AI legal assistant
Paste a contract or upload an NDA and get back a structured review in minutes: the risky clauses flagged, the missing protections named, and a draft response you can send to the other side. Zeplik reads only the documents you provide, so the analysis is grounded in your deal, not a generic template. When an answer would cite case law, Zeplik verifies the citation against a real court database instead of inventing one, the single failure that makes a general chatbot dangerous for legal work.
Who it is for
- Founders and operators signing vendor, SaaS, and employment agreements without in-house counsel on call
- In-house and solo attorneys who want a fast first pass before they spend billable time
- Procurement, sales, and ops teams triaging NDAs and order forms at volume
- Anyone who needs to understand a contract before they sign it
How it works
Upload the document
Drop in the contract, NDA, policy, or letter as a PDF or DOCX, or paste the text straight into the chat. Nothing is assumed about your situation. The review is built only from what you share, and you can add context (which side you are on, what you care about) in plain language.
Ask for the review you need
Say what you want in your own words, for example "flag the risky terms in this SaaS contract" or "is this NDA mutual and what should I push back on?" Zeplik routes to the right legal skill without you naming it, and you can steer: focus on liability, on IP, on termination, on data.
Review, redline, and respond
You get a structured result with the working shown: clause-by-clause risk, suggested safer language, and, if you ask, a drafted email or redline you can send. Everything stays in one conversation so you can iterate, ask follow-ups, and export the final version.
What you can ask
Real asks in plain language, and the concrete deliverable you get back. Open the assistant to try your own; nothing runs until you send it.
Review this SaaS subscription agreement before I sign it and flag anything risky.
A clause-by-clause pass that surfaces auto-renewal traps, uncapped liability, one-sided indemnities, and data terms, each with why it matters and safer language to propose.
Is this NDA mutual or one-way, and what should I push back on?
A triage that classifies the NDA, calls out an overbroad definition of confidential information, a missing residuals clause, and a term that never ends, plus the specific edits to request.
Draft a firm but professional response to this demand letter.
A drafted reply grounded in the letter you uploaded that acknowledges receipt, disputes the claims you flag, and preserves your position, ready for your attorney to review and send.
Assess our overall legal exposure on this vendor deal, here is the master agreement.
A risk assessment that ranks the exposures (liability, IP ownership, termination, compliance) from the actual agreement text, with the highest-risk items first and the questions to raise before signature.
Check whether our data-processing addendum meets GDPR requirements.
A structured gap check against GDPR data-handling expectations: lawful basis, sub-processor terms, data-subject rights, transfer mechanism, and breach notice, with the gaps named against your own DPA.
Prep me for a legal review meeting with counsel on these three contracts.
A briefing that summarizes each agreement, lists the open issues and decisions, and gives you the questions to ask, so the billable hour is spent deciding, not reading.
What it can do
The assistant routes to the right skill for you. Each one below is a focused, ready-to-run workflow; follow a link to see exactly what it does.
Review and triage
Read a document you upload and turn it into a decision. These are the skills that most legal chats start with.
- Contract reviewClause-by-clause risk, redlines, and safer language for any agreement you paste.
- NDA triageIs it mutual, is it overbroad, what to push back on, in one pass.
- Employment agreementsOffer letters, non-competes, and IP-assignment terms read for what binds you.
- Vendor due diligenceLiability, termination, and obligations across a vendor's paper before you commit.
Risk and compliance
Move from a single document to a defensible position on exposure and obligations.
- Legal risk assessmentRank the exposures in a deal from the actual text, highest risk first.
- Compliance checkTest terms against a regulatory checklist you provide and name the gaps.
- GDPR data handlingData-processing and privacy terms checked against GDPR expectations.
- Worker classificationRun the applicable jurisdiction test on a proposed engagement and flag misclassification gaps before you hire.
- Open-source license reviewMap copyleft obligations across a dependency list to how you actually ship, and flag non-OSI licenses.
Draft and prepare
Produce the work product: briefs, responses, and the prep that makes counsel time count.
