Legalcomplex builds and runs its own products for legal intelligence. The design and engineering behind them are available to your team. Our principles are to build token-efficient, supply chain-secure solutions.
Each one links straight at the running thing, not at a page about it.
A research workspace that shapes a grounded answer into whatever display: 19 of them, from a table to a timeline to a map.
So you can see what the AI actually found, check it at the source, and hand it to someone else on any device without rebuilding it.
One search across tax, compliance and the wider law, with legislation and case law cross-linked across eight jurisdictions.
So a question that spans countries stops meaning a different national site, and a different search box, for every one of them.
Fifteen years of funding rounds and M&A across more than 31,900 tracked companies, kept current by a daily sync.
So founders, investors and strategy teams can see who raised what and who bought whom in one place, instead of assembling it from press releases.
A single secure gateway to every interaction your organisation has with an AI model, running inside your own infrastructure.
So regulated enterprises, law firms and government entities can put AI on real work without a document ever leaving their domain.
A contract platform where the contract is the core object, e-sign is the way in, and the identity check behind it is reusable.
So a business is verified once instead of at every counter, and signing stops being the slowest part of the deal. Suriname first.
Six tools over one protocol, putting the Spark market data inside the AI client your team already uses.
So you ask market questions in Claude or Cursor, where the work already happens, instead of keeping another tab open to look them up.
A monthly engagement for teams building AI into legal, tax and regulated work. Not an agency and not a generalist consultancy: someone who has built, shipped and verified in this market, working on your product.
Interfaces for work that has to be checkable. Design canvases, behaviour prototypes that run without a model call, and a component library the shipping code actually uses.
See the deck →Architecture, retrieval, AI integration and the plumbing underneath: adapters per jurisdiction, schema migrations against a live writer, deploys that roll back on their own.
See what shipped →The part most AI work skips. Evals that fail both ways, registries gated on a committed run, and checks designed so that an absent input can never read as a pass. It is also why Monocle retrieves instead of recalling: the S3 benchmark sweeps 2,716 citation statements past a model with no sources attached, and every model tested lands within a couple of points of chance.
See the benchmark results →Roadmap decisions grounded in the Spark dataset rather than in anecdote: who raised, who bought whom, which segments are actually funded.
See the market dashboard →All of it is live and public, so you can check any claim on this page against the thing itself.
One search across tax, compliance and the wider statute book, answering from the official sources rather than from memory.

The design system behind Monocle, and a prototype you can drive end to end without spending a token.

A curated dataset of who funded, bought and built in legal tech, served as a dashboard anyone can read.

Thirty-one thousand company profiles, built to be found and to stay findable.

The site is its own sample. No third party analytics, CRM, monitoring or services tools connected except for AI services providers.

Public-sector engagements where the same discipline applies: definitions that hold, and a structure someone else can still run after we leave.
AI governance and infrastructure across the branches of Defence: a NATO-compliant AI platform built on Sabaio-like platform, and the legal evaluation frameworks based on AI Act, the GDPR and NATO's rules.
Hear us on AI governance →The Metadata Project: a metadata library of more than 8,000 definitions, a model browser to navigate it, and the guide and governance that keep it usable after delivery.
Read about the project →Engagements are limited, so each one gets real attention. Scope and rate are agreed per engagement.
Tell us what you are building and where it is stuck. If it is not a fit we will say so quickly.
Apply for an engagement