Legal Entity Resolver MCP Server

MCP server for the Mamba Labs Legal Entity Resolver actor on Apify.

Give it a company domain and it returns the registered legal entity behind it: legal name, company number, jurisdiction, status, LEI and VAT number. One flat row per domain, 24 fields, ready for Clay or a CRM.

Install

bash

npx -y @mambalabsdev/mcp-legal-entity-resolver

Claude Desktop

json

{
  "mcpServers": {
    "mamba-legal-entity-resolver": {
      "command": "npx",
      "args": ["-y", "@mambalabsdev/mcp-legal-entity-resolver"],
      "env": { "APIFY_TOKEN": "your-apify-token" }
    }
  }
}

Get an Apify token at console.apify.com/account/integrations.

Tool

`resolve_legal_entity`

Company domain in, the registered legal entity behind it out.

Input	Type	Required	Notes
`domain`	string	yes	A single company domain, for example monzo.com. Protocol and path are stripped.
`legal_name_hint`	string	no	Skips the domain lookup and goes straight to the registers with this name. Use it when you already have the legal name and just want the register record.
`jurisdiction_hint`	string	no	ISO-2 country code. `GB`/`UK` asks Companies House, `US` asks SEC EDGAR, `FR` asks the French Sirene register (SIREN), `NO` asks the Norwegian Brreg register (organisasjonsnummer). Each asks GLEIF alongside. Any other code asks GLEIF alone. Leave empty to query all five.
`match_strictness`	enum	no	`exact` accepts a register record only when the normalized legal names are equal. `fuzzy` returns the best scoring candidate with a confidence below 100 and a warning in `rejected_candidates`. Default `exact`.
`validate_vat`	boolean	no	Runs any VAT number found on the company's own pages through the EU VIES service and returns the name VIES holds for it, as a cross-check against the register name. Default `true`.
`skipCache`	enum	no	`false` uses the cache, 90 days for a resolved company and 7 days for a null. `true` forces a fresh look. Default `false`.

A null is the product, not a gap

Register search endpoints are fuzzy and they always return something. Taking the top search result gives you a confidently wrong company number more often than not. This actor accepts a match only when the legal names are identical after normalization, which is why roughly 6 domains in 10 resolve instead of 10 in 10, and why the 6 are worth acting on.

Read match_method, match_confidence and rejected_candidates before acting on a match. fuzzy strictness is a research mode: it will hand you a confidently wrong company on most domains.

Three registers are queried: UK Companies House, GLEIF and SEC EDGAR.

Billing

You are charged per domain resolved, plus the platform's actor start event. Prices from the live pricing record, read 2026-10-05:

Event	Price per event (USD)	Fires when
`domain-resolved`	$0.007 on the Free plan, down to $0.00595 on higher Apify plans	One domain looked up: the company's own pages fetched and read for a legal name, then up to three company registers queried for an exact match. Charged whether or not a register matches, because the work is the same either way and charging only for matches would push this Actor toward guessing. Not charged when the domain cannot be reached at all.
`apify-actor-start`	$0.00005	Charged when the Actor starts running. Number of events charged depends on Actor memory (one event per GB, minimum one event).

Cached results are 90 days for a resolved company and 7 days for a null. The actor's Apify page carries the current prices. Running this server consumes Apify credits.

What this server does and does not do

It is a thin client for the Apify actor. It passes your input through and returns the actor's output unchanged. Every behavior described above lives in the actor, not here.

This is not a company database and not a credit or risk product. It does not score companies, rate them, or tell you whether to trade with them. It answers one question: which registered legal entity sits behind this domain.

Errors are surfaced, never swallowed. An invalid input, an invalid token, an exhausted balance, a timeout, or a run that returns anything other than a dataset all come back as an explicit tool error rather than as an empty result.

How each call runs

Each call starts the actor run, polls it until it finishes, then reads the dataset. The run is allowed 1,800 seconds. If the run is still going when this call stops waiting, the call returns the run id and a console link instead of a timeout, so the result is never lost.