LangChain OpenRegistry
sophymarine/langchain-openregistry
Acesso em tempo real a 27 registros nacionais de empresas via agentes LangChain
Ver no GitHubToolkit LangChain pré-configurado para acessar em tempo real 27 registros nacionais de empresas (Companies House, RNE França, Handelsregister Alemanha, InfoCamere Itália, BORME Espanha, OpenDART Coreia e mais). Permite que agentes de IA consultem dados governamentais atualizados sem cache, preservando todos os campos originais da resposta.
Instalação
npx convez-ai sophymarine/langchain-openregistry1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Verificação de dados corporativos e rastreamento de cadeias de propriedade entre jurisdições
- Due diligence automatizada e pesquisa de compliance com dados oficiais
- Integração de dados de registro público em pipelines de agentes IA
Veredito

Solução pronta para produção que democratiza acesso a dados governamentais fragmentados através de uma interface unificada. Útil para automação legal, financeira e de compliance, mas requer familiaridade com LangChain e agentes assíncronos.
README
A tradução para o português deste README ainda está na fila.
langchain-openregistry
LangChain toolkit preconfigured for the OpenRegistry MCP server — live, real-time access to 27 national company registries (UK Companies House, France RNE, Germany Handelsregister, Italy InfoCamere via EU BRIS, Spain BORME, Korea OpenDART, plus 21 more).
Every tool call is a real-time query against the upstream government API at the moment the agent asks. No data is cached or aggregated, so the agent always sees the registry's own response with all field names preserved.
Install
pip install langchain-openregistry
Usage
Anonymous use — no signup, no API key beyond the LLM provider's own:
from langchain_openregistry import OpenRegistryToolkit
from langchain.agents import create_agent
from langchain_openai import ChatOpenAI
toolkit = OpenRegistryToolkit()
tools = await toolkit.aget_tools()
agent = create_agent(ChatOpenAI(model="gpt-4.1"), tools)
result = await agent.ainvoke({
"messages": [
{
"role": "user",
"content": (
"Find Tesco PLC on Companies House and walk its corporate"
" ownership chain across jurisdictions."
),
}
]
})
print(result["messages"][-1].content)
In a synchronous context, toolkit.get_tools() is a thin asyncio.run() wrapper around aget_tools().
Authenticated tier
The free anonymous tier allows 20 calls/min per IP and a cross-border fan-out of 3 countries / 60s. For higher limits, complete the OAuth 2.1 flow at openregistry.sophymarine.com/account and pass the resulting bearer token:
toolkit = OpenRegistryToolkit(oauth_token=OPENREGISTRY_TOKEN)
Restricting the tool surface
OpenRegistry exposes ~30 tools. To keep the agent's prompt lean, allowlist just the ones you need:
toolkit = OpenRegistryToolkit(
allowed_tools=[
"search_companies",
"get_company_profile",
"get_officers",
"get_persons_with_significant_control",
],
)
What's under the hood
This package is a thin convenience wrapper around langchain-mcp-adapters' MultiServerMCPClient with the OpenRegistry endpoint baked in. If you need MCP prompts, resources, or fine-grained session control, reach into the underlying client:
toolkit = OpenRegistryToolkit()
client = toolkit.client # MultiServerMCPClient
async with client.session("openregistry") as session:
...
License
MIT — see LICENSE.
OpenRegistry is a platform by Sophymarine. The hosted service is not affiliated with any of the national company registries it proxies.