# AIVory

AIVory - From all-purpose AI assistant to code compliance tool. A journey of building, pivoting, and finding the real problem worth solving.

From AI Butler to Code Compliance Tool What started as a simple development pipeline somehow became the most complex project I’ve ever built. AIVory began life as an internal tool – just a way to streamline some repetitive coding tasks. But then we added a productivity feature here, a communication helper there, and before we knew it, we had something entirely different on our hands.
The Accidental Evolution I remember the exact moment everything clicked with the first version. I was drowning in my usual morning email avalanche when our prototype casually said, “You have 47 new emails, but only 3 need immediate attention.” That’s when I realized we weren’t building a development tool anymore – we were solving information overload.
The technical challenges were fascinating. Building a multi-agent system where different AI models collaborate behind a single interface? That’s not exactly well-documented territory. Each agent needed to be specialized – one for email analysis, another for document processing, another for chart generation – but they all had to work together seamlessly.
The real game-changer came when we added voice control. Being able to say “Hey Siri, ask AIVory for an update” while walking to a meeting and getting a perfect summary of what I’d missed – that felt like magic. We spent months getting that natural conversation flow right.
We launched in May 2025 after 18 months of development. Full multi-tool calling agent, mobile apps, complete data separation per client, works with self-hosted LLMs. The tech was genuinely cool.
The Part Where It Didn’t Work But here’s the honest truth: our focus group didn’t understand it. They wanted “single-product” solutions – replace this specific role in my company. Our all-purpose assistant was harder to communicate. And by the time we launched, Microsoft had published their “Work Trend Index 2025” describing precisely our product. Big players told us they’d been working on the same thing for months.
We couldn’t compete. That hurt.
The Pivot Since May, Jenni and I have been testing different concepts. And I learned something from actually using AI to code: it makes you faster AND lazier. When you’re getting hundreds of lines every few minutes, you stop inspecting. You glance, accept the gray areas, move on.
That cost me weeks of fixes before our original launch. Weeks I could have spent learning if anyone actually wanted what we built.
What bugged me: no one is tackling this problem. The software that exists is either management dashboards or generic linters. What we actually need is real-time inspection AS the AI writes code. A Cloudflare for code – catch issues before they hit your codebase, let AI fix its own mistakes on the fly.
So we pivoted. Hard. Took everything we’d learned about multi-agent AI systems and pointed it at code compliance.
What It Does Now AIVory scans your code in real-time as you write – whether you’re typing it yourself or letting Copilot/Cursor/Claude generate it. GDPR, HIPAA, SOC 2, PCI-DSS, ISO 27001, OWASP Top 10, CCPA – the whole regulatory alphabet soup.
The killer feature is one-click fixes. It doesn’t just tell you what’s wrong, it shows you exactly how to fix it. Install the plugin, 30 seconds of setup, and you’re catching compliance issues before they hit your codebase.
We kept the multi-agent architecture from the original AIVory – different models specialized for different compliance frameworks, all working together. But instead of analyzing emails and documents, they’re scanning code. Turns out that’s significantly more complex. Each framework has different requirements, different edge cases, different contexts where the same code might be compliant or not.
Getting that real-time scanning fast enough without destroying IDE performance took months of refinement. But it works.
Why This Actually Matters AI coding assistants write functional code fast. They don’t always write compliant code. They’ll give you something that works perfectly, compiles cleanly, and violates three different regulations. By the time you catch it in an audit, fixing it costs 30-50x more than if you’d caught it while writing.
AIVory is the Cloudflare for code I wish I’d had when building the original version. Catch issues early, let AI fix its own mistakes, less noise.
Try It The JetBrains plugin is live. Free tier gives you OWASP Top 10 scanning with 100 scans per day. Premium is $39/month if you need more. VS Code and Claude Code Agent integration coming soon.
Building AIVory has been the most interesting journey I’ve had in years. From development pipeline to information assistant to compliance tool – each evolution taught us something. This latest version feels honest. We’re solving a problem I actually experienced, not one I theorized about.
Check it out at aivory.net.