Kranth · Terms of Service

#1. The deal

These terms cover your use of Kranth (the "Service"): the websites and apps at kranth.com, kranth.ai and their subdomains, the Kranth API, the official SDKs (Python, TypeScript, Go and Rust), the Kranth MCP connector, and the Kranth GitHub App. The Service is operated by Kranth AI, Inc., a Delaware corporation ("Kranth", "we", "us"). You are the "Customer" or "you".

By creating an account, calling our API with an API key, installing the Kranth GitHub App on a repository you control, or clicking "I agree" when we show it, you accept these terms on behalf of yourself or the entity you represent and confirm you have authority to do so. You must be at least 16. If you don't agree, don't use the Service.

These terms incorporate the Privacy Policy (/privacy), Cookie Notice (/cookies), Acceptable Use Policy (/aup), and Refund Policy (/refund). If those conflict with these terms on personal-data processing, the Privacy Policy controls.

#2. What the Service does

Runs synthetic-audience simulations using AI personas with distinct biases, and returns their reactions, the objections they share and a synthesized verdict.
Runs debates, in which personas argue a motion against each other and a judge model scores the result.
Runs Recon, in which research agents read the public web about your idea and return cited findings and a score.
Runs Interview Mode: a panel of AI personas questions you about a decision you name, in text or by voice, and returns a readiness assessment: how well you defended the decision, which questions you did not answer, and what you would need to answer them.
Bills credits proportional to persona count × model tier, or to booked session length for Interview Mode.
That's it. Nothing here is investment, legal, medical, or therapy advice.

Interview Mode specifically does not tell you what to do. It scores how you defended a decision under questioning. It does not tell you to take the job, raise the price, or kill the product line, and you should not read its output as a recommendation to do any of those. The panel is synthetic: it is not a lawyer, an investor, a doctor, an accountant, or a customer, and nothing it says is professional advice or a substitute for it. Decisions you make afterwards are yours.

#3. Your account

One human per account. No sharing logins across teams; use team seats instead (Starter includes 1, Pro 3, Scale 10, Enterprise by contract).
You're responsible for everything that happens under your API keys. Treat them like passwords: keep them out of git, rotate them when you suspect leak, revoke through /app/api-keys.
Tell us if you suspect unauthorized access. Email [email protected].
We are not liable for losses caused by your failure to keep credentials secret.

#Access license

Subject to these terms we grant you a limited, non-exclusive, non-transferable, revocable license to use the Service for your internal business (or personal) purposes. We and our suppliers keep all IP in the Service. These terms transfer no ownership.

You may not: (i) license, sell, rent, or commercially exploit the Service except via a written reseller agreement; (ii) reverse-engineer, decompile, or create a competing service from the Service except where law forbids that restriction; (iii) scrape the Service to train a model or build a substitute; (iv) remove proprietary notices.

We may modify, suspend, or discontinue features with notice where practicable. We do not owe you support except as your plan states.

Feedback. If you send suggestions, you grant us a perpetual, irrevocable, worldwide, royalty-free license to use them without attribution. Do not send feedback you consider confidential.

#4. Acceptable use

You agree to follow the Acceptable Use Policy at kranth.ai/aup. In short: no illegal content, no impersonation of real persons in personas without consent, no targeted harassment, no election-interference workflows, no CSAM or violence-incitement. Full list in the AUP.

We can suspend or terminate your account immediately and without refund for AUP violations.

#5. Pricing + credits

Pricing lives at kranth.ai/pricing, where plans and credit packs are listed.
Credits debit on sim accept (pre-authorization). Unconsumed credits refund automatically on sim.cancel and on sim.failed.
Debates reserve credits for the maximum turn count at start; unspent turns refund automatically on completion or failure.
Interview Mode charges for the session length you book, at the start of the session. Ending early does not refund the remainder. The panel and its questions are generated up front against the length you chose. A session that fails before the first question refunds in full.
Credits don't expire while your account is open, whether they came from a plan or a top-up.
We can change pricing for future renewals. Existing subscribers stay on their current plan until renewal; we email 30 days in advance of any change that affects you.
Tax (VAT / sales tax) is added by Stripe per your billing address.

#6. Refunds

Unused credits are refundable for 30 days. Consumed credits are not. Email [email protected] with the account email within 30 days of charge and we'll refund your unused balance within 5 business days.

"Consumed" means credits already spent running a sim, debate, recon, interview, or improvement. That's live model compute we already paid for on your behalf, so it can't be returned.
Applies to monthly + annual subscriptions and top-up packs. We refund the unused portion to your original payment method via Stripe and tell you the math.
Runs that fail or are canceled refund their credits automatically (Section 5). You never pay for compute that didn't run.
Free tier doesn't charge, so there's nothing to refund.
We may refuse refunds during AUP enforcement or on repeated refund-then-reactivate cycles.
EU/UK consumers keep their statutory cooling-off rights (see the Refund Policy); where the law gives stronger rights, those win.

#7. Your content + your sims

You own your ideas. Idea text you submit is yours. We never claim ownership of your inputs.
You own the per-persona reactions generated from your inputs, on the plan you ran them.
You own your interview transcript and its readiness assessment. The answers you give a panel are your content, same as idea text. Interviews are private to your organization and have no public share link.
We retain a license to use de-identified, aggregated usage data (counts, percentiles, model-mix breakdowns) to operate and improve the Service.
Training data is opt-in only, off by default. See the Privacy Policy section on AI training data. Enterprise plans contractually never train.
Public sharing. You can publish a sim or debate at a public share link (/s/…, /d/…), including voice audio for voice debates. You control visibility and can unpublish at any time; published content must follow the AUP, and we may take down publicly shared content that's reported and violates it.
Integrations. If you connect Slack, Discord, Telegram, or Linear, we deliver notifications and verdicts to those services at your direction. Their own terms govern what happens on their side.

