In your AI client
The MCP connector puts the sourced foundation in front of your assistant. Ask in your own words; every answer carries its European source or a named gap.
Data feeds
Pauhu®'s streams deliver Europe's sourced foundation as it changes. The moment a European source publishes, the update arrives at your endpoint, normalised, sourced, machine-ready. One continuous flow carries twenty-four European languages across fourteen domains. Every Connect subscription includes the streams.
Integration
Pauhu speaks MCP, the Model Context Protocol that AI assistants and agents already use to reach external data. As a data recipient you connect once, and Europe's sourced foundation is available inside your AI client, your code, and your data pipelines, with the source URL on every row or a named gap. Standard MCP, sourced at origin, one flat subscription. One connection reaches all fourteen Common European Data Spaces, and REST and the structured stream are included in the same subscription. The foundation stays sourced at origin; you receive it cited, hosted for you.
The MCP connector puts the sourced foundation in front of your assistant. Ask in your own words; every answer carries its European source or a named gap.
Read sourced rows over REST, routed by domain, each with a paragraph-precise identifier and the source URL behind it.
Point your data warehouse at the stream cursor; the moment a European source publishes, the delta arrives at your endpoint.
Translation memory
The Language data space is the European translation memory and term base, served read-only over an API, and built to be safe for a regulated buyer. Your CAT tools and your systems use it live over that API. It stays read-only, and your text stays on your side, so the cited source of truth stays exactly as published and your confidential text stays yours. The foundation stays at origin, secured and current for you. Every match carries its European source, and it is kept current as Europe publishes. Pauhu returns the exact, cited matches; your CAT tool scores the fuzzy ones and your translator decides.
An authoritative reference you consume. Your queries and your text stay on your side, and the source of truth stays exactly as published.
Your CAT tools and your systems use the term base and translation memory live over the API. AI assistants reach it over MCP. It stays at Pauhu, live and current; you reach it in place. Exact, cited matches in; the fuzzy scoring stays in your tool.
Refreshed as European sources publish, across the twenty-four EU languages; the term base spans fifty-one, the migrant languages included. A static memory ages from the day it ships; this one stays current as Europe publishes.
Machine stream
The streams are typed, schema-stable, paginated, and machine-consumable. Every delta carries:
The sourced rows that changed.
The source URL backing each row.
A paragraph-precise identifier.
A timestamp.
The European publication channel that triggered the update.
Your code reads structured rows, routes by domain, stores against your cursor, alerts on the changes that matter to you. Where a domain is still filling, the stream emits honest-gap markers so your code can route those too. Both behaviours are part of the product.
Update model
Every European source Pauhu® curates has a publication channel. EUR-Lex publishes the Official Journal. EMA publishes safety updates. EuroVoc publishes vocabulary refreshes. The moment that channel emits a new instrument, amendment, recital, terminology entry, or scientific record, Pauhu® normalises it, sources it, and pushes the delta to your stream cursor.
Your systems sit ahead of the publication cadence, on the foundation, with the source URL in hand.
Coverage
The streams cover the same fourteen domains as the foundation: agriculture, cultural heritage, energy, finance, green deal, health, language, manufacturing, media, mobility, public administration, research and innovation, skills, tourism.
Where the foundation is dense, the streams carry sourced rows row by row, with source URLs and paragraph-precise identifiers. Where it is still filling, the stream carries an honest gap that names what would close it. The cited-or-honest-gap contract holds in the stream the same way it holds in the prose answer and the REST response.
Consume
For ingest at scale, point your data warehouse at the Pauhu® stream cursor. The cursor advances as you consume; resume from any point. For event-driven workflows, subscribe to the push channel; updates arrive at your webhook the moment Europe publishes.
curl -H 'Authorization: Bearer YOUR_KEY' \
'https://api.pauhu.eu/v1/stream?domain=public-administration&cursor=YOUR_CURSOR' Full schema and authentication shapes resolve at /docs/. Every Connect plan includes the streams; pricing is on /pricing/.
Close