About Recon
A company is more than what it knows. It runs on a way of working: the decisions, the context, the way it handles a problem and treats its people, most of which lives in a few people’s heads and a thousand scattered threads. Recon is the verified record of that, built so your team and your AI tools can know how your company actually works instead of guessing.
Why we are building it
For most of a company’s life, how it works passed from person to person. Now teams are bringing in AI tools and agents to do the work, and that transfer breaks. The most capable model in the world still starts as a stranger guessing at your context. It is the new hire who knows nothing about you.
So we are building the layer that gives people and agents what a new hire spends months absorbing: what the company knows, how it works, what it has decided, and where it is stuck. Recon reads the systems you already run, your docs, conversations, tickets, help desk, code, and database, and makes one record that a teammate reads in Slack and that Claude or ChatGPT read over MCP.
It only counts if it is true
The worst thing a company brain can do is hand someone a confident answer that is wrong. Act on it once, send it to a customer once, and you have burned your own trust in the system. A lot of what people call AI hallucination starts there, with bad context.
So Recon does not just capture, it verifies. Every belief traces back to the exact line it came from. When two sources disagree, Recon keeps both receipts, marks the claim disputed, and supersedes the old version only when the source of truth actually resolves it, with the history kept. We hold ourselves to the same bar: we measure our own record and publish where it still falls short, because a record you cannot measure is one you should not trust.
Read the longer argument: the hard part of a company brain is truth, not captureCapture gets you a company brain. Truth is what makes it usable.
Where this goes
The verified record is the foundation, and it is shipped today. On top of it we are building two things. Skills capture how your people actually work, the plays a team runs again and again. Agents act on what changes in the record and draft the work for you, still holding it for your approval. Memory first, then skills, then agents.
Because every agent reads one shared, verified context layer, they stop paying to re-derive your company on every run. They get faster and cheaper on the work that matters, they can hand off to each other, and they can propose what they learn back into the same record, with receipts and review.
Who is behind it
Recon is built by Pratik, an engineer, in the open. The short version of why: he kept watching real questions about a business route back through the few people who happened to hold the context, and decided that context should live somewhere the whole team, and every AI tool, can read. Follow along on X at @chaibytesai.
Talk to us
We are early. We are onboarding a handful of teams directly and working a lot of this out in the open, so we are not going to hand you a finished pitch. But if any of this resonates, whether you want to put it to work or you just think hard about the same problem, we would genuinely love to talk.