Sealwax -- Instant NDAs for Freelancers and Independent Contractors - Vertical Agent Spec

One-line definition

An agent that drafts, routes, and archives mutual or one-way NDAs for freelancers negotiating new client engagements.

The workflow it owns end-to-end

Trigger: freelancer describes the engagement (type of work, parties, jurisdiction, approximate deal size) via a short web form or conversational interface.
Agent asks 5-8 clarifying questions: unilateral vs. mutual, IP ownership carve-outs, confidentiality duration, governing law preference, any known sensitive categories (trade secrets, source code, client lists).
Agent generates a draft NDA with plain-language annotations on each clause, flags any inputs that suggest the document is outside its scope, and presents the draft for freelancer review before sending.
Agent routes the document to all parties via e-signature link, tracks signature status, and sends reminders to unsigned parties at 24 and 72 hours.
On full execution, agent stores the signed PDF, logs the key dates (expiration, any renewal triggers), and sends a confirmation to both parties.

What it knows that a generic LLM doesn't

Common freelance NDA failure modes: omitting carve-outs for information already in the public domain, forgetting to specify that IP ownership is a separate agreement from confidentiality, using "perpetual" confidentiality language that courts routinely void.
Jurisdiction-specific enforceability rules for the most common freelance states: California's blanket ban on non-competes embedded inside NDAs, New York's consideration requirements, Delaware's business-entity defaults.
Appropriate confidentiality duration norms by engagement type: 2-3 years for software and tech, 1-2 years for creative work, indefinite claims on genuine trade secrets (with a note that indefinite is frequently reduced by courts).
When a client's NDA request is structurally lopsided: a unilateral NDA that binds only the freelancer while the client shares nothing is a negotiating signal, not a legal requirement, and the agent can flag it.
E-signature routing conventions: which states require wet signatures for NDAs to be enforceable (almost none), and when a DocuSign audit trail is vs. is not sufficient evidence of execution.
Plain-language translation vocabulary specific to freelance engagements: translating "residuals" clauses, "work made for hire" language that should not appear in an NDA but sometimes does when clients mix up documents.

What it explicitly declines

Legal advice on whether an existing NDA was breached, whether a clause is enforceable in a specific dispute, or what remedies are available after a violation.
Any document outside the NDA and basic confidentiality agreement scope: employment contracts, non-competes as standalone agreements, IP assignment agreements, licensing deals.
Jurisdiction coverage outside the US, UK, Canada, and Australia. The agent's clause logic degrades significantly outside those systems and it should say so rather than generate plausible-sounding output.
Representation, indemnification, or any assurance that the generated document will hold up in court. The agent produces a document, not a legal opinion.

Tools and integrations required

HelloSign API or DocuSign API for e-signature routing, status tracking, and executed document retrieval.
Stripe for per-use payment processing at the moment of document generation.
AWS S3 or equivalent object storage with per-user access controls for signed document archiving.
Puppeteer or WeasyPrint for rendering the final formatted PDF from the agent's structured output.
SendGrid or Postmark for signature-status emails and expiration reminders.
Optional: Google Drive or Dropbox API for users who want executed documents pushed to their own storage rather than held in Sealwax's system.

Trust escalation: when it pings a human

Any clause request involving non-compete language, liquidated damages, or injunctive relief. These are outside NDA scope and escalate legal complexity beyond what the agent should handle without flagging.
When the freelancer mentions a prior dispute or potential breach with the same counterparty. The adversarial context changes what the document needs to do, and the agent is not built for that.
When the stated deal size exceeds $75,000 or when one party is a publicly traded or regulated entity. Complexity at that level typically warrants a licensed attorney reviewing the draft before it goes out.
When the requested jurisdiction is outside the agent's four covered legal systems. It should stop, explain the gap, and suggest the user consult a local attorney rather than produce a document it cannot validate.

Pricing model

Per-completed-workflow pricing at $6 per finalized, executed NDA makes more sense than a monthly subscription for a low-frequency use case. A freelancer who signs five NDAs per year pays $30, which is defensible relative to the time saved and the $300-500 a lawyer would charge for the same document. The honest concern is unit economics: at $6 per transaction and a realistic 2,000 active users completing an average of four workflows per year, annual revenue is roughly $48,000, which does not support a team. The model works at scale (50,000+ active users) or with a higher price point justified by the time-savings argument, but the freemium path described in the GTM plan actively undermines both by training users to expect no-cost access.

Differentiation from a generic LLM wrapper

The differentiation is thinner than most vertical agent pitches claim, and this product should reckon with that directly. A freelancer who is comfortable with Claude or ChatGPT can already produce a usable NDA draft in under two minutes at no marginal cost. Sealwax earns its fee by owning the full workflow loop: routing signatures, tracking execution, archiving the signed copy, and surfacing expiration dates, which are the steps that actually fail in practice when a freelancer does it manually. The moat is operational, not generative. For users who are not fluent with AI tools, the structured intake process reduces the blank-page problem. For users who are fluent, the value proposition is closer to "I don't want to manage the signature logistics myself" than "I need AI to write this." That is a real and payable value, but it competes with DocuSign's own template library and HelloSign's basic free tier, not just with generic LLMs.