Terms of Service

Version 1.10 · Effective 23 September 2026 · Last updated 2 November 2026

These terms are the agreement between Thalient Labs Inc. (“Thalient Labs”, “we”, “us”) and you, the person or organisation using Hyperresearch. They cover the website at hyperresearch.ai, the console at console.hyperresearch.ai, the API at api.hyperresearch.ai, the hosted MCP server, Hyperresearch Answers at answers.hyperresearch.ai, and everything reachable through them (together, the “Service”).

By creating an account, starting a run, or calling the API with a key we issued, you accept these terms. If you are accepting them for a company, you are saying you have authority to bind it, and “you” means that company.

1. Who we are

Thalient Labs Inc. is a Delaware corporation, incorporated 16 March 2026.

Mailing address: 1001 S. Main St., Suite 500, Kalispell, MT 59901, United States. Representative: Jordan Gibbs. Email: support@hyperresearch.ai.

2. What the Service does, and what it does not

You give us a research question. We run an automated pipeline: it searches the web and scholarly indexes, fetches and reads the pages it finds, and writes a report with numbered citations back to the material it used. Alongside the report you get the source list, the extracted claims, a check of the report’s cited claims against their sources (on a Deep run, a verification record: citation checks, quote integrity, a retraction sweep and an independence audit) and the run’s intermediate artifacts. Large language models from third-party providers do the writing and the judging. We also offer a workspace vault that keeps what your runs read, a hosted MCP server that exposes runs and the vault to your own agents, and standalone verification endpoints that check citations in documents we did not write. Hyperresearch Answers is a separate service, paid for per call from prepaid credit (§4): its ask tool answers one question from sources it quotes, or says it could not find or confirm an answer, and its search tool returns ranked sources with quoted passages.

Three things follow from that, and they are the most important sentences in this document:

• Reports are machine-generated and can be wrong. A model can misread a source, cite a page that does not say what the sentence claims, miss the decisive document, or state something confidently that is false. Our verification step catches some of this and is not designed to catch all of it. Check anything you rely on against the cited source.
• Nothing the Service produces is professional advice. It is not legal, medical, financial, tax, investment, engineering or safety advice, and no lawyer-client, doctor-patient, fiduciary or other professional relationship is created by your use of it. Do not use a report as the sole basis for a decision that needs a licensed professional.
• We read public web pages; we do not break into anything. The pipeline fetches pages that are publicly reachable. It does not log in on your behalf, does not defeat paywalls, and does not bypass access controls. See §7.

Sign-ups are open, and §7 and §13 say when we may suspend or close an account. Features move, and the prices described in §4 are the prices as of the effective date above. The Deep tier additionally requires a payment method on file before a run will start, because it commits hours of compute before there is anything to show you — including the Deep run included at no charge when you add one.

Reports are research summaries, not professional advice

A Hyperresearch report is a research summary. AI models write it from the sources the pipeline found and read (public web pages, licensed articles and anything already in your vault), and it may be incomplete or out of date. It is not medical, legal, financial or any other professional advice, even when it is about your own situation.

• Talk to a qualified professional before you act on a report. For a question about your health, that means a doctor or pharmacist; for a legal question, a lawyer; for a money question, a qualified financial adviser. Do not start, stop or change a medication because of a report.
• A report does not create a professional relationship. Nothing in a report, and nothing about using the Service, makes us your doctor, lawyer, financial adviser or other professional, or creates a professional relationship between you and us or anyone a report cites.
• In an emergency, call your local emergency number. If you or someone else may be in danger, or you think it is a medical emergency, do not wait for a report or rely on one.

A report on a health, legal or money question may open with a short notice that says the same. This section applies to every report, whether or not it carries one.

