Terms of Service

01Accepting these terms

Using the service means you accept these terms, and confirm you have read sections 3 to 6 (and section 13, if you use the Agent Transaction Check).

By creating an account, connecting a wallet, or submitting a contract for audit, you agree to these terms. If you are agreeing on behalf of a company or other organisation, you confirm you have authority to bind it, and "you" means that organisation.

You specifically acknowledge that you have read and understood sections 3 to 6, which describe how the audit is performed, the limits of what it can tell you, and who may rely on a report, and section 13 if you use the Agent Transaction Check. If you do not agree, do not use the service.

02Who operates this service

In these terms, "we" and "us" mean the operator of Grisoco AI, reachable at contact@grisoco.com.

03What the service does

An automated audit of the code you submit: tools plus AI review for Solidity, a compile check plus AI review for DAML. No human reads your code.

Grisoco AI performs an automated security audit of smart-contract source code that you submit. Two languages are supported, and you choose which review runs: Solidity (EVM) and DAML (Canton). They are different reviews, not one review pointed at different files — what an audit round combines depends on which you pick.

For Solidity, a round combines:

• A deterministic tool suite — Slither, Aderyn, Mythril, Echidna and Solhint, each run in an isolated sandbox with no network access.
• A structured AI review pipeline — independent agents, a jury and a judge, across more than one model family, whose role is to interpret, correlate and prioritise what the tools found.

For DAML, a round combines:

• A compile check only. There is no third-party static analyzer in a DAML audit. The only deterministic check is daml build, which proves the package compiles and has no opinion about who is authorized to do what. Nothing equivalent to Slither or Mythril runs, and no such tool is implied by the price.
• The same structured AI review pipeline — but with review agents aimed at DAML's own failure classes (who can act, who can see, who can proceed, and what a future package version would break) rather than at EVM ones, which do not exist in DAML. Every security finding in a DAML report comes from these agents and the jury. The report states this on its face.
• Ledger verification of authorization findings. Where an authorization finding can be expressed as a command, it is submitted to a disposable Canton ledger as the party who should not be able to act; a ledger that accepts it is recorded as the evidence. Privacy and liveness findings are not mechanically verified — there is nothing to replay — and the report says which is which.

The output is a findings report: a prioritised list of issues worth a human's attention, with the reasoning behind each.

The audit is performed by software, not by people. No qualified security engineer reads your code, forms a professional opinion, or stands behind the conclusion. That is the substantive difference between this and an engagement with an audit firm, and it is reflected in the price.

The service also supports remediation rounds: after a report, you can fix or dispute individual findings and resubmit, and the service compares the new submission against the previous round.

The audit is performed on the code you submit, exactly as submitted. For Solidity, it does not inspect deployed bytecode, your deployment process, your keys, your governance, your economic design, your off-chain infrastructure, or any dependency you do not include. For DAML, it likewise does not inspect anything off-ledger — keeper and automation processes, participant node configuration, party allocation and identity management, or obligations enforced by agreement rather than by code — and upgrade-related findings are reasoned from the single package version you supply rather than from a comparison against another.

The service also offers the Agent Transaction Check, described in section 13.

04What no audit can promise

No audit proves a contract is safe. This one will miss issues, may report false ones, and is not a certification.

Specifically, and without limiting anything above, the service:

• Will miss vulnerabilities, including critical ones. Automated tools and language models are systematically weak at exactly the issues that cause the largest losses: flawed business logic, broken economic assumptions, governance and privilege design, cross-contract and cross-protocol interactions, and anything requiring an understanding of what your protocol is supposed to do.
• Will sometimes report findings that are not real, or misjudge the severity of ones that are. Findings are a starting point for human review, not conclusions.
• Is not a certification, accreditation, attestation of security, or seal of approval, and is not issued by any accreditation body.
• Involves no human review. Nobody at Grisoco AI reads your code or checks the findings before they reach you.
• Is not financial, investment, tax or legal advice, and nothing in a report is a recommendation to deploy, launch, invest in, or transact with any contract.
• Creates no professional, advisory or fiduciary relationship between you and us, and no duty of care beyond delivering the automated audit you paid for.

