AI integration
Assistants that answer only from your own content, agents that act through your APIs, and gated MCP servers so an AI can work inside your ERP, CRM or accounting — on rails, logged, within limits you set.
Two decades of solutions consulting under one roof. APIs, integrations, and the enterprise plumbing behind them — designed, built, and quietly maintained. One contract, one Architect, one accountable point of contact.
Most of what pays for itself isn't a chatbot on your website. It's plugging an assistant into the systems you already run — including the fifteen-year-old one nobody wants to touch. A custom MCP server is, in plain terms, a small set of locked doors into that old system: the AI can look up what you allow and do the few things you allow, nothing else, and every move is logged. No migration, no rip-and-replace.
today Two hundred mails a day. Twelve of them matter. You find out which ones by reading all two hundred.
after An assistant sorts the real requests from the noise, files the rest, and leaves the standard replies drafted for you to approve or rewrite.
today Quotes, stock and client history live in software from 2009. Two people know how to drive it, and both of them are busy.
after A custom MCP server gives an assistant a narrow, logged set of doors into that old system. "Is PO 4412 shipped?" gets answered in seconds, by someone who never learned the screens.
today Every quote is a template someone copies, edits, gets wrong, and sends two days late.
after "Quote Acme, forty hours, the usual rate" becomes a real draft in your own system — numbered, taxed, ready for you to check and send.
today Manuals, contracts and price lists in a shared folder. When someone needs a number, they call you.
after Staff and customers ask in plain language and get the right answer with the source attached — or an honest "not in the documents", never a guess.
If none of these are your problem, that's fine — you have a different one. Thirty minutes is usually enough to tell whether AI actually helps here, or whether it would just be expensive theatre.
Book a quick call →One contract, one calendar, one accountable pair of hands. When a project needs a specialist, I bring them in and run the room.
Assistants that answer only from your own content, agents that act through your APIs, and gated MCP servers so an AI can work inside your ERP, CRM or accounting — on rails, logged, within limits you set.
REST, GraphQL, webhooks, event-driven. Pragmatic, documented, debuggable. Systems that connect and stay connected.
ERP, CRM, and core platform integrations. Less duct tape. Fewer vendors-in-the-middle. Workflows that actually measure.
Tailored applications when off-the-shelf doesn't fit. Web apps, internal tools, Shopify storefronts, RAG-backed assistants.
Decades of pattern recognition for hire. The right technology, sequenced the right way, with failure modes already mapped.
High-availability infrastructure built for sensible scale and easy maintainability. Long-running systems that don't wake you at 3am.
The bias is toward technology that's still around in five years. Modern enough to be productive, mature enough to be quiet at 3am.
Systems that don't just run tasks but anticipate them. RAG pipelines, embeddings, retrieval over private data, agents on rails. Routine work becomes leverage.
New — packaged AI chatbot builds. See the offer →Software adapts to the business, not the other way around. Tailored web apps, internal tools, Shopify storefronts — built for flexibility and a long tail.
Show up to the right people at the right time. AdWords creation and management, continuously optimized for visibility, qualified leads, and ROAS.
On-chain programs in Rust with Anchor, paired with TypeScript clients and SDKs. Token mints, staking, escrow, and custom instructions — tested on localnet, deployed to mainnet, observable end-to-end.
Tools that came out of client work and earned their own page.
A chatbot that answers only from your own documents — citations, refusals, French and English, client dashboard with usage, limits and cost control.
product page →The schema-driven engine behind every APIgoat application — describe the data model, get the database, admin, API, permissions and MCP server. The next project starts further along.
how it works →Self-hosted grid and trend bot with paper mode by default, server-side risk rails, and an AI agent that re-fits the grid hourly — inside limits it cannot change.
product page →Rémi, a real-time voice tutor that teaches French to 9–12-year-olds, remembers each learner between sessions, and gives parents transcripts, progress and spend limits.
apigtutor.apigoat.com →A focused crypto companion for tracking, alerting, and answering the simple questions traders actually ask.
cryptobboy.apigoat.com →Lightweight invoicing built to disappear into the rest of your workflow. No SaaS bloat, no monthly minimum.
billboy.apigoat.com →This site uses Google Analytics to count visits. It loads only if you accept. Details in the privacy policy.