01.
Sources and fetchers
Fetchers work in tiers, from an API to an RSS feed, a plain HTTP page and, last, a headless browser, so a regulator with no feed or API is still read where it publishes.
All projects
// RegTech · 2026
Regulatory intelligence for emerging-market fintech
Client
EMRules, in-house
Services
Core technologies
TypeScriptLangChainLangGraphNx
Status
Live
// The Brief
Put emerging-market regulation in one English feed.
// Background
From scattered rules to one feed
Compliance, legal and expansion teams at fintechs licensed in more than one emerging market answer to every central bank and supervisor there, each publishing in its own language and format.
We built every layer: tiered fetchers for each regulator, an AI pipeline that extracts and translates each document, and a platform with plans, a REST API and an MCP server.
Live at emrules.com, reading each regulator where it publishes every 24 hours, with English beside the original and a 14-day trial over the API or MCP that needs no card.
241
Regulator websites, feeds and APIs
129
Countries with a listed regulator
39
Original languages across the articles
// Challenges
What made it hard
Each regulator publishes through a channel of its own: a website, an RSS feed or an API.
Some regulators publish no feed or API, so their pages must be read the way a browser does.
Every document has to reach the reader in English, with the original kept beside it.
Every AI run has a cost, so spend must be tracked per run and stopped when it runs away.
Usage has to be metered and billed by plan, with limits enforced on every request.
// Value delivered
What it delivers
One English feed of rules and notices, with the original publication beside each article.
A feed teams can narrow by document type, regulatory topic and credibility score.
A REST API and an MCP server, so code and AI agents read the same feed with one key.
Plans with usage billing, rate and quota limits at the API edge, and a no-card trial.
A cost ledger on every pipeline run, with a circuit breaker that can halt it.
// Inside the Product
What we built
Each part of the product in turn: what it does for the people who use it, how we built it, and the real screens where it runs.
02.
Extraction and translation
An agentic pipeline built with LangChain and LangGraph extracts each publication into a structured record, summarises it and translates it to English. The reader sees the English beside the original, which prevails wherever the two differ.
03.
Classification and credibility
Each article is classified by document type, from regulation and circular to enforcement action and press release, given a credibility score and tagged by regulatory topic, so a team can filter the feed to what it must act on.
04.
API, MCP and plans
A REST API and an MCP server serve the same feed with one key, to code or to agents in Claude, Cursor and GitHub Copilot. Plans bill by usage, with rate and quota limits at the API edge.
// Under the Hood
Stack and integrations
Your engineers can judge the build for themselves: the languages and frameworks the product runs on, and the services, platforms and tools it connects to.
// Technologies
What it runs on
TypeScriptLangChainLangGraphNxNestJSNext.jsPostgreSQL
// Integrations
What it connects to
PolarClaudeClaude CodeCursorGitHub CopilotVS Code CopilotGoogle GeminiGemini CLI
// See It Live
Open the product
See the work running outside this case study: each link opens the live product or one of its public pages in a new tab.
What's next
Let's ship
the next one.
Tell us what you're building and where it stands. We'll scope it with you on a call and give you a straight answer on fit.
- 01
Book a call
Pick a time on our calendar and tell us what you want to build.
- 02
Shape the scope
Together we name the users, the first release and what success looks like.
- 03
Get a proposal
We send a written plan with the scope, the steps and the engagement that fits.
- 04
Start the work
We begin with Discovery and show you working software early and often.