The brief.
Contracts decide who owes what to whom — and most of the people signing them can't afford a $400/hour lawyer to translate the small print. Lexly was founded on the premise that AI could close that gap: a freelancer, founder or small-business owner should be able to draft, understand and sign a legally binding contract in the time it takes to write the brief.
Devmint partnered with Lexly to engineer the platform end-to-end — the AI generation pipeline, the contract analysis surface, multi-jurisdiction e-signature, the billing layer, and the operational infrastructure to keep it running under real legal scrutiny. Today the product is live, processing contracts across six jurisdictions and covering an estimated 4.5 billion people who fall under those legal frameworks.
What we shipped.
- Module · 01Contract ForgePlain-English brief → legally formatted contract. Jurisdiction-aware clauses, party-aware terms, and a redline-ready output the user can export to Word or send directly to e-sign.
- Module · 02Contract SenseUpload any contract. Get a risk score, a clause-by-clause explanation in plain English, and a list of the terms that would normally cost $400/hr to spot.
- SigningMulti-party e-signLegally binding signatures across ESIGN (US), eIDAS (EU/UK), IT Act 2000 (India) and three more frameworks. Identity verification, SHA-256 tamper seal, audit trail to PDF.
- SecurityEnd-to-end encryptionAES-256 at rest, TLS 1.3 in transit, per-tenant encryption keys, and a privacy posture that doesn't keep your contracts on a vendor's server.
How we built it.
The technical question wasn't “can an LLM write a contract” — it was “how do you ship a system a lawyer would be willing to put their name next to.” The answer was layered: retrieval over jurisdiction-specific clause libraries, deterministic validators on every output field, an eval harness running against a corpus of human-reviewed contracts, and a human-in-the-loop surface for high-risk edits.
1. Jurisdiction-aware generation.
Each supported framework (ESIGN, eIDAS, IT Act, ICA, ETA, CECA) has its own clause library, validation rules and signing requirements. The generation pipeline routes through the right library before the LLM ever sees the brief — so the model is never asked to invent the law, only to translate the user's intent into a library the legal team has already vetted.
2. Contract analysis as structured output.
Contract Sense doesn't produce free-form summaries. Every analysis is a structured object — clauses, parties, obligations, durations, risk flags — that the UI renders. Free-form explanation is generated from the structured object, not the other way around. That separation is what makes the risk score auditable.
3. Signing as a first-class concern.
We built the signing layer ourselves rather than wrapping DocuSign, because the unit economics of a $0–$99/month product can't carry a $0.50-per-envelope vendor markup. SHA-256 document locking, AES-256 storage, IP-anchored consent records, and a regulator-friendly audit export.
The stack.
AI
- Anthropic Claude · OpenAI GPT
- Custom retrieval over clause corpus
- Structured output validators (zod)
- Langfuse for traces + evals
App + signing
- Next.js 15 · TypeScript
- Postgres · Redis
- Stripe billing · multi-tier plans
- Native e-sign · SHA-256 + AES-256
“Devmint shipped a working platform faster than I've seen anyone ship a static marketing site. Six jurisdictions on day one. The retrieval architecture means a new framework is a configuration change, not a rewrite.”
— Founder · Lexly
Outcomes.
Lexly is live in production at lexlyai.com. The platform currently supports six legal frameworks, covering — by the team's own estimate — roughly 4.5 billion people. Pricing starts at $0 and ceilings at $99/month, against enterprise CLM tooling that starts at $60,000/year.
Devmint continues to operate the AI layer on a quarterly basis: weekly eval reviews, monthly cost reviews, and new-jurisdiction onboarding as the product expands.