DataLedger MCP server for AI agents
Make your AI actually know UK companies.
Connect Claude or another AI agent to DataLedger and ask about any UK company in plain English. Owners, ultimate parents, subsidiaries and accounts, all from filed records, not guesses.
What is MCP? A standard way to plug outside data into your AI, like adding a new skill. The DataLedger MCP server gives AI agents structured UK company financials, ownership records and group structures from Companies House.
Connect your AI25 free credits on sign up, no card needed. Set-up takes minutes and needs no code.
Being asked to use AI at work? This is a good place to start. Tested with Claude, and works with Cursor, VS Code, n8n and other tools that support MCP.
I found 38 matches. The strongest three:
- Example Fabrication Ltd, Wigan. Net assets £2.1m, up 11%. Owner born 1964, 75 to 100%.
- Example Precision Ltd, Bolton. Net assets £1.4m, up 6%. Owner born 1960, 75 to 100%.
- Example Tooling Ltd, Preston. Net assets £1.9m, up 9%. Owner born 1958, 50 to 75%.
Shall I put all 38 in a table for Excel, or research the top three?
Why connect DataLedger
AI is the interface. DataLedger is the record.
Everyone is being asked to use AI. The difference is whether it knows what it is talking about. Connect DataLedger and the same AI you already use becomes more powerful, more accurate and more flexible.
More powerful
Your AI can answer the questions it could not before: who owns this company, how old are they, is it growing, and what else sits in the same group.
More accurate
Each answer comes from filed records, not memory or web pages, so you can check where it came from before you rely on it or show it to a client.
Yours to shape
Filter, compare, refine and reshape in a sentence. Ask for a shortlist, a table for Excel, a briefing or a report. The data does not decide how you work.
A real session
Here is one in full.
This is a real Claude session with the DataLedger MCP server connected. The user typed one plain English question. Claude searched 71 companies, pulled financial data on the most promising candidates in parallel, and returned a structured acquisition shortlist.
Salford area: acquisition shortlist
Criteria: active, individual owners only, over 5% asset growth, equity under £1m, non-dormant, service SIC codes
Tier 1: Strong targetsCompany names in this example have been changed. The searching, filtering and analysis are as the session returned them.
How it works
You ask. Your agent looks it up.
You never see the technical part. Your agent decides which lookups to run, runs them against DataLedger, and writes the answer in plain English.
Connect your AI
Three steps. Whichever AI agent you use.
Get a free key
Sign up at hub.dataledger.uk. You get 25 free credits straight away, with no card.
Add DataLedger to your AI
Paste one web address and your key into your AI tool. That is the whole set-up.
Ask your question
Switch DataLedger on in a chat and ask in plain English, as in the demo above.
Every route uses the same three details. Pick the tab that matches your tool. You need a DataLedger API key, free from hub.dataledger.uk.
The simplest route, with no code. Works in Claude on the web and in the desktop app. Tested by us.
- Get your API key. Sign up at hub.dataledger.uk.
- Add a custom connector. In Claude, go to Customize, then Connectors, then Add custom connector.
- Enter the server URL:
https://api.dataledger.uk/mcp - Add your key. Set Authentication to No sign-in. Under Request headers, choose
x-api-keyand paste your key as the value. Click Add. - Switch it on and ask. In a chat, click +, then Connectors, and make sure DataLedger is on.
Claude Desktop config file (if you prefer a file)
Recent versions that support native HTTP connections can use this. Open the file from Claude Desktop settings, then Developer, then Edit config.
{
"mcpServers": {
"dataledger": {
"type": "http",
"url": "https://api.dataledger.uk/mcp",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}Older versions can use mcp-remote as a bridge. Node.js must be installed.
{
"mcpServers": {
"dataledger": {
"command": "npx",
"args": [
"mcp-remote",
"https://api.dataledger.uk/mcp",
"--header",
"x-api-key:YOUR_API_KEY"
]
}
}
}For developers and anyone building with an AI coding assistant. Paste one block into your client, add your key, and restart it.
One command writes the config for you and picks it up on next start.
claude mcp add --transport http dl-mcp \ https://api.dataledger.uk/mcp \ --header "x-api-key: YOUR_API_KEY"
Add this to ~/.cursor/mcp.json, or .cursor/mcp.json inside a project.