- Demand letterA payment, breach, or cure demand drafted from your facts, gated on a privilege and FRE 408 check first.
- Cease and desistA C&D calibrated to your enforcement posture, or a triage of one you received into an options memo.
- Legal briefA short, structured brief summarizing the issues in a dispute or matter.
- Legal responseA drafted reply to a letter or claim, grounded in the document you uploaded.
- Meeting briefingA pre-read that turns a stack of contracts into a decision agenda.
- Signature routingWho signs, in what order, and what has to be in place first.
Why a general chatbot is risky for legal, and what is different here
The dangerous failure in legal AI is not a wrong summary you can catch. It is a fabricated case citation that looks real. General chatbots have invented case names, reporter numbers, and holdings with total confidence, and lawyers have been sanctioned for filing them. Zeplik is built to not do that.
The failure. A general model asked for supporting case law will often invent a citation that does not exist, complete with a plausible reporter number.
How Zeplik grounds it. When an answer would cite a case, Zeplik searches CourtListener, a real public database of US court opinions, and cites only cases that actually come back with a real name, court, date, and reporter citation. If nothing verifiable is found, it says so and refuses to invent one.
The failure. A general model treats your specific contract as an average of every contract it has seen, so it hallucinates clauses that are not in your document.
How Zeplik grounds it. Every review is grounded only in the file you upload or paste. The analysis quotes and points to your actual text, so you can check each flag against the source.
The failure. A general model presents analysis as if it were legal advice, with no honest boundary.
How Zeplik grounds it. Every legal result carries a uniform disclaimer: it is draft analysis for a qualified attorney to review, not legal advice, and it creates no attorney-client relationship. The moat is honesty about what the output is.
What it does not do
Knowing the boundary is part of using it responsibly. This is where a person, not the assistant, has to own the call.
- It is not your lawyer. The output is a draft for a qualified attorney to review. It does not create an attorney-client relationship and should not be filed or sent as-is on anything that matters.
- Jurisdiction matters and it will not always know yours. Law varies by state and country. Tell it the governing law, and still have a licensed attorney in the right jurisdiction confirm anything you rely on.
- It reads what you give it. If a schedule, exhibit, or side letter is not in the upload, it is not in the analysis. Provide the full document set for a complete review.
- Verified does not mean applicable. A case citation checked against CourtListener is real, but whether it actually supports your position is a legal judgment for your attorney, not a lookup.
This is draft analysis to be reviewed by a qualified attorney -- it is not legal advice and creates no attorney-client relationship. Rely only on the documents and facts you provide; do not treat any cited authority as verified unless it appears in your materials.
Frequently asked questions
- Is the AI legal assistant free?
- Zeplik is free to start and every new account includes credits. Legal skills run like any other Zeplik chat, so you only spend normal usage credits on the model, with the cost shown before you send. Verifying a case citation against the court database is free.
- Does it give legal advice?
- No. It produces draft analysis and work product for a qualified attorney to review. It is not legal advice and does not create an attorney-client relationship. Treat it as a fast, structured first pass, not a substitute for counsel.
- How does it avoid making up case citations?
- When an answer would rely on case law, Zeplik searches CourtListener, a real public database of US court opinions, and cites only cases that come back verified with a real name, court, date, and reporter citation. If it cannot verify an authority, it tells you rather than inventing one.
- What documents can I upload?
- Contracts, NDAs, employment agreements, policies, demand letters, and similar documents as PDF or DOCX, or just paste the text. There is no account linking or connector to set up. The review is built only from what you provide.
- Is my document confidential?
- Your documents and conversations are tied to your account. Zeplik does not require you to connect any external legal system, and you control what you upload. Review the privacy policy for the specifics that apply to your data.
- Can it redline a contract and draft a reply?
- Yes. It suggests safer language clause by clause and, if you ask, drafts an email or response grounded in the document, so you can take the review straight into negotiation. Everything stays in one conversation for follow-ups.
More on Zeplik
Review your first document
Upload a contract or paste an NDA and get a structured review you can act on. Free to start, and citations are verified, not invented.