You’ve never done research like this.
Propab is an autonomous AI research platform. Its hosted agents don’t just answer questions — they execute the whole study: planning the work, pulling the data, running the tools, reasoning across millions of papers, and delivering results you can defend.
Your best hours shouldn’t go to the busywork of research.
Swapping the same parameter by hand, run after run, just to see what moves.
Juggling databases and APIs to track down the one dataset you actually need.
The export → convert → upload dance, every time a file format doesn’t match.
Good hypotheses you never test, because there’s no bandwidth to chase them all.
None of that is the science. It’s the tax on it. Propab pays the tax — the retrieving, the converting, the parameter sweeps, the dead-end runs — so your intelligence goes where it actually matters: the questions only you can ask.
One question in. The agent does the rest.
Decomposes it into a testable hypothesis tree — nine hypotheses, ranked.
Generates data — co-sedimentation across six phospho-variants. assay-run-14 · 1,240 rows.
Updates its beliefs as evidence lands — Ser262 accelerates nucleation +0.25; Ser356 ruled out −0.28.
Returns ranked, defensible hypotheses — each with its data and citations.
A new way to work that meets you where you already are.
Every real breakthrough began with a method nobody trusted yet. Autonomous research is that method now. You don’t abandon your tools, your data, or your judgment — the agent plugs into them. You stay the scientist. It becomes the tireless lab that never sleeps.
What a Propab agent can do — today.
Give it a goal. It plans the study, runs the work, and returns defensible results — across science and engineering.
Native access to the sources you already cite. It reaches your data instead of asking you to move it.
The agent calls the right instrument for the job — analysis, simulation, retrieval — without you wiring anything.
It reads what your field reads. No conversion, no export dance — the native formats of real primary data.
It reasons over the published record, not a summary of it — and shows its citations, every time.
Every capability is auditable. You can open any plan, tool call, or citation the agent used.
One goal in. An auditable study out.
It decomposes your question into a research plan you can inspect and edit.
Pulls primary data from 250+ sources across 200+ formats.
Calls the instruments the work needs and iterates on what it finds.
Reads across millions of papers and reconciles evidence with your results.
Results, methods, and citations — auditable from question to conclusion.
The difference isn’t speed. It’s what you can attempt.
When an iteration takes weeks, you test the few ideas you’re surest of. When it takes hours, you test the ones you used to shelve.
- 01Frame the question
- 02Hunt for data across tools
- 03Wire the scripts by hand
- 04Wait on a collaborator
- 05Reconcile the results manually
- 01State the goal
- 02The agent retrieves the data
- 03It runs the tools and iterates
- 04It reasons over the literature
- 05You get an auditable result
How Dr. Lena Okafor ran a three-week study in an afternoon.
Lena studies membrane proteins. She gave a Propab agent one goal: find candidate stabilizing mutations. The agent pulled sequence and structure data from connected databases, ran folding and stability tools, cross-referenced forty years of literature, and returned a ranked shortlist — each candidate with its reasoning and citations. Work that used to mean weeks of coordination across three tools and two collaborators came back before her coffee was cold.
“I didn’t change how I think about the problem. I changed how long it takes to test a thought.”
“It screened 12,000 candidate electrolytes overnight and handed me three worth synthesizing.”
“It traced a regulatory variant across 40 years of papers and lab data in an afternoon.”
“It reran the ensemble with corrected boundary data before the meeting started.”
Research isn’t bottlenecked by ideas. It’s bottlenecked by everything between an idea and the evidence for it — the retrieving, the wiring, the waiting, the reconciling.
We’re building the engine that clears it — so a single researcher can command the output of an entire lab. Retrieval, tooling, parameter sweeps, the dead-end runs: all of it handled in parallel, at a scale no team could staff. You still ask the questions and make the calls — Propab just gives you a thousand hands to test them with. V1 is the first step. We’d like you to take it with us.
Autonomous scientific discovery.
V1 puts a working research engine in your hands. The vision is larger — an engine that doesn’t just accelerate research but makes genuinely novel discoveries on its own, across every domain of science. We’re building the system that closes the loop from question to discovery. This is the first version of that engine.
Change the speed of your research. Not the way you work.
It connects to your data and formats. You migrate nothing.
Every plan, tool call, and citation is inspectable. Nothing is a black box.
Hand it one task. Trust is earned one result at a time.
The questions researchers actually ask.
What is Propab?
Propab is an autonomous AI research platform. It gives researchers hosted AI agents that plan a study, retrieve data from 250+ databases, run hundreds of tools, reason across millions of papers, and deliver auditable results — executing complex science and engineering tasks end to end.
How is it different from ChatGPT or a research assistant?
A chatbot answers questions. A Propab agent executes the work — it runs tools, generates and analyzes real data, and returns results with citations you can audit, not just text.
What can it do today, in V1?
V1 agents execute end-to-end research tasks with native access to 250+ databases and APIs, hundreds of on-demand tools, 200+ scientific data formats, and millions of research papers.
Does it work with my existing data and tools?
Yes. Propab connects to the databases, formats, and instruments you already use. You don’t migrate your stack or change your workflow — the agent comes to your data.
Can I trust the results — does it hallucinate citations?
Every plan, tool call, and citation is inspectable. Propab reasons over the actual published record and shows its sources — auditable from question to conclusion. Nothing is a black box.
When does it launch and how do I get access?
Propab V1 launches in Fall 2026. Access rolls out cohort by cohort from the waitlist — join below to be in the first cohort.
Be in the first cohort.
V1 ships Fall 2026. Early access is limited and rolled out cohort by cohort.
No spam. One email when your cohort opens.