#8. Service availability

We aim for high availability but don't promise specific uptime SLAs on Free / Starter / Pro tiers.
Scale + Enterprise plans get a 99.9% monthly uptime commitment, with credits as the remedy. Specific SLA in your order form.
Scheduled maintenance gets emailed 48 hours ahead when avoidable.
Force majeure (outages of Stripe, Anthropic, Cloudflare, etc., upstream of us) suspends the SLA for the duration.

#9. LLM provider behavior

When you pick a model, we route your prompt + persona context to that model's provider (Anthropic, Google, Together AI, Groq, xAI, or your on-prem Ollama for Enterprise). GPT-family models, and some providers when direct routing is unavailable, are routed through OpenRouter as an intermediary. If you pick a model mix, we distribute personas across the mix's providers; that's the mix doing what it says. Voice debates additionally send turn text to Cartesia for speech synthesis. A voice interview additionally streams your microphone audio from your browser to Cartesia for live transcription (see the Privacy Policy and Subprocessors); choosing to type instead sends Cartesia nothing inbound. Each provider has its own policies: they might refuse certain content, or insert their own safety filters. We pass their behavior through without trying to override it.

The same applies to an interview: if a provider fails a session repeatedly, the session ends and its credits are refunded in full, however far in you were. You do not pay for our provider's outage. If a provider refuses to answer a question mid-session, that question is surfaced as a failure rather than quietly replaced.

If a model fails repeatedly on a sim, we refund your credits + surface the error. No silent fallback to a different model: you picked it, and we honor it. Hard fail beats soft fail. One narrow exception: if the final verdict-synthesis step fails after your sim has already run, we return the statistical verdict (scores + breakdowns) without the LLM-written narrative rather than discarding a completed sim.

#10. Liability

We provide the Service "as is" and "as available". No warranties beyond what the law mandates.
Our maximum liability for any claim is capped at the fees you've paid us in the 12 months before the claim.
We're not liable for indirect / consequential damages (lost profits, business interruption, etc.).
Nothing in these terms excludes liability we can't lawfully exclude (gross negligence, fraud, etc.).

If you're an enterprise customer, your order form may modify these caps.

#11. Indemnity

You agree to defend + indemnify us against claims arising from your violation of these terms, the AUP, or the rights of any third party.

We agree to defend + indemnify you against claims that the Service itself infringes a third-party IP right, provided you notify us promptly and let us run the defense.

#12. Suspension + termination

You can cancel anytime in /app/billing → "Cancel subscription". Cancellation takes effect at period end (you keep the credits you paid for until then).
We can suspend or terminate your account for AUP violations, non-payment, chargebacks, or court order. We'll always give a reason if it's lawful for us to.
On termination by us for cause: no refund.
On termination by you, or by us without cause: unused credits are refundable under Section 6.
Cancelling a subscription doesn't delete your data. It stays in your workspace until you delete it.
If we terminate your account, you have 30 days to export your data, unless the law or the reason for termination prevents it. After that we delete it.
Deleting your account or workspace removes it from our database immediately. Backups roll off within 7 days.

#13. Modifications to these terms

Material changes: 30 days notice via email to the account owner.
Non-material (typos, clarifications): we update the doc and bump "Last updated"; no notice.
The current version always lives at kranth.ai/terms. Prior versions available on request to [email protected].

#14. Disputes

Governing law: Delaware, USA, without regard to conflict-of-law rules. Disputes go to state or federal courts in Delaware. Either party may still seek injunctive relief for IP anywhere, or file an individual small-claims action where that court has jurisdiction. If your jurisdiction gives you a stronger consumer right, that overrides this section to the extent of the conflict.

We prefer to resolve disputes informally first. Email [email protected] (or write to the mailing address below) and wait 30 days before filing, unless you need emergency injunctive relief.

#14a. State consumer notices

These apply only to the extent that state's law covers you.

California. Complaints: Complaint Assistance Unit, Division of Consumer Services, California Department of Consumer Affairs, 1625 N. Market Blvd. Suite N112, Sacramento, CA 95834, (800) 952-5210. Civil Code § 1789.3 notice: the provider is Kranth AI, Inc., c/o Stable, 2810 N Church St STE 89663, Wilmington, DE 19802. Write that address or [email protected]. CCPA rights: Privacy Policy. California residents waive Civil Code § 1542 for the release in these terms to the extent they release us (unknown claims).
Colorado / Connecticut / Virginia / Texas / others. State privacy rights (access, delete, opt out of sale/sharing/targeted ads): Privacy Policy.
Nevada. Right to direct us not to sell covered information: [email protected]. We do not sell for money today.

#15. Miscellaneous

No agency / partnership. These terms don't create a joint venture, agency, or franchise relationship.
Assignment. You can't assign your account without our consent. We can assign these terms to a successor entity.
Severability. If a section's unenforceable, the rest stays.
Entire agreement. These terms + the Privacy Policy + Cookie Notice + AUP + Refund Policy + your order form (if any) are the whole deal.
Export. You will not export or re-export the Service or technical data in violation of U.S. export law.
Electronic communications. You consent to notices by email or in-product post. That satisfies any legal requirement for writing.
Accessibility. We aim for WCAG 2.1 AA on public pages. Difficulty using the Service: [email protected].

#16. Contact

Customer support: [email protected]
Legal: [email protected]
Security: [email protected]
Privacy: [email protected]
Mail: Kranth AI, Inc., c/o Stable, 2810 N Church St STE 89663, Wilmington, DE 19802, United States

If anything's confusing, ask before you sign up. We'd rather rewrite this than litigate it.