, contextual customer service, auditable compliance, regulatory reporting.
Read post - marketing
Software Zechim builds for marketing agencies
Five AI agents that attack what consumes time inside an agency: reports, creative, briefing, account audit, competitor monitoring.
Read post - retail
Software Zechim builds for Brazilian retail
Five AI agents that make a real difference in an e-commerce, marketplace, or physical-store operation. No hype, integrating with the stack you already run.
Read post - opinion
Brazilian startups are burning Anthropic credit for nothing
I audit a lot of AI-startup Anthropic bills. Three patterns show up in almost every one. None are about the model; all are about operation.
Read post - mcp
Authenticating a remote MCP server when the client can't send a header
Claude Desktop and Cursor send an Authorization header. The Claude.ai web Connectors panel only lets you paste a URL. Here is the query-string key fallback and how to make a credential in a URL safe to ship.
Read post - vercel
Blocking vulnerability scanner bots on Vercel and Next.js
Vercel logs full of HTTP 500s for paths like /wp-Blogs.php that don't exist on the site. The fix uses two layers: Vercel Firewall at the edge plus Next.js middleware. Root cause and patterns to block.
Read post - mcp
MCP vs custom SQL chatbot: when each one wins
Why we chose MCP for the public Acme Store demo, and when a proprietary SQL chatbot still makes more sense. Criteria, not religion.
Read post - claude
How we cut Claude inference cost 60% with model routing
Haiku for the tool loop, Sonnet only for the final synthesis. Aggressive prompt caching. Real numbers from what changed in Zechim's demo cost.
Read post - manifesto
Why we don't do AI without a data foundation first
Most AI projects don't fail because of the model. They fail because of the data. Why we insist on building the foundation before the agent.
Read post