Your AI agent can browse the web, write code, and manage calendars. Can it answer "is this company real?" before it processes a payment or sends sensitive data?
Counterparty verification is standard practice in financial services. Confirming that the entity you're dealing with is registered, active, and not sanctioned. Every bank does it. Every payment processor does it. AI agents skip it entirely, because the data is scattered across government registries, sanctions lists, and commercial databases that weren't built for programmatic access.
Here's how to fix that in about 5 minutes.
The problem in practice
Say your agent is processing an invoice from a UK supplier. Before authorizing payment, you want to know four things: Is the company actually registered at Companies House? Is it active or dissolved? Who are the directors and persons with significant control? Are any of them on sanctions lists?
Without a verification layer, your agent either skips these checks (risky) or you build custom integrations to Companies House, OFAC, EU sanctions lists, and UN consolidated lists. That's weeks of work and ongoing maintenance for each jurisdiction you add.
MCP: zero code
If your agent supports MCP (Claude Desktop, Cursor, Windsurf, or any MCP-compatible client), connect Strale's MCP server:
{
"mcpServers": {
"strale": {
"type": "streamableHttp",
"url": "https://api.strale.io/mcp"
}
}
}
Your agent now has 271 tools available. Ask it to verify a company and it finds and calls the right capability:
"Verify that Acme Trading Ltd is a real, active UK company and check the directors against sanctions lists."
The agent calls kyb-essentials-uk, gets back registration status, officers, beneficial ownership, and sanctions screening results. Structured JSON with provenance metadata and a quality score.
LangChain / CrewAI: three lines
from langchain_strale import StraleToolkit
toolkit = StraleToolkit(api_key="sk_live_...")
tools = toolkit.get_tools()
That gives your agent access to the full capability set. The toolkit handles discovery, input formatting, and response parsing.
For CrewAI:
from crewai_strale import StraleToolkit
toolkit = StraleToolkit(api_key="sk_live_...")
tools = toolkit.get_tools()
agent = Agent(
role="Compliance Analyst",
tools=tools,
goal="Verify counterparties before transactions"
)
Also available for Semantic Kernel, OpenAI Agents SDK, Google ADK, and Pydantic AI.
Direct API call
curl -X POST https://api.strale.io/v1/execute/kyb-essentials-uk \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{"company_name": "Acme Trading Ltd"}'
Response:
{
"result": {
"company_name": "ACME TRADING LTD",
"company_number": "12345678",
"status": "active",
"incorporated": "2019-03-15",
"registered_address": "10 Downing Street, London",
"officers": [...],
"persons_with_significant_control": [...],
"sanctions_check": {
"is_sanctioned": false,
"lists_checked": ["OFAC SDN", "EU Consolidated", "UN Security Council"]
}
},
"meta": {
"quality_score": { "grade": "A", "score": 92 },
"provenance": {
"source": "companies-house.gov.uk",
"fetched_at": "2026-04-04T10:23:15Z"
},
"transaction_id": "txn_abc123",
"verify_url": "https://api.strale.io/v1/verify/txn_abc123"
}
}
The meta block gives your agent and your compliance team everything needed for an audit trail: where the data came from, when, how reliable it is, and a public URL to verify the result independently.
What's available beyond the UK
271 capabilities across 27 countries. KYB bundles for UK, Norway, Sweden, and Australia that combine company data, sanctions, and beneficial ownership in one call. IBAN validation across 80+ countries. EU VAT validation via VIES. Sanctions screening against OFAC, EU, and UN lists. Domain reputation and WHOIS lookups for verifying a counterparty's web presence. Beneficial ownership via UK Companies House PSC register, with Nordic registries next.
Honest limitation: beneficial ownership is UK-only right now. If your counterparty is registered in Singapore or Delaware, you'll get company data and sanctions screening but not the ownership chain. I'm working on expanding that.
All priced per call, €0.02 to €2.50. No monthly minimums. Charged only on success.
How the scoring works
Every capability is tested by 1,500+ automated test suites. The results are published as a dual-profile quality score. One profile measures the capability itself (correctness, schema compliance, error handling). The other measures the reliability of its upstream data source. A sanctions check scoring 95 is more useful to your agent than one scoring 60, and the score tells you which is which before you pay for the call.
Methodology: strale.dev/trust
We're based in Sweden and bootstrapped. If you're building agents that need to verify counterparties, suppliers, or partners, we'd like to hear what data you need that isn't available yet.
strale.dev. Free capabilities available, €2 trial credit, no card required.
Top comments (0)