The semantic layer for UK commercial finance data.

Registers below us. Your decisions above us.

Creditbase sits between the public record and your judgement. Below us: Companies House, The Gazette, iXBRL accounts, procurement and payment-practices data. Raw, rate-limited, each speaking its own dialect, none sharing identities or timelines. Above us: your underwriting, your monitoring, your models, your agents.

We do not compete with either side. We make the layer in between dependable, and we keep it open.

Six kinds of meaning

A semantic layer is not a bigger database. It is a system of meaning with six parts:

Entities

One resolved identity for each company, person, funder and address across every source, with match confidence shown

Events

A canonical vocabulary of commercially meaningful happenings, from charge.created to accounts.overdue, mapped from raw register payloads

Relationships

Typed, evidence-backed links: controls, secures, succeeds, shares a director with

Measures

Normalised financial concepts from full, abbreviated and micro-entity iXBRL accounts, with published definitions

Time

Point-in-time semantics: what was observed, when it took effect, what superseded it

Provenance

Every assertion cites its source, arrival time, transformation and confidence

Why agents need it

An LLM pointed at raw registers is confidently wrong: it merges namesakes, misses superseded filings and misreads micro-entity accounts. Connected to the semantic layer through MCP, the same agent reasons over resolved entities and point-in-time facts, and can cite the register document behind every claim.

What you can build on it

• Company intelligence inside your own product
• Portfolio monitoring in your CRM or servicing platform
• Underwriting and risk features with documented lineage
• Agent tools through the MCP server
• Research datasets that support honest backtests

Access

The layer is reachable however you build:

• REST API
• MCP server
• Webhooks and event streams
• Bulk and warehouse delivery
• A free low-volume tier for prototypes and research

An open schema, published to be adopted

The event vocabulary, feature definitions and model cards are published openly, and we want other people to use them: fintechs, academics, software companies, even competitors. A shared vocabulary for UK company data makes everything built on it better, and it is how open infrastructure is supposed to work.

The first public schema release is being prepared with pilot partners now. If you want to review it before it ships, get in touch.