OWL AI Legal Research: Six AI agents working as a paralegal team

Hi Product Hunt — I'm Hobie Cunningham, founder of OWL AI Agency. We didn't set out to build "another legal AI chatbot." We set out to fix a specific failure mode: attorneys and paralegals using general-purpose AI for research, getting fluent prose back, and only discovering at cite-check time that a case name or holding never existed. Courts have sanctioned lawyers for exactly this. That problem kept showing up in conversations with small and mid-size firms — not "AI is bad," but "we need verification built into the workflow, not bolted on after." So we built a **six-agent paralegal pipeline**, not a single prompt: 1. **Research Agent** — pulls from live public legal databases (Cornell LII, Oyez, CourtListener, GovInfo, SupremeCourt.gov) 2. **Precedent Agent** — maps the authority chain (e.g. Katz → Jones → Riley → Carpenter) 3. **Analysis Agent** — IRAC issue spotting and application 4. **Citation Agent** — Bluebook formatting + Table of Authorities 5. **Brief Writer** — court-ready prose 6. **Filing Agent** — assembles the package (memo, brief, certificate of service) Each agent has a narrow job. The output isn't a chat reply — it's a **structured document set** with retrievable source URLs attached. **Try it without signup:** our public demo runs the full pipeline on *Carpenter v. United States* (585 U.S. 946) — a real SCOTUS Fourth Amendment case. Watch six agents work, then read the generated memo, brief, and TOA: https://owl-ai-agency.com/legal-... If you're comparing approaches, we published an honest breakdown of why we don't recommend generic ChatGPT for filing-bound work: https://owl-ai-agency.com/compar... **What we're NOT claiming:** this replaces attorney judgment. Every output is a first draft meant for professional review before filing. What it does replace is the 4–8 hours paralegals often spend on source gathering and initial memo structure on complex matters. **Who it's for:** law firms and legal ops teams that want batch research runs, consistent deliverables, and a private workspace after intake — not consumers asking quick legal questions. **Stack for the curious:** FastAPI backend on Render, static demo on GitHub Pages, agents orchestrated in Python. No proprietary case law database — we query public sources so firms can audit what was retrieved. I'd love your feedback on three things: - Does the demo make the verification story clear in the first 60 seconds? - Is "six agents" compelling or over-engineered-sounding to you? - Would you trust this more if you saw X? (tell me what X is) Happy to go deep on architecture, pricing ($3k / $6k / $12k packages), or the citation verification flow. Thanks for checking us out — and thanks to PH for hosting builds like this. — Hobie https://owl-ai-agency.com | hobiecunningham@owl-ai-agency.com **Topics:** Legal Tech, AI, Productivity, Developer Tools