Frequently Asked Questions
Welcome to the Marvin help center. Below are quick answers to the questions we hear most. For deeper walkthroughs, see getting started and how-to guides.
How does Marvin know what to research?
You give Marvin a question or topic — in plain language — and it takes it from there. There's no script to follow or workflow to configure.
Once you've framed the question, Marvin investigates autonomously: it reads relevant literature, forms hypotheses, runs analyses, and reaches conclusions. When it's done, you get back structured findings you can read, cite, and act on. See what is Marvin for more on how Marvin thinks.
Can I guide Marvin mid-run?
Yes. Marvin runs autonomously, but you're always in the loop and can redirect it at any time.
You can monitor progress live as it works, and if the direction needs to change, you can pause the run, update your instructions, and resume. This lets you steer without micromanaging every step.
Where does the data come from?
Marvin pulls from academic literature databases — primarily arXiv, Semantic Scholar, PubMed, and more.
This means Marvin is strongest on questions grounded in published research. Connecting your own data sources is on the roadmap but not yet available; see Can I connect my own data sources? below.
Is my data secure?
Your work is private to your account. It isn't shared with other users, and it isn't used to train models unless you explicitly opt in.
You stay in control of what Marvin can access and how findings are stored. If you have specific compliance or security questions, reach out via contact & support.
What models does Marvin use?
Marvin is powered by frontier large language models, giving it strong reasoning, reading comprehension, and analysis capabilities.
You don't need to pick a model or configure anything under the hood — Marvin handles that for you based on the task. For more on what this enables, see what Marvin can do.
Can I connect my own data sources?
Not yet — this is on the roadmap. Today Marvin works with academic literature databases.
We're working on letting you bring your own datasets and sources. If you want to be notified when this lands, let us know through contact & support.
How does billing work?
Every plan includes a monthly research capacity, plus a wallet of top-up credits you can add anytime.
You can track usage as a percentage bar in the sidebar. If you go beyond your monthly capacity, additional usage is billed at provider cost with no markup. Credits never expire and are drawn before any overage kicks in. See plans and billing for the full breakdown.
What happens if I run out of credits?
If you run out of credits or capacity, your runs pause — they aren't canceled or lost.
To resume, you can add more credits, upgrade your plan, or wait for your monthly capacity to reset. Any partial work Marvin had completed is preserved, so you pick up where you left off.
What's the difference between Marvin and a chatbot?
A chatbot gives you single-turn replies based on what you type. Marvin does sustained, multi-step research.
That means Marvin can read many papers, form hypotheses, run analyses over time, and return structured findings rather than a one-off answer. It's built for questions that take real investigation to answer well. See what is Marvin for more.
Do I need ML expertise to use Marvin?
No. You frame the question in everyday language, and Marvin handles the investigation.
That said, domain expertise helps you interpret the results — especially when judging how a finding applies to your specific situation. Marvin tells you how confident it is in each finding so you can weigh them appropriately.
Can I review past runs?
Yes. Every project keeps a full history, and you can revisit any past run.
Use the Time Machine iteration control to step back through a run's progress and see how findings developed. This makes it easy to audit conclusions or pick up an earlier thread. See navigating the UI for where to find it.
How do I stop a run?
Hit the Stop Marvin button whenever you want to halt a run.
The run pauses and saves its partial findings, so nothing is lost. You can resume later or leave it as-is — the work done up to that point stays available.
What's a "seat"?
A seat is one concurrently running research project — not one user.
You can have as many total projects as you like, but the number of seats on your plan determines how many can actively run at the same time. See plans and billing for seat counts per plan.
Is there a free tier?
Yes. The free tier lets you chat with Marvin and look up literature — at no cost.
Autonomous research runs aren't included on free; those require Academic, Pro, Team, or Enterprise. The free tier is a great way to explore how Marvin works before upgrading. See plans and billing for plan details.