Structured data from unstructured filings. We turn official public filings — iXBRL, XBRL, structured JSON and PDF — from registries such as Companies House, Erhvervsstyrelsen, Brønnøysundregistrene and SEC EDGAR into normalized financial data: revenue, profit, equity and ratios, using a consistent schema regardless of country of origin.
Filings without structured data (PDF or plain HTML reports) are read with AI, and an MCP server (in beta) lets AI agents query the data directly. Both are available today.
Example (abridged): filing → normalized API response
A traditional database answers a relatively simple question:
Nexus Filings addresses a different problem:
The registries already publish the data. Nexus makes it usable across markets.
A researcher can work with one country's filing format. A data product cannot scale that way. Nexus provides one normalized interface across jurisdictions, with the provenance of every value.
A database can store information on thousands of companies. But storing documents doesn't necessarily mean understanding their content, relating it, normalizing it or making it comparable.
Nexus Filings starts from the original filings and regulatory documents and builds a layer of information on top of them. If you already work with US filings through SEC EDGAR, Nexus sources US companies directly from EDGAR too — and extends the same normalized schema to markets EDGAR doesn't cover, like the UK, Danish and Norwegian registries: one data layer instead of one document format per country.
Each market has its own reporting frameworks and taxonomies — including UK accounting standards, IFRS and Danish GAAP — so financial data is structured differently across registries.
Documents downloaded straight from each country's official registry — iXBRL or XBRL when available, and PDF or unstructured HTML read with AI when not — with no intermediaries and no data aggregators.
Context, period and taxonomy are resolved and mapped to a single schema — revenue, profit, equity, ratios — regardless of country of origin.
A REST request returns the same format, always, with the provenance of every field marked.
Every value is either sourced directly from the original filing or derived from other sourced values using a documented calculation. A missing value is never invented — if it can't be derived with confidence, it is marked as unavailable.
We start with the registries where financial filings are already rich and structured. The roadmap continues across Europe and Asia.
Add one URL to Claude Code, Cursor or any MCP-compatible agent and ask in natural language. Use your API key in a header, or sign in with your account (OAuth 2.1) for remote connectors that require it. It uses the same plans and usage units as the REST API.
search_companies
Find a company by name and get its id. Uses no allowance.
get_financials
Normalized financial statements for a company, with the formula of every derived field.
batch_financials
Several companies in one call (Pro, Pro Team and Business).
screen_companies
Filter processed companies by net margin and jurisdiction (Pro, Pro Team and Business).
Start for free. 1 company returned = 1 usage unit.
Batch and screening requests are charged by the number of companies returned, not by the number of API calls. If a request returns 50 companies, it uses 50 units from your monthly allowance.
To try the schema and explore the coverage.
For analysts and builders with regular usage. Try it free for 7 days, no card.
A shared workspace, not a shared login.
For products and data teams in production.
New companies processed: the first request for a company that has not been processed yet (typically a non-listed company) downloads and normalizes its filings on the spot. Each plan includes a monthly number of these. Companies that are already processed don't count against it. See the FAQ.
Annual billing is 2 months free compared to paying monthly (16.7% off) — the same monthly usage allowance and limits apply, they just renew every month instead of once a year.
No. Nexus Filings delivers normalized public historical data. It does not recommend buying or selling anything — analysis and decisions are the responsibility of whoever uses the API.
From each country's official registries — Companies House in the United Kingdom, Erhvervsstyrelsen in Denmark, Brønnøysundregistrene in Norway, and the SEC's EDGAR system in the United States. We don't buy or resell data from aggregators: we parse the iXBRL, XBRL and structured JSON that companies are legally required to file, and when a company files only a PDF or an unstructured HTML report we read it with AI. A value that is not in the filing is never invented.
The unit is a company record: 1 company returned = 1 usage unit, and there is no charge per API call. Searching by name uses 0 units. A company's detail uses 1 unit. A batch uses 1 unit per company returned (ids not found or not processed yet cost 0), and a screening uses 1 unit per company in its results (0 results, 0 units). If a request returns 50 companies, it uses 50 units from your monthly allowance. On the Free plan each request returns a single fiscal year (the latest, or the one you choose with fiscal_year), so viewing several years takes several requests; paid plans return the whole history in one request. Failed requests and retries while a new company is being processed cost nothing: you are charged once, when the data is delivered.
Companies in the main catalog (large and listed) are already processed and answer instantly. When you request a company that nobody has requested before, its filings have to be downloaded, read and normalized on the spot, which takes real processing. Each plan includes a monthly number of those first-time requests (Free 5, Pro 50, Pro Team 100, Business 500). Requests for companies that are already processed never count against it.
Registry aggregators usually cover identity, directors and corporate structure. Nexus Filings specializes exclusively in structured financial data — they are complementary layers, they don't compete for the same thing.
Some requests require retrieving, parsing or normalizing a filing that has not been processed yet. In those cases the response can take longer than a request against already-normalized data, up to a few minutes the first time. Instead of holding the connection open (many HTTP clients time out anyway), the API returns 202 immediately and you repeat the same request every few seconds until it returns 200 with the data. Retries while it is processing don't cost anything: you are charged once, when the result is delivered. See the technical detail in the documentation.
Yes, for one thing: when a company files only a PDF or an unstructured HTML report (common for large and audited companies), the document is read with an AI model to extract the figures. The model is instructed never to guess: anything that is not written in the filing stays empty and is marked as unavailable.
Yes, in beta. Point Claude Code, Cursor or any MCP-compatible agent at the MCP URL from your dashboard, using your API key or signing in with your account. It exposes search, company financials, batch and screening, and uses the same plans and usage units as the REST API.
EDGAR is the United States SEC's filing system, and it's actually one of our direct sources: US companies in Nexus Filings come straight from EDGAR's own structured data. UK, Danish and Norwegian registries publish similar filings, but each in its own format (iXBRL, XBRL, structured JSON, PDF, plain HTML) and without a single shared schema like EDGAR's. Nexus Filings normalizes all of them — EDGAR included — into one schema, one API and an MCP server, with the provenance of every value.
Yes. Once your email address is verified, you can start a 7-day free trial of the Pro plan from your dashboard. No card is needed. During the trial you get every Pro benefit, with up to 10 new companies processed instead of 50. When it ends your account returns to the Free plan automatically and nothing is charged. There is one trial per person.
Yes, with Pro Team or Business. Each is a shared workspace: the owner invites members by email, and every person — owner included — gets their own individual API key, never a shared one. Everyone in the workspace draws from the same monthly allowance; it is not multiplied per user. Only the owner manages billing and can invite or remove members. Pro Team includes 3 seats (up to 2 extra) and Business includes 6 (up to 4 extra), each extra seat $25/month, billed together with the plan.