3. Your account

You need an account to use the Service. Sign-in is handled by Clerk (see the Privacy Policy): an emailed one-time code, a passkey, or a Google or GitHub account. A console session lasts 14 days from its last use and at most 30 days from sign-in, after which you sign in again. You are responsible for everything done under your account and under any API key issued to your workspace, and for keeping those keys secret. API keys are shown once at creation and stored by us only as a salted hash — we cannot recover one for you, only revoke it and issue another. Tell us at support@hyperresearch.ai if you believe a key or a session has been compromised.

You must be at least 18 years old, or be using the Service on behalf of an organisation that authorises you to do so. The Service is not for children and we do not knowingly collect data from them.

A workspace can have several members. Whoever owns the workspace controls who can see its runs, reports and vault, and can remove members. If you invite a colleague into your workspace you are giving them access to the research in it.

4. Prices, runs and payment

Runs are sold at a flat price per run, in US dollars, in two sizes. The prices as of the effective date above are:

Item	Price
Light run	$9 per run
Deep run	$49 per run

Current prices are also shown on the pricing page at hyperresearch.ai/pricing and again at the point of purchase.

A run’s price is fixed when the run is created and shown to you before you start it.

Reruns. Restarting a finished run from a later step costs a percentage of the run price, banded by where it re-enters: 25% from step 12 onward, 50% from steps 8–11, and 100% below step 8 — because re-entering early means doing substantially all of the work again.

Metered add-ons. Some things are billed by use rather than by the run: scholarly lookups, vault fetches, verification pairs, and vault storage per gigabyte-month. Pay-as-you-go includes an allowance before metering starts: 500 scholarly lookups and 5 GB of storage per month. The console shows your current usage and the allowance it counts against under Usage & billing.

Included run. Add a card and your first Deep run is included at no charge. It is granted once per account and once per card, the card must stay on file to use it, and we grant a limited number per day, so one may not be available immediately. Light runs and all other runs are charged at the prices in §4. A new account cannot start runs until a payment method is on file; accounts opened before 15 October 2026 keep any included Light runs they have not yet used, and while such an account has no payment method on file it runs one run at a time and is limited to a set number of runs per day. We may withhold included runs where we reasonably believe an account was created to collect them repeatedly.

How payment works. Stripe Inc. processes all payments. Card details are entered on Stripe’s own pages and are never sent to, or stored on, our servers — we hold a Stripe customer reference and nothing more. Usage is recorded in our own ledger in whole cents and reported to Stripe as meter events, so your invoice equals our ledger. Taxes are added where we are required to collect them.

Late or failed payment. If a charge fails we may retry it, and we may suspend runs on the account until it clears. Suspension for non-payment does not delete your reports or vault.

We may change prices. A change to run or add-on prices applies to runs you start after we publish it.

Answers and prepaid credit

Hyperresearch Answers is paid for per call from prepaid credit in US dollars, not by the run. This subsection covers it; §5 covers runs.

Your balance. Credit belongs to your account: every workspace and key on the account draws from the same balance. It pays for Answers calls only. The console shows your balance and the calls it paid for.

Adding credit. You add credit through Stripe Checkout, and the smallest top-up is $10. Checkout saves your card with Stripe so you can top up again without entering it.

Auto-reload. Auto-reload is off unless you switch it on. While it is on and your available balance falls below the threshold you set, we charge your saved card the amount you set, up to a monthly limit. You can switch it off at any time in the console or through the API.

Free credit. A new account may receive a one-time free credit to try Answers, in the amount shown on the site. It is limited to one per person and one per account, and we grant a limited number per day, so one may not be available immediately. Free credit has no cash value and cannot be transferred or refunded. We may withhold it, or remove any unspent part of it, where we reasonably believe it is being collected through more than one account or otherwise abused.

