Put Real Ownership Data
Inside Your AI
One endpoint connects Claude, ChatGPT, Cursor or any MCP client to government-sourced registry data. Ask who owns a company, get the chain back with the registry it came from.
Ask in plain language. Get the filing.
The hard part is not finding a company. It is knowing which filing a figure came from, and when it was filed. That is what comes back.
Every figure traces back to the registry that issued it. No blended scores, no third-party estimates, nothing your compliance team has to defend to an auditor.
What your agent can call
Grouped by the job it does. Each tool returns structured data plus the source registry and retrieval date.
Find and verify the entity
Resolve a name, number or website to one legal entity before anything else happens.
search_companiesName, registration number, VAT or jurisdictionget_companyStatus, legal form, incorporation, registered addressverify_vatVAT number validation against the issuing authoritymatch_by_websiteDomain to registered entity
Trace ownership and control
The core of the product. Layered shareholding resolved across jurisdictions.
get_shareholdersDirect holders, percentages, share classesget_ownership_chainFull upward chain with each intermediate entityget_group_structureParents, subsidiaries, sister entities
People behind the entity
Officers and their other appointments, for conflict and network checks.
get_officersDirectors and executives with appointment datesget_officer_appointmentsEvery company one person is attached to
Watch it change
Ownership is a moving target. Register interest once and get told when it moves.
watch_companyMonitor an entity for filings and ownership changesget_changesWhat changed, when, and in which filingget_data_coverageWhat is available in a given country before you ask
Connected in about a minute
No SDK, no server to run, no engineering ticket. Paste a URL and your existing key.
-
Get a key
Choose a plan, or use the key from the Zavia.ai plan you already have. Same key, same quota as the API.
-
Add the server
Paste the config into your client. Every MCP client accepts the same shape — only the file location differs.
-
Ask a question
Try: "Trace the ownership of Detected Ltd up to the ultimate beneficial owner at a 10% threshold."
// claude_desktop_config.json { "mcpServers": { "zavia": { "url": "https://mcp.zavia.ai/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }
// .cursor/mcp.json { "mcpServers": { "zavia": { "url": "https://mcp.zavia.ai/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } }
# Streamable HTTP · list the tools curl -X POST https://mcp.zavia.ai/mcp \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'
Transport: Streamable HTTP. Auth: bearer token in the Authorization header. Calls draw on your plan quota — nothing separate to buy.
Your AI already guesses at ownership
Ask any model who owns a mid-market European company and it will produce something confident and unsourced. Compliance cannot sign that.
Model without registry access
- Answers from training data of unknown vintage
- No source to cite in a file review
- Silently misses layers held offshore
- Cannot tell you what it does not know
Model with Zavia.ai connected
- Data pulled live from the issuing registry
- Every field carries source and retrieval date
- Chains resolved across jurisdictions, not one filing
- Coverage gaps returned as gaps, not filled in
Built for the teams that have to prove it
Onboarding without the manual chase
Resolve UBOs at your own threshold during onboarding instead of emailing the client for a structure chart and waiting a week.
Due diligence that shows its working
Every answer arrives with the registry behind it, so the file stands up when someone asks where the number came from.
See the structure before the data room
Map parents, subsidiaries and minority holders before you commit deal team hours to a target.
Know when control moves
Register a watchlist through the agent and get told when a shareholding or director changes at source.
Before you connect
What is MCP, in one sentence?
Model Context Protocol is an open standard for connecting an AI assistant to a live data source. Instead of pasting registry extracts into a chat, the assistant queries Zavia.ai directly and cites what it finds.
How is this different from your REST API?
Same data, same key, different consumer. The REST API is for your engineers to build against. The MCP server is for an AI assistant to call on its own, so a compliance analyst can ask a question in plain language without anyone writing code.
Where does the data come from?
Directly from official government registries in over 100 countries. Every response carries the source registry and the date it was retrieved.
Is the data live, or cached?
Company records, officers and shareholdings are retrieved from the registry at the time of the call. Where a registry publishes on a filing cycle rather than continuously, the response shows the date of the most recent filing so you can see how current it is.
Can I set my own UBO threshold?
Yes. Pass the threshold with the call — 25%, 10%, or whatever your policy requires. Holders below the threshold are returned as well and marked as such, so nothing is hidden from the reviewer.
What happens when a jurisdiction does not publish ownership data?
The tool returns the gap as a gap. It names the jurisdiction, states what is unavailable there, and stops. Nothing is filled in with an estimate or a modelled figure.
Can our compliance team audit what the agent retrieved?
Every tool call and its response is attributable — the tool invoked, the parameters passed, the registry that answered and the retrieval timestamp. That is the record you keep on file, not a chat summary.
Which clients work with it?
Any MCP-compatible client. That includes Claude, ChatGPT, Cursor, VS Code Copilot and Windsurf, as well as agent frameworks that speak Streamable HTTP.
Does this cost extra on top of my plan?
No. The MCP server is included on every plan and uses the same API key and quota as the REST API. Rate limits follow your tier. Plans start at $99/month.
Is our query data used to train models?
No. Queries are processed to serve the request and are not used to train models. Data is transmitted over TLS and hosted in ISO-certified data centres, in line with GDPR.
Ask your first ownership question today
Plans start at $99/month and include MCP access at no extra cost. Connect in about a minute and see the chain for yourself.