You may accurately say your contract was audited automatically by Grisoco AI. You must not describe it as "certified" or "verified" by us, drop the word "automated", or otherwise use our name in a way that implies a guarantee, a human review, an audit-firm engagement, or a professional opinion we have not given.

05Your responsibility for your own contracts

Your contracts’ security stays with you: get a professional human audit before deploying anything of value.

The security of your contracts is yours alone. Using this service does not transfer any part of that responsibility to us, no matter what the report says.

You agree that before deploying any contract that will hold, control or move assets of value, you will:

• Obtain an independent security audit from a qualified professional firm — a human one — and act on it;
• Have a qualified human engineer review every Grisoco AI finding rather than acting on the report mechanically; and
• Apply your own testing, review and operational safeguards.

If you choose to deploy on the strength of an automated audit alone, you do so entirely at your own risk and against our express recommendation, recorded here.

06No reliance by anyone else

A report is for you only. Share it whole and unaltered, and never present it as a certification.

If you share a report, you must share it in full and unaltered, and you must not present it as an audit, a certification, or a statement that your contract is secure. If you present a report in a way that misleads someone, that is your act, not ours, and section 19 applies.

07Who may use the service

You must be 18 or over, not in a sanctioned country, and follow the laws that apply to you.

You must be at least 18 years old. You must not be located in, or ordinarily resident in, a country subject to comprehensive trade sanctions, and you must not appear on any applicable restricted-party list. You are responsible for complying with the laws that apply to you, including any that govern your use of cryptocurrency.

08Your code and your rights

You keep your code and your report. We use the code only to run the audit, never to train models.

You keep ownership of everything you submit. We claim no rights in your source code beyond what is needed to run the audit.

You grant us a limited, non-exclusive licence to store, process, transmit and analyse your submitted code for the purpose of delivering that audit — which includes transmitting it to the AI providers and sandbox infrastructure listed in our Privacy Policy. That licence ends when the processing ends. We do not use your code to train models.

You confirm that you own the code you submit, or otherwise have the right to submit it for analysis. Do not submit code that is subject to a confidentiality obligation you would breach by sharing it with a third-party processor.

The report belongs to you, subject to section 6. You may use, copy and share it, internally or publicly, without further permission from us.

09Acceptable use

Do not submit code you have no right to, attack others, bypass limits, or try to extract our pipeline.

You must not:

• Submit code you have no right to submit, or that contains personal data, credentials or secrets.
• Use the service, or anything it produces, to attack, exploit or gain unauthorised access to any system or contract you do not own or are not authorised to test.
• Attempt to circumvent payment, rate limits, usage limits, or the sandbox isolation.
• Submit content designed to make the service attack a third party, exfiltrate data, or behave outside the audit it was asked to perform — including instructions embedded in source code or comments and aimed at the AI pipeline.
• Resell, sublicense or white-label the service without our written agreement, or offer its output to others as if it were your own audit.
• Scrape, reverse engineer, or attempt to extract the prompts, models or internal structure of the analysis pipeline.

10Accounts

Sign in by email, Google, GitHub or wallet. Keep your account secure; sessions last 30 days.

You may sign in with an email address, a Google or GitHub account, or a wallet signature. You are responsible for keeping access to your account secure, and for everything done through it. Sessions expire after 30 days.

If you connect a GitHub account, you can revoke that connection at any time from GitHub, and the stored token becomes unusable.

11Payment

Pay per round in USDC. Payments are irreversible and generally non-refundable; a failure on our side is not charged twice.

The service is pay-per-round, in USDC on-chain. Current prices are listed on the pricing page, and the exact amount is shown to you before you pay. How a round is priced:

How payment works

• You are quoted a price before the round runs, and you send USDC to the address shown.
• A payment reference is valid for 15 minutes. If you do not broadcast in that window it expires, though a genuine late transaction is given a short grace period to land.
• If a round fails on our side, your payment reference is released and can be used again — you are not charged twice for one failure. Where a round is paid from an account balance, a failed attempt is credited back.
• After a successful round you can re-download the report for one hour without paying again.

Except where these terms say otherwise, or where a refund is required by law, payments are non-refundable. You are paying for an audit round to be run, not for any particular result: a report that finds nothing, or that you disagree with, is a round that was delivered. Prices may change, but a change never applies to a round you have already paid for. You are responsible for any taxes arising from your use of the service.