What a call costs. Each call is charged at the price published for it on hyperresearch.ai at the time of the call. Before a call runs we hold the most it can cost from your balance; when it finishes we charge as follows and release the rest:

• an ask that returns an answer, or that shows the question rests on a false premise, is charged its price. If you allow a call to search harder and it does, it is charged the higher price published for that, and you can cap what any one call may cost;
• an ask that ends without an answer it can stand behind (it found nothing, or could not confirm what it found) is charged a reduced amount: at least one fifth of its price, more when the search cost us more, and never more than its price;
• a search that returns sources is charged its price. A search that finishes with no sources is free if it cost us less than half a cent, and otherwise is charged one and a half times what it cost us, rounded up to the cent and never more than its price.

When nothing is charged. A call that fails on our side (an error, a timeout, or our engine unreachable or at capacity) costs nothing, and its hold goes back to your balance. Nothing is charged for a call we refuse because your available balance is below its price, because too many of your calls are already running, or because your balance is frozen. A call still unfinished after ten minutes is released at no charge. A retry sent with the same Idempotency-Key returns the first result without charging again.

Expiry and closing your account. Credit does not expire while your account is open. When your account is closed, any credit left on it is forfeited, unless the law requires otherwise. If we discontinue Answers, or you close your account because you do not accept a change to these terms (§16), we will refund the unused credit you paid for.

Refunds. Credit you paid for is not refundable, except for a billing error or where the law requires it. If a top-up is refunded, or disputed with your card issuer, its amount comes off your balance. While a dispute is open, or if a refund leaves the balance below zero, the balance is frozen and calls are refused until it is resolved.

Answers results. What §2 says about reports applies to Answers results too: they are machine-generated, can be wrong, and are not professional advice. Check anything you rely on against the quoted source. Answers results are made largely of third-party material, and §6 (“Third-party material and our providers”) says how you may use them.

5. Refunds

We would rather refund a run than argue about it, and most of this is automatic.

Automatic, no need to ask. When a run fails because of us, the charge is reversed without you doing anything. That covers a run that was admitted and never started; a run that started and then stopped making progress until our watchdog ended it; a run interrupted by a platform restart; and a run that died because a model provider refused us, ran out of credit, rejected the request we built, or returned output we could not use. Any included run you spent on such a run is handed back to you. The run page says in plain words what happened and whether you were charged.

Partial charges by where a run stopped. If a run fails for a reason that is not ours and stops before the report is written, you are charged the Light price if it got past the reading stage and nothing at all if it did not. If the report was written, the run price stands.

Stopped by you. A run you abort is billed by the same by-where-it-stopped rule. We do not refund a run you cancelled because you changed your mind about the question.

Quality complaints. We do not automate these, and we do not hide behind “the output is generated text”. If a report is unusable — citations that do not support their sentences, a question plainly not answered, sources that are not real — write to support@hyperresearch.ai within 30 days with the run id. A person reads it, and the normal outcome is a refund or a credited rerun. This is our policy, not a warranty: §9 still describes the limits of what we promise.

What is not refundable. Usage-metered add-ons you actually consumed (lookups performed, chat turns taken, storage used) and runs whose reports you have used.

6. Your content and our content

Yours. Your research questions, the files you upload, the reports and Answers results the Service produces for you, and the notes, claims and source text kept in your workspace vault are “Customer Content”. You own your questions, your uploads, and the reports and answers the Service writes for you as works. The source text, quotes, excerpts, titles and links inside Customer Content still belong to whoever owns them: you hold them in your workspace and may use them as this section allows, but you do not own them. We claim no ownership of any Customer Content, and no licence to it beyond what running the Service requires: to process your question, to send the necessary text to the model and search providers listed in the Privacy Policy, to store the results in your workspace, to show them to your workspace members and your authorised agents, and to keep backups and the billing and audit records described in the Privacy Policy.

We do not train on your content. We run no models of our own and we do not license Customer Content to anyone for model training. We send your prompts and the source text the pipeline read to third-party providers in order to generate your report, under the paid API terms of those providers, which is not the same thing as consenting to training; the Privacy Policy names every provider and links its terms. We cannot promise more than the strongest commitment our providers give us, and if you need a specific provider excluded from your workspace, write to us — the platform supports excluding providers.