{
"mcpServers": {
"dataledger": {
"url": "https://api.dataledger.uk/mcp",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}Add this to .vscode/mcp.json. VS Code asks for your key the first time and stores it securely, so it never sits in the file.
{
"inputs": [
{
"type": "promptString",
"id": "dataledger-key",
"description": "DataLedger API key",
"password": true
}
],
"servers": {
"dataledger": {
"type": "http",
"url": "https://api.dataledger.uk/mcp",
"headers": {
"x-api-key": "${input:dataledger-key}"
}
}
}
}Add this to ~/.gemini/settings.json.
{
"mcpServers": {
"dataledger": {
"httpUrl": "https://api.dataledger.uk/mcp",
"headers": {
"x-api-key": "YOUR_API_KEY"
}
}
}
}ChatGPT's MCP connectors expect an OAuth sign-in, and DataLedger uses an API key. The practical route is a custom GPT with Actions, which calls the DataLedger API directly. The result is much the same: a GPT that searches UK companies and reads financials and ownership in conversation.
- Create a custom GPT. In ChatGPT, go to Explore GPTs and choose Create. Add the instructions below.
- Add an Action. Choose Create new action, set Authentication to API key, and the method to Bearer. Paste your DataLedger key.
- Paste the OpenAPI schema. Copy it from our AI agents and LLM workflows page into the Action schema field.
- Test and save. Run a company search in the preview. Once it returns results, save the GPT.
Suggested GPT instructions
You are a UK company data assistant powered by DataLedger. Use the DataLedger API when the user asks for UK company searches, company profiles, financial data, ownership data, SIC filtering, location filtering, or Companies House-derived structured data. Explain clearly when data is based on electronic Companies House filings. Do not imply that turnover or profit and loss data is available for every company. Where profit and loss data is unavailable, explain that many UK SMEs do not disclose full P&L figures. When searching, prefer structured filters such as SIC code, local authority, company status, incorporation date, assets, liabilities, equity, employees, and available P&L figures. Return concise summaries first, then structured tables where helpful.
For workflow and automation builders who want an AI agent step that can look up companies. Shown for n8n. Other tools with an MCP client step need the same three details.
- Add the tool. In an AI Agent workflow, add the MCP Client Tool node as one of the agent's tools.
- Set the endpoint:
https://api.dataledger.uk/mcp - Set the transport: HTTP Streamable.
- Set authentication: choose Header Auth, with the header name
x-api-keyand your key as the value. - Tell the agent when to use it. In the agent's instructions, say it should use DataLedger for UK company, ownership and financial questions.
If you are building your own agent with an SDK or framework that supports remote MCP servers, point it at the DataLedger endpoint and send your key in the x-api-key header. Those are the same three details as every other route.
- Prototype with MCP, scale with the APIIf your agent calls the data repeatedly, the API gives you more control over cost and speed.
- Self-host if you need controlThe server is open on GitHub, so you can adjust the system prompt and control exactly what it requests. Node.js is required.
- API v2 documentationEndpoints, parameters and credit costs.
- dl-mcp on GitHubThe MCP server source, for self-hosting.
MCP or API
Same data. Pick how you want to use it.
You can also use the DataLedger API directly. MCP is the easy way to bring the same data into an AI tool you already use, with nothing to build.
Connect once, then just ask
- Plug into Claude and other AI tools in minutes
- Your AI decides which lookups to run
- No code. Ideal for exploring, research and prototypes
Build it into your own product
- Search by name or criteria, then fetch the full profile with owners, ultimate parent and subsidiaries
- Full control of cost and speed for repeated calls
- API v2 documentation or the API overview
Prefer a spreadsheet to ask questions of? See custom datasets.
Cost and limits
Know what a question costs, and what the data covers.
You pay in credits and see where they go. Start free, then buy a one-off pack when you need more. There is no subscription to manage, and packs never expire.
Empty results are not charged. 25 free credits on sign up, no card needed.
Prefer a plan? Monthly plans start from £10 and credits roll over. See full pricing
- Financials from electronically filed accountsOver 3 million of the over 5 million companies on record. Turnover is only available where a company reports it.
- Ownership as registeredWhere the register ends at a trust or an overseas entity, the chain stops there and the record says so.
- Owner age is approximateBased on the year of birth Companies House publishes. No R&D data.
- Built only from open dataEvery field comes from public records. The server is read-only.