12Reports and attestations

Each report carries an attestation that we produced it — not a statement that your contract is secure.

Reports are provided as Markdown, PDF or DOCX. Each report carries a cryptographic attestation, which lets us confirm a report was produced by the service and has not been altered.

The attestation says only that we produced the file. It is not a statement about your contract's security, and it is not a certification. It is also best-effort and time-limited: the signing key may be rotated, and a report signed before a rotation may no longer verify afterwards. Keep your own copy of any report that matters to you.

13Agent Transaction Check

The check compares what a transaction or signature would do with what you say it should do. “No risks detected” is not a guarantee, and whether to sign is always your decision.

What the check does. The Agent Transaction Check is an automated service, offered through our API and MCP server, for software that signs blockchain transactions or messages on someone's behalf (an "agent"). You send us a proposed transaction or signature request with a statement of what it is meant to do (an "intent"), such as a payment of a stated amount of a stated token to a stated address. You may also send limits of your own (a "policy"). We simulate the transaction against the chain's state at a stated block, or decode the message to be signed, and compare the result with the intent and policy. The answer (a "verdict") is allow, warn or block, with a reason for each check. Every check is a fixed, rule-based comparison. No AI model and no person is involved in producing a verdict.

In particular:

• The chain can change between the check and the moment the transaction lands. Balances, prices, contract code and the order of other transactions may differ by then, and the transaction may do something other than what was simulated.
• A simulation is not execution. A contract can be written to behave differently when simulated, and a contract can move assets in ways our checks cannot see.
• Our checks are limited in scope. We check only the tokens on our published list, and only the asset movements and spending permissions that tokens record in the standard way.
• A signature check reads only the message. It cannot tell how, when or by whom a signature will be used once given, beyond what the message itself says.
• We rely on third-party node providers for chain data. If their data is wrong or out of date, the verdict will be too.
• We check the transaction against your intent, not the intent itself. We do not check whether the intent is right, whether the recipient is who they claim to be, or whether the purchase is legitimate. If an agent has been misled into a wrong intent, a transaction that faithfully carries it out will be allowed.
• Warn and block verdicts can also be wrong. A legitimate transaction may fail a check.

You decide, and you sign. We never hold keys, and we never sign, submit, delay or reverse a transaction. Whether to sign is decided by you or your software, and stays your responsibility whatever the verdict. You are responsible for your own safeguards, including spending limits and the security of your keys. You are also responsible for what your software does when the check is unavailable, slow, or returns an error. We recommend that it does not sign in that case.

Your users. If you build the check into a product that others use, you must not tell them that we guaranteed, certified, approved or verified a transaction or signature as safe. You must not present a verdict as meaning more than this section says. Your users acquire no rights against us, and section 6 applies to verdicts just as it does to reports.

Acceptable use. Section 9 applies to the check. In addition, you must not use it to design, test or refine transactions or messages intended to take assets from anyone, or to get around the check or any similar protection. You must not resell verdicts as your own service without our written agreement.

Payment. Each check is charged from your account balance at the price listed on the pricing page, whatever its verdict. A check that returns no verdict because of a failure on our side is not charged. Section 11 otherwise applies, including how price changes work.

Liability. Sections 17 and 18 apply to the check. Without limiting them, we are not liable for any loss of funds, tokens or digital assets resulting from a transaction or signature that was checked, whatever the verdict. Nor are we liable for losses from one that went unchecked because the check was unavailable. For claims relating to the check, our total aggregate liability is limited to the fees you paid us for checks in the three months before the event giving rise to the claim.

14Third-party services

The service relies on third parties listed in the Privacy Policy; we are not responsible for their outages or terms.

The service depends on third parties, including cloud infrastructure, a hosted database, AI model providers, blockchain RPC providers, and the OAuth providers you choose to sign in with. They are listed in our Privacy Policy.

We are not responsible for those third parties' own acts, outages or terms. A public blockchain in particular is outside anyone's control: transactions on it are permanent and publicly visible.

15Availability and changes

No uptime guarantee. We may change the service, but will not knowingly break a round you have paid for.