Reports contain third-party material. A report quotes, paraphrases and cites pages we do not own. You own the report as a work; you do not thereby acquire rights in the sources it quotes. If you republish a report, the ordinary rules about quoting other people’s work apply to you, and the citations are there so you can check them.

Third-party material and our providers. Reports, Answers and search results, and the source text kept in your vault contain material from third parties: the websites and data sources the Service reads, and the search and page-reading providers we use to find and read them, which the Privacy Policy names (today they include Parallel Web Systems, Exa Labs, Valyu and Firecrawl). We give you a limited, non-exclusive, non-transferable, revocable licence to use that material inside your own research, products and services, for the person or agent whose request produced it, including showing it to your own end users as part of the response to that request, while your account is in good standing. That licence does not let you, and you must not let anyone else:

• sell, resell, sublicense, rent, publish or otherwise make results available as a dataset, a feed, or a search, answer or data service of your own;
• copy, cache or store results in order to serve them to other users, other customers or later requests, beyond keeping your own record of the calls you made and the reports you received;
• use results to build or add to a database, index, data brokerage or other data product;
• use results to train, fine-tune or create training data for a machine-learning model;
• use the Service or its results to build, improve or offer a product or service that competes with the Service or with any of our providers; or
• remove any copyright, attribution or other notice that a source or provider includes.

You must also comply with the terms and licences of the sources a result comes from. If you make the Service or its results available to your own users, you must bind them to terms at least as protective of this section as this section is, and you are responsible for what they do with them. If you use results in a legal, medical, financial or similarly high-risk setting, tell your users that your product relies on third-party search providers, and have a qualified person review results before they are relied on. This section protects our providers’ rights in what their services return as well as ours, and we may enforce it on their behalf. Section 7 and Section 11 apply to a breach of it.

Ours. The Service itself — the pipeline, the prompts, the models’ configuration, the console, the API, the denylists, the verification logic and everything we wrote — stays ours. You get a non-exclusive, non-transferable right to use it while your account is in good standing, and nothing more. Do not copy it, resell raw pipeline access, or try to extract our prompts or internal configuration.

Feedback. If you send us a suggestion we may act on it without owing you anything for it. We will not publish your name or your research as a case study without asking you first.

7. Acceptable use

These rules exist because the pipeline reads the open web on your behalf and because a research engine is a tempting tool for things that are not research.

Do not use the Service to:

• break any law, or to help someone else break one;
• gather material for harassment, stalking, doxxing or surveillance of a private individual, or to compile dossiers on people who have not consented;
• obtain access you are not entitled to — do not ask the pipeline to work around a login, a paywall, a rate limit or a technical access control, and do not use it to violate a website’s terms of use or to scrape a source that has told us or you to stop;
• produce material that infringes someone’s copyright, trademark, trade secret or privacy rights, or that is defamatory;
• produce or seek instructions for weapons, malware, or anything intended to cause serious harm; or sexual content involving minors, which we report;
• generate content presented as human-written where the context makes that deceptive, or as professional advice from a licensed professional;
• evade a spending cap, a rate limit, an allowance, a denylist entry, an authentication check, or a suspension;
• resell, sublicense or white-label raw pipeline access, or benchmark the Service for publication without telling us;
• resell, redistribute, store or reuse reports, Answers results or search results in a way §6 (“Third-party material and our providers”) does not allow;
• probe, load-test or attack the infrastructure, or interfere with anyone else’s use of it.

How we fetch. The pipeline records robots.txt for every page it reads and obeys it for crawl-shaped fetching — following citations onward from a page — and for workspaces that ask us to obey it everywhere. Opening a single page that appeared in search results is treated as a person clicking a link, which is what it is. We maintain a shared denylist that the pipeline will not fetch at all: financial institutions, government login portals, health portals, adult sites, data brokers, and any site that has asked us to stop. A workspace can add domains to that denylist. It cannot remove one. If you operate a site and want us to stop fetching it, email support@hyperresearch.ai and we will add it.