Companies House MCP servers compared
Look up one company, or find the ones you do not know yet.
There are several MCP servers for Companies House. Most are open-source wrappers around the free Companies House API. They suit checking a company you already know. DataLedger is built for finding companies by their financials and owners.
Wrappers vary, so check the one you are considering. DataLedger is built on the same open Companies House data and adds the structuring on top.
Questions
Common questions about DataLedger and AI
What is an AI agent, and what is MCP?
An AI agent is an AI tool that can take steps for you, such as searching, reading records and comparing results. MCP (Model Context Protocol) is an open standard that lets an agent use outside tools and data, like adding a new skill. Connecting DataLedger over MCP lets your agent look up UK company records instead of guessing from memory or web pages.
Do I need to be technical to use DataLedger with AI?
No. Connecting DataLedger is a one-off task that takes a few minutes: you paste a web address and your API key into your AI tool. After that you type questions in plain English, and your agent searches the records, reads the results and answers. It can also put them in a table for Excel.
How does DataLedger make my AI more accurate?
On its own, an AI answers about UK companies from memory and web pages, which can be out of date or wrong for ownership and accounts. With the DataLedger MCP server connected it looks up the filed Companies House records, so each answer about owners, group structures and financials traces back to a record you can check.
Which AI agents work with the DataLedger MCP server?
The DataLedger MCP server has been tested with Claude. It works with Claude on the web and desktop, Claude Code, Cursor, VS Code, Gemini CLI, n8n and other tools that support remote MCP servers with an API key header. ChatGPT connects through a custom GPT using the DataLedger API.
Can my AI see a UK company's ultimate owner and subsidiaries?
Yes. With the DataLedger MCP server your agent can follow the registered ownership chain up through parent companies to the ultimate owner, and list the subsidiaries beneath a company. Where the register ends at a trust or an overseas entity, the chain stops there and the record says so.
Should I use the DataLedger MCP server or the API?
Use the MCP server to connect an AI tool you already use, with nothing to build. Use the API when you are building a product or workflow that calls the data repeatedly. Both use the same DataLedger data and the same credits, and many people prototype with MCP and then build with the API.
Is there a Companies House MCP server?
Yes, there are several. Many are open-source wrappers around the free Companies House API, which are good for looking up a company you already know. The DataLedger MCP server adds structured financials, growth ratios, owner type and age, and ownership chains, so an AI agent can find companies by those criteria, not just look one up. It is hosted, so there is nothing to install.
How do I connect Claude to Companies House data?
Get a free DataLedger API key, then in Claude go to Customize, then Connectors, and add a custom connector with the server URL https://api.dataledger.uk/mcp. Set authentication to No sign-in and add your key as an x-api-key request header. If your account does not show request headers yet, use the Claude Desktop config file instead.
Can ChatGPT access Companies House data?
Yes, through a custom GPT with Actions that calls the DataLedger API with your API key. ChatGPT's own MCP connectors currently expect an OAuth sign-in, which the DataLedger MCP server does not use.
Why not just let my AI search the web for UK company data?
Web search returns pages, not structured records, so it cannot reliably retrieve filed financials, registered owners, owner ages or group structures. The DataLedger MCP server gives your agent structured Companies House data, so each answer traces back to a filed record.
How much does the DataLedger MCP server cost?
DataLedger gives you 25 free credits when you sign up, with no card needed. A company search costs 1 credit and a company lookup costs 1 credit, with extra credits for financials, owners, parents and children. After that you can buy one-off credit packs from £5 that never expire, or take a plan from £10 per month.
Is the DataLedger MCP server read-only?
Yes. The DataLedger MCP server only retrieves data. Nothing is written, changed or deleted, and all usage is tracked in credits so there are no surprises.
What are the limits of DataLedger's data?
Financials come from electronically filed accounts, which covers over 3 million of the over 5 million UK companies on record. Ownership is as registered at Companies House, owner age is approximate, and there is no R&D data. Turnover is only available where a company reports it, so many searches use net assets, growth, debt and cash instead.
Ready to connect your AI to UK company data?
Start with 25 free credits. Search live financials, ownership records and corporate structures from any MCP-compatible AI tool.
Get your free API keyView API docsRelated: Ownership intelligence What is PSC data? AI agents and LLM workflows Datasets
Last reviewed 2 November 2026. Questions? contact@dataledger.uk