Launching today
Perch AI
Built for work where a wrong answer has consequences
6 followers
Built for work where a wrong answer has consequences
6 followers
A verifiable AI coworker for legal, finance, research, and coding. It reads your files, does the work, and shows you the source behind every claim.
Hey Product Hunt, I wanna introduce Perch.
I'm a law student at Emory, and this started with a problem I kept running into. AI tools would confidently invent case citations that did not exist. In my finance and operations internships I watched the same thing happen with numbers. That is fine for a rough draft and genuinely dangerous the moment the work has consequences, like an audit, a financial model, or a legal memo. Everyone I knew either avoided AI for real work or used it and then re checked everything by hand, which erased the time it was supposed to save.
So I built Perch: an AI coworker that shows the source behind every claim. It reads your actual files, does the work, whether that is forensic analysis, spreadsheets, research, or code, and cites where every number and statement came from, so you can verify it in one click instead of re doing it yourself.
A few things that make it different:
• Verifiable by default. Every answer traces back to a source in your files, and numbers are computed, not guessed.
• Works where you work. Web, a Mac Desktop app that runs on your approved local folders, and a CLI.
• Built for consequential work. Legal, finance, and research, not just quick chat.
I built the whole thing myself, from the engine to the models, and wrote up the failure modes and design principles behind it at perchai.app/research if you want to go deep.
It is free to start, no credit card, and Pro is $10/mo during early access. I'll be here all day and I would genuinely love your honest feedback, especially if you work in law, finance, or research. What would it take for you to trust an AI with real work?