Rate limits and fair use. The API is rate limited per key and per account, and each workspace has a concurrency ceiling (2 concurrent runs on Pay-as-you-go). When you hit a limit the API says so and tells you when it resets. We may change these limits to keep the Service up. An account that repeatedly hits them through automation designed to hit them is abuse, not fair use.

What happens if you break these rules. For something ongoing and serious — active abuse, illegal content, an attack on the platform, a credible legal demand — we may suspend or terminate access immediately and without notice, and we will tell you why as soon as we reasonably can. For anything else we will contact you first and give you a chance to fix it. You can appeal any suspension by replying to that message or writing to support@hyperresearch.ai; a person, not a filter, decides the appeal. If we terminate you for a breach, your reports and vault remain exportable for 14 days.

8. Copyright complaints and takedowns

If you believe material in the Service infringes your copyright, send a notice to support@hyperresearch.ai with the subject line “DMCA” including: the work you say is infringed; the URL or run id where the material appears; your contact details; a statement that you have a good-faith belief the use is not authorised; a statement that the information in your notice is accurate and that you are authorised to act for the rights holder, under penalty of perjury; and your signature, electronic or physical. We will remove or disable the material we find and tell the customer whose workspace it was in, who may send a counter-notice to the same address.

If you operate a website and want the pipeline to stop fetching it, you do not need a copyright claim — see §7, “How we fetch”. That is the faster route and we treat it as a standing instruction.

9. No warranty

The Service is provided “as is” and “as available”. To the fullest extent the law allows, we disclaim all warranties, express or implied, including merchantability, fitness for a particular purpose, non-infringement, and any warranty arising from course of dealing. In particular, and without limiting that:

• we do not warrant that a report is accurate, complete, current, or that its citations support the sentences they are attached to;
• we do not warrant that the Service will be uninterrupted, available at any particular time, or free of errors; and
• we do not warrant that any run will finish, or finish in any particular time.

There is no service level agreement. We do not commit to an uptime figure, a support response time, or a maximum run duration. The wall-clock times shown next to each run size are targets, not promises. We will say so plainly rather than publish a number we have not measured.

Some jurisdictions do not allow the exclusion of certain warranties; where that is true, this section applies as far as it is permitted and no further.

10. Limitation of liability

To the fullest extent the law allows:

• Neither side is liable for indirect damages. That means no liability for lost profits, lost revenue, lost data, loss of goodwill, business interruption, or any indirect, incidental, special, consequential or punitive damages, even if we were told such damages were possible.
• Our total liability is capped. For all claims arising out of or relating to these terms or the Service, our aggregate liability is limited to the greater of (a) the total amount you paid us for the Service in the twelve months before the event giving rise to the claim, and (b) one hundred US dollars ($100).
• The cap does not cover everything. Nothing in this section limits liability for death or personal injury caused by negligence, for fraud or fraudulent misrepresentation, for our own wilful misconduct, or for anything else that cannot be limited by law.

These limits apply to us, and to our officers, employees and suppliers. They reflect a deliberate allocation of risk that the prices in §4 are set against: a $9 report cannot carry unlimited exposure for what someone does with it.

11. Indemnity

You will defend, indemnify and hold us harmless from any third-party claim, and from the losses, damages and reasonable legal costs that come with it, to the extent it arises from: your breach of §7 (Acceptable use) or of §6 (“Third-party material and our providers”), including a claim by one of our providers; your Customer Content; your republication or commercial use of a report or of Answers results; or your infringement of someone’s intellectual property or privacy rights through the Service. We will tell you promptly about any such claim, let you control the defence of it (as long as any settlement releases us fully and admits nothing on our behalf), and cooperate at your expense.

