The infrastructure documents are built on.

DARIL helps companies generate reliable documents at scale — a single deterministic layer any agent or workflow can call.

Talk to the founders See how it works
Call DARIL
Our promise

Same inputs. Same output. Every time.

Is it your approved language? Of course. DARIL is a single deterministic layer that any system, agent, or workflow calls when it needs a reliable document.

Party: Acme Corp Term: 24 months Governing law: New York If deal value > $1M, add Indemnity Rider B
Rendered by DARIL, 10,000 times
Run 1Identical
Run 2Identical
Run 10,000Identical
The problem

AI agents are everywhere. Reliable documents are not.

Every company needs documents out the door fast, and right. The deal closes and the contract has to go out tonight. Today there are two ways to produce it. Both are broken.

Not safe

AI generation

Fast, but probabilistic. Ask for your standard contract and it “improves” it: restructured payment terms, a rewritten indemnity, a provision nobody asked for. Grounded in your language, but not your language, and different every time. That's not automation. That's liability.

Does not scale

Legacy automation

Deterministic but brittle — every change means more hand-coding, and it doesn't scale. Hand-coded conditional logic in proprietary markup languages. $50K and 8–12 weeks per document set. Most organizations plateau at a handful of templates before hitting the authoring wall.

Deterministic + intelligent

DARIL

Templates authored through conversation, documents rendered with algorithmic precision. AI is the interface. The platform is the guarantor.

The Stripe for document generation

DARIL is the best of both worlds.

It's shared infrastructure that any agent calls. Upload a document or describe what you need. DARIL discovers the logic, codifies it through conversation, tests every permutation automatically, and renders documents with algorithmic precision.

Stripe made payments programmable and guaranteed. DARIL does the same for documents.

How to use DARIL

Chat with DARIL

Domain experts author and use templates through guided conversation, on any device. What took months takes hours.

Call DARIL via MCP

Any LLM or AI tool calls DARIL through the Model Context Protocol. The model handles reasoning. DARIL guarantees the output.

Access via agent / API

Fully headless. Agentic workflows call DARIL directly, no human in the loop, with deterministic results every time.

The team

We didn't research this problem. We lived inside it for 20 years.

We're founding team members of Contract Express — one of the most successful document automation platforms ever built. Acquired by Thomson Reuters.

Geoff Goldberg

Geoff Goldberg

Co-founder

20-year pioneer in document automation. Founding team at Contract Express (VP/GC, Product Manager). Former Director of Document Automation Services at Thomson Reuters. Knows exactly why the first generation fails to scale — and what it takes to fix it.

Anatoly Soyfer

Anatoly Soyfer

Co-founder

MBA. CMO of Contract Express, where he drove 300% ARR growth. Separately scaled a small business from $1M to $40M ARR in under three years. Integrated Contract Express into Thomson Reuters' product portfolio post-acquisition.

20 years inside the problem
  1. Contract ExpressFounding team
  2. Acquired by Thomson Reuters2015
  3. DARILNow

Teams we built document automation for at Contract Express

Why now

Five convergent forces.

  1. 2023 LLMs infer structure Models can now reliably infer document structure and semantics — enabling inductive discovery and conversational authoring rather than manual encoding.
  2. 2024 The coding agent paradigm The model that lets developers converse with AI to write code now lets domain experts converse with AI to author templates.
  3. 2024–2025 MCP creates the distribution layer DARIL sits underneath enterprise AI tools as callable infrastructure — transforming it from a product into a layer.
  4. 2026+ Regulation demands determinism The EU AI Act and Colorado's AI Act require auditability and human oversight. DARIL is compliant by architecture, not by process.
  5. 2026+ Agentic ecosystems take flight Agent-to-agent ecosystems need a first-class deterministic document layer. That layer is DARIL.
The shape

Why an octahedron?

It's an octahedron — a structure you've likely seen in bridge trusses, space frames, and molecular models. One of the most efficient load-bearing geometries in engineering. We chose it because, like DARIL, it is an interesting piece of infrastructure.

Opposing pyramids sharing a rectangular base that represents a document. Each apex projecting onto it — where intelligence meets structure.

  1. AI suggestion
  2. Approval
  3. Template
  4. Document

AI suggests. A user approves. A template is created. A document emerges.

Ready to talk?

We'd love to hear from you.

DARIL is in private development. We're selectively engaging enterprise design partners and investors.