The service is offered without any uptime commitment or service-level agreement. We may change, suspend or discontinue any part of it, and we may impose or adjust usage limits. We will not knowingly break a round you have already paid for.

16Suspension and termination

You can stop at any time. We may suspend or end access if these terms are breached.

You may stop using the service at any time and ask us to delete your account, as described in the Privacy Policy.

We may suspend or terminate your access if you breach these terms — in particular sections 4, 6 and 9 — or where we reasonably need to in order to protect the service or other users. Where it is practical and lawful to do so, we will tell you why. Unused account balance is refundable on termination unless the termination was for a breach of section 9.

17Disclaimer of warranties

The service is provided “as is”, without warranties, except where the law says otherwise.

To the fullest extent permitted by law, the service and every report it produces are provided "as is" and "as available", without warranties of any kind, whether express, implied or statutory — including any implied warranty of merchantability, fitness for a particular purpose, accuracy, completeness, or non-infringement.

Without limiting that, we do not warrant that the service will be uninterrupted or error-free; that it will detect any particular vulnerability or class of vulnerability; that its findings are accurate, complete or correctly prioritised; or that a contract it screened is secure, correct, or fit to deploy.

Nothing in these terms excludes or limits any liability that cannot lawfully be excluded or limited — including liability for fraud, or for death or personal injury caused by negligence — and if you deal with us as a consumer, your statutory rights are unaffected.

18Assumption of risk and limitation of liability

You carry the risk of deploying. Our liability is capped at what you paid for that round (for transaction checks, see section 13).

You assume the risk of deploying smart contracts. Smart contracts are immutable, adversarial, and hold value that cannot be recovered once taken. You understand that an automated audit reduces that risk only partially, and that the residual risk stays with you.

To the fullest extent permitted by law, we are not liable for any indirect, incidental, special, consequential or punitive damages, nor for any loss of profits, revenue, business, data, goodwill, tokens or digital assets — including, for the avoidance of doubt:

• any exploit, hack or loss of funds affecting a contract we screened;
• any vulnerability the service failed to find, whatever its severity;
• any decision you or anyone else took in reliance on a report, including a decision to deploy; and
• any loss arising from a finding that was wrong, incomplete or wrongly prioritised.

Our total aggregate liability arising out of or relating to the service is limited to the amount you actually paid us for the audit round giving rise to the claim or, for the Agent Transaction Check, the amount set out in section 13. This limit applies however the claim arises, whether in contract, tort (including negligence), or otherwise, and whether or not we were told such loss was possible.

You agree that this allocation of risk is reasonable, and that it is reflected in the price — an automated audit costs a small fraction of a human engagement precisely because it does not carry a human engagement's assurance or its liability.

19Indemnity

You cover our costs for claims arising from your breach, misuse, or misrepresenting a report or verdict.

You will indemnify us against claims, losses and reasonable costs arising from: your breach of these terms; your submission of code you had no right to submit; your use of the service in a way section 9 prohibits; your presentation of a report as an audit or certification; your presentation of a verdict in a way section 13 forbids; or any claim brought by a third party who relied on a report or verdict you shared.

20General

Standard provisions: severability, no waiver, assignment, and this being the entire agreement.

If any provision of these terms is held unenforceable, the rest remain in force and that provision is limited to the minimum extent necessary. Our failure to enforce a provision is not a waiver of it. You may not assign these terms without our consent; we may assign them to a successor in connection with a transfer of the service. These terms, together with the Privacy Policy, are the entire agreement between us about the service. We are not liable for failures caused by events beyond our reasonable control.

21Governing law

Indian law governs these terms, and Indian courts have jurisdiction.

These terms are governed by the laws of India, and the courts of India have jurisdiction over any dispute, except that either party may seek injunctive relief wherever appropriate.

Before starting formal proceedings, please contact us — most disputes can be resolved quickly and directly.

22Changes to these terms

We may update these terms, with notice before any material change takes effect.

We may update these terms. The "last updated" date at the top always reflects the current version. For material changes we will give notice — by email to your account address, or prominently in the workspace — before they take effect. Continuing to use the service after that point means you accept the updated terms.

23Contact

contact@grisoco.com