12. Confidentiality

Neither side will disclose the other’s non-public information learned through this relationship, except to people who need it and are bound to keep it confidential, or where the law compels disclosure. Your research questions, reports and vault are your confidential information; we treat them that way. Our prices are not confidential — they are published.

13. Suspension, termination and what happens to your data

You can leave at any time. You can delete individual vault sources from the console, export your whole vault as a tar.gz of Markdown, download any report as Markdown or PDF, and export your workspace audit log.

Deleting a workspace. A workspace can be deleted through the API (DELETE /v1/workspaces/{id}), which erases its vault, its stored sources, its reports and every run artifact. There is no button for it in the console yet, so the reliable route today is to email support@hyperresearch.ai and we will do it.

Closing an account. You can close your account yourself, from Account in the console. We do not delete anything straight away: closure is scheduled for 14 days out, you can cancel it from the same screen at any point in those 14 days, and the account keeps working normally until then — so export whatever you want to keep before the 14 days are up. After that we delete its data as described in the Privacy Policy, which is honest about what survives. If you would rather we did it, write to support@hyperresearch.ai.

On termination, by either side: your right to use the Service stops immediately; you remain liable for usage already incurred; and your reports and vault stay available for export for 14 days, after which we may delete them. Sections 6 (ownership), 9, 10, 11, 12, 14 and 15 survive termination.

We may discontinue the Service. It is a beta from a young company. If we shut it down or drop a feature you rely on, we will give you at least 30 days’ notice by email, refund any unused prepaid subscription time, and keep exports working through the notice period.

14. Export controls and sanctions

You may not use the Service, or let anyone use it through your account, in breach of US export control or sanctions law. You confirm that you are not located in a country or territory subject to comprehensive US sanctions, that you are not on a US restricted-party list (including the SDN list and the Entity List), and that you are not acting for anyone who is. You will not use the Service for any end use prohibited by the US Export Administration Regulations.

15. Governing law and disputes

These terms are governed by the laws of the State of Delaware, without regard to its conflict-of-laws rules. Any dispute arising out of or relating to them or to the Service goes to the state or federal courts located in Delaware, and both sides consent to the personal jurisdiction of those courts and waive any objection based on venue or inconvenient forum.

There is no arbitration clause and no class-action waiver in this agreement. We have not asked you to give up the right to a court or to join with others, and you have not.

If a dispute arises, please email support@hyperresearch.ai first. Most of what reaches us is a billing question that takes one reply.

16. Changes to these terms

We may change these terms. When we do, we will update the version and effective date at the top of this page, and for any change that materially reduces your rights or increases your obligations we will email the address on your account at least 14 days before it takes effect. Continuing to use the Service after the effective date of a change means you accept it; if you do not, stop using the Service and write to us to close your account, and we will refund unused prepaid subscription time.

17. The rest

Assignment. You may not assign this agreement without our written consent. We may assign it to a successor in a merger, acquisition or sale of assets, and will tell you if we do.

Entire agreement. These terms and the Privacy Policy are the whole agreement between us about the Service, and replace any earlier draft or placeholder text shown in the console.

Severability. If a provision is unenforceable, the rest stays in force and the unenforceable part is read as narrowly as it needs to be to become valid.

No waiver. Not enforcing something once does not waive it.

Independent parties. We are not partners, agents, employees or joint venturers.

Notices. We give notice to the email address on your account, or in the console. You give notice to support@hyperresearch.ai, or by post to the address in §1.

Force majeure. Neither side is liable for a delay or failure caused by something outside its reasonable control, including a failure of a cloud provider or model provider we depend on. Our refund policy in §5 still applies to a run such a failure kills.

18. Contact

Thalient Labs Inc., 1001 S. Main St., Suite 500, Kalispell, MT 59901, United States.

Questions about this document, about a charge, about a report, or about anything it should say and does not — email us. A person reads it.