Magic + Clarity

Legal clarity for public procurement, at machine speed.

A working multi-agent AI platform for legal analysis of Ukraine’s public procurement — surfacing what is deliberately hidden.

Live magilarity.com
Open the live system Explore the platform info@magilarity.com
X-RAY · clause analysis
The Problem

8–16 hours of a lawyer’s work per tender.

Hidden discriminatory requirements
Criteria written to exclude all but one bidder.
Tailored specifications
Specs reverse-engineered from a single product.
Anticompetitive collusion
Coordinated bids that fake competition.
One-sided contracts
Terms stacked against the supplier.
Legal help out of reach for SMBs
Hours of counsel that small business can’t afford.

MAGILARITY automates this entire process — from a tender number to a ready complaint.

What you get

Enter a tender number — get a full legal analysis.

Violations report
Every discriminatory clause, flagged with legal grounds.
Document checklist
Exactly which documents the tender requires.
Bid analysis
Every submitted document checked against every requirement.
Contract analysis
One-sided terms surfaced before you sign.
Documentation builder
For the buyer: draft documentation with risk warnings as you build it.
Complaint
A draft complaint with grounds, precedents, the fee and the deadline.
Architecture

10 specialized AI agents

Each agent does one thing well and passes its result to the next — and every conclusion is anchored to a quote, a norm and, where practice exists, a decision number.

In service
oopz_researcher
Mines OOPZ appeal rulings for precedent.
In service
tender_doc_researcher
Parses tender docs for discriminatory terms.
In service
bid_researcher
Checks every submitted document against the requirements.
In service · beta
complaint_creator
Drafts the complaint against the documentation — fee and deadline included.
View docs →
amku_researcher In service
Builds the knowledge base of AMCU evidence and legal reasoning for investigations.
In service · beta
td_creator
Tender documentation builder for public buyers — with risk warnings from real appeal practice.
View docs →
In service
news_curator
A daily procurement news feed from official sources — in two languages, on the site and in Telegram.
View docs →
Next wave — designed
Designed
investigation
Orchestrates bid-rigging investigations following the competition authority’s own methodology.
View docs →
Designed
complaint_researcher
Drafts OOPZ complaint decisions for case officers.
View docs →
Designed
bid_creator
Generates the supplier’s bid document package from the requirements checklist.
View docs →
Market

The scale of the market.

₴800B+
Annual ProZorro volume
200K+
Active participants
30K+
Public buyers
€2T
EU procurement market

The same system scales to any jurisdiction.

Knowledge base

What the system reads.

Auto-updated every day
~287,500
Legal & regulatory documents
4,402
OOPZ appeal rulings
87
AMCU antitrust decisions
170
Acts in the norm graph — split into articles, parts and paragraphs
~59,000
Norm units across three edge layers, resolved on a date
~782,000
Indexed Prozorro procurements

Legislation is not stored as flat text: every act is split into articles, parts and paragraphs, and every query is answered as of a date. That is how the system tells “in force” apart from “applicable”: the new procurement law has been in force since 24.06.2026 but applies only from 24.03.2027.

Roadmap

Where we’re headed.

Seven agents are in service, three are designed. The constraint is not code but the volume of knowledge in the bases. We promise no delivery dates.

Tender documentation analysis Shipped
Bid analysis Shipped
Complaint drafting Shipped
Tender documentation builder Shipped
Norm graph with nightly self-sync Shipped
Public site, news feed and cabinet Shipped
Production deployment (EU), HTTPS, backups Shipped
Populating the appeal and competition practice bases In progress
Category coverage in the documentation builder In progress
investigation — bid-rigging investigations Next
complaint_researcher — the appeals body’s side Next
bid_creator — the supplier’s bid package Next
Semantic search over practice, court practice Next
Documentation

Detailed technical specs.

Architecture of each agent, analysis methodology, infrastructure and roadmap — open access.

tender_doc_researcher
12-step pipeline, 16+2 analysis criteria, 13 procurement categories, multi-lot tenders and TD amendments.
View docs →
bid_researcher
Cascade sieve for document verification, ensemble LLM for tech specs, SAFEGUARD mechanism, 18+ chapter DOCX report in phase chronology.
View docs →
oopz_researcher
10-step OOPZ ruling pipeline, two-factor anti-hallucination, self-improving system, quality scoring A+→D.
View docs →
amku_researcher
Full-text AMCU ruling analysis (Gemini 2.5): evidence, legal reasoning chains, objections & rebuttals — the knowledge base powering investigations.
View docs →
platform
The public site, cabinets for both sides of the market, the API and task queue, plans and quotas, notifications, the operations console.
View docs →
infrastructure
Eight services in production, five databases, the norm graph and its nightly chain, document processing, security, LLM usage.
View docs →
roadmap
What is done, what is in progress, what comes next — and what we deliberately will not do. No promised dates.
View docs →
complaint_creator
The four pillars of a complaint, grounds selection, and fee and deadline resolved from the norm graph on the procurement’s dates.
View docs →
news_curator
Collection rules, own bilingual summaries, the holiday calendar, legislation changes from the graph delta, Telegram publication.
View docs →
Repository
All project documentation is publicly available on GitHub.
MagilarityAI/magilarity.com →

Built entirely with Claude and Claude Code. Zero prior programming experience → a production-ready platform in 6 months.