ChatGPT gets protocol questions wrong, and a confident wrong answer is worse than no answer. Cora is built for this: cited answers in about 10 seconds, grounded only in your documents, with the exact page every time.
Your coordinators are resourceful. When they need a protocol answer at 6pm, they'll find one. The question isn't whether they'll use AI; it's whether the AI they use was built for regulated environments. Cora was.
When your coordinator pastes a protocol into ChatGPT, she doesn't get a warning. The sponsor doesn't get notified. And if the sponsor's legal team ever asks whether the full protocol text was shared with an unauthorized third party, the honest answer is yes. Cora was built so the honest answer is always no.
AI inference runs through AWS Bedrock with zero data retention. Your prompts and answers are not persisted by the model layer.
Your documents are never used to train any AI model. Every query runs through a clinical-research pipeline, not a public chatbot.
Every query, answer, and timestamp is logged and exportable. And we sign a confidentiality agreement and DPA with your site, so you can name us as an authorized vendor under your sponsor agreement. ChatGPT can't be that.
Your coordinators know ChatGPT exists. Some of them are already pasting protocol text into it. Here's why that's a problem, and why Cora is built for this.
Your first study is free. Each additional study is one predictable price. Every coordinator, every question, every amendment. No base fee, no per-seat fees, no query limits, no surprise charges.
Founding seats are full for now
We're focused on getting these sites running well before we open the next ones. Join the waitlist and you'll be first in line when they do.
Your first study is free. $0/year
Your annual exposure without Cora
Your first study is free. $300/year for each study after that. Founding seats are full for now. Join the waitlist for the next opening.
Running 5 studies? Your first is free, so that's $1,200/year ($100/month) for the other four. Every answer comes with page-cited sources your coordinators can verify in seconds.
Add a paid study and change your mind within 30 days of your first paid study? Every dollar on that study back. No questions, no process.
Upload a protocol and run at least 10 substantive queries. If Cora has not saved your coordinators meaningful time within 90 days of your first paid study, we refund your paid studies for the year.
Every query your team runs is logged to Cora's dashboard. You see questions asked and answer quality in real time, not as a billing surprise.
No IT department. No training sessions. No 6-week implementation. Setup takes less time than a coffee break, and once you're live, finding an answer takes about 10 seconds instead of 20 minutes.
Drag and drop your protocol, lab manual, ICFs, pharmacy manual, and any supporting documents. Cora processes each one in 2-4 minutes. No reformatting. No tagging. Just drop the PDFs.
Type like you’d text a colleague. “What are the exclusion criteria for renal function?” “When is the Day 15 visit window?” Cora searches every document in your study bundle at once.
Every answer comes with confidence scores, page-number citations, and clickable links to the source. Low-confidence answers are flagged automatically. Verify it yourself in one click.
Protocols aren't documents you skim. They're 300-page decision trees: stacked amendments that override each other, dosing matrices with footnotes, visit schedules that cross-reference six other sections. A general LLM reads the words; it doesn't parse the structure. So when your coordinator asks ChatGPT about the exclusion criteria and gets an answer that Amendment 3 already superseded, she doesn't know it's wrong. She answers the PI confidently. She replies to the sponsor query. She trains the new hire on it. Cora's pipeline is built for that structure: tables, amendments, cross-references. Every answer is grounded in an exact page from your documents, so when it's right you know why, and when the protocol changes, Cora changes with it.
Tables, dosing matrices, stacked amendments, cross-referenced visit schedules. Cora parses the structure a general LLM flattens, so retrieval is right before the answer is written.
Every answer is verified against the source text before your team sees it. Low-confidence answers are flagged, and every claim links to the exact page.
Up to 30 documents per study bundle plus 50 site SOPs. One question searches all of it at once.
Cora started because our co-founder was a research coordinator who got tired of putting screening calls on hold to spend 20 minutes hunting for one answer in a 300-page protocol PDF.
We didn't build a generic AI chatbot and slap “clinical” on it. We built the tool that would have saved us from amendment chaos, from juggling five protocols across different tabs, and from being too embarrassed to message our study manager the same question twice.
Every feature in Cora exists because we lived the specific pain it solves. That's why it works.
CEO & Co-founder
Former clinical research coordinator and site support specialist. Lived through the answer-hunting, built Cora to give sites the tool she wished she had.
CSO & Co-founder
Software developer and AI harness engineer specializing in regulated environments. Builds the infrastructure that keeps Cora fast, accurate, and secure.
Request early access. We'll reach out when your spot is ready.
Your data stays secure. Zero data retention on the AI, no training on your documents, full audit trail from day one.