Legal Engineering
Food Law Built into AI Tools
I design and build legal knowledge bases, automations and AI tools for EU and Polish food-law compliance. I start from the legislation and the registers and finish with a tool your regulatory, QA or marketing team actually uses.
Legislation turned into data, rules and processes
Legal engineering sits where law meets technology. I turn legislation – including Regulations (EC) No 1924/2006, (EU) No 432/2012, (EU) No 1169/2011, (EU) 2015/2283 and the Polish Act on Food and Nutrition Safety – into structured data, rules and process steps that a tool can check consistently, with a reference to the source.
Legal analysis stays with the lawyer. A tool organises the material, shows the source and the status and flags risks – it does not approve a label and does not replace a legal opinion.
What I have built and maintain
These tools are publicly available, so you can see how I work before we discuss a project.
supplemental.pl MCP server
A public, read-only MCP server at https://mcp.supplemental.pl/mcp – no login, no key. Eight tools: searching and retrieving supplemental.pl articles, the EU Register of nutrition and health claims, the EU Novel Food Catalogue, C.L.A.I.M.S. claims, the GIS notification steps and the Nutrimetr calculator. Works with Claude, ChatGPT, Cursor and other MCP clients.
C.L.A.I.M.S.
An AI tool for health-claims analysis (app.foodlaw.ai): claims authorised under Regulation (EU) No 432/2012 and the botanical claims “on hold”. Botanicals on hold are treated as not authorised. It is not an official European Commission register.
Open C.L.A.I.M.S. →Nutrimetr
A nutrition calculator: a per-100 g table from the recipe, %NRV, preliminary eligibility for nutrition claims under Regulation (EC) No 1924/2006, allergens and QUID hints. Deterministic calculations, no AI. It does not replace laboratory testing.
Open Nutrimetr →What I can build for your company
Every project starts with a specific process in your company, not with the technology.
Legal knowledge bases
Structured regulatory knowledge used by people and by AI models.
- structured sources: legislation, EU registers, guidance and case law,
- every entry with its legal basis, the date of the legal status and a link to the source,
- PL/EN/DE language versions for teams and export markets,
- update rules for changes in legislation and registers.
Claims checking
Repeatable checks of claims before they reach the label, the website or an influencer.
- matching an ingredient and dose to claims under Regulation 432/2012,
- flagging botanical claims on hold as not authorised,
- checklists for marketing and e-commerce,
- an escalation path to the lawyer for borderline cases.
Label and GIS notification workflows
Processes in which nothing gets lost between R&D, QA, marketing and the designer.
- label checklists under Regulation 1169/2011 (ingredients, warnings, QUID, allergens),
- GIS notification steps under Article 29 of the Polish Act on Food and Nutrition Safety and the data required,
- novel food status of ingredients from the Novel Food Catalogue,
- a clear distinction: a GIS notification ≠ a novel food authorisation.
Structured regulatory data
EU registers and internal product data in a format that can be searched and combined.
- data models for ingredients, claims, novel food statuses and recipes,
- linking data with nutrition values (as in Nutrimetr),
- export to spreadsheets, an API or a database,
- source and licence noted for each dataset (e.g. CC BY 4.0 for European Commission data).
MCP and AI assistant integrations
Connecting regulatory knowledge to the tools your team already uses.
- MCP servers – public or internal – for Claude, ChatGPT, Cursor and other clients,
- read-only tools with a source in every answer,
- tool descriptions and disclaimers that limit the risk of misuse,
- connecting to the public supplemental.pl server where that is enough.
Internal tools for regulatory teams
Simple applications and automations tailored to one process.
- calculators, search tools and checklist forms,
- automated pre-screening of documents and label drafts,
- answer and report templates with room for the lawyer’s decision,
- documentation and handover to your team.
How a project runs
- Process and sources. You describe the process (e.g. claims checking or preparing a notification) and share sample documents. We agree which legislation and registers are the source of truth.
- Model and rules. I translate the legislation into a data structure and rules with their legal basis, and identify which decisions stay with a person.
- Prototype on your examples. We test the tool on real labels, recipes and claims from your portfolio.
- Deployment and maintenance. I hand the tool over with documentation and we agree how the data will be updated when the law changes.
Who it is for
- producers of food supplements and functional food,
- importers and distributors placing products on the Polish and EU market,
- regulatory, QA and marketing teams that repeat the same analyses,
- teams deploying AI assistants that want answers grounded in legal sources.
Ground rules
- A tool does not replace a legal opinion. A result from C.L.A.I.M.S., the MCP server or Nutrimetr is a map, not approval of a label.
- Every tool answer points to its source: a legal act, a register or an article.
- Botanical claims on hold are not authorised claims – every tool labels them that way.
- Scope and billing depend on the process and the number of sources; we agree them after the first conversation. I do not publish a price list.
- The introductory call is to assess whether the project is a fit. It is not a free legal opinion.
Have a process worth automating?
Describe the process, the sources and the tools your team uses in a few sentences. I will reply whether and how it can be turned into a knowledge base or an AI tool.