Skip to content
OpeniDevelopers

AI and MCP

Openi inside your agent. Through MCP or the API.

Bank data, tax invoices and reconciliation as tools for Claude, ChatGPT, Cursor and any MCP client.

Request accessDocumentation
Reconcile May's transactions from the Itaú account with the invoices issued.
  • GET /v1/items/{itemId}/accounts/{accountId}/transactions
  • POST /v1/rule-sets/{ruleSetId}/run
  • GET /v1/matches

Connect from any MCP client

  • Claude
  • ChatGPT
  • Cursor
  • Gemini
  • GitHub Copilot
  • Windsurf
  • n8n
  • LangChain

Open Finance via MCP. Accounts, transactions and investments in the conversation.

The agent creates the connect session, waits for the user's consent and reads the bank data already normalized.

Explore Open Finance
Connect the client's Itaú account and list May's transactions.
  • POST /v1/connect-sessions
  • GET /v1/items/{itemId}/accounts/{accountId}/transactions

Tax invoices via MCP and API. Issue and look up, by agent or by code.

NF-e, NFS-e and NFCom with the same operations on both paths: ask the agent or call the endpoint straight from your backend.

Explore Docs
Issue the NF-e for order 10472 and tell me how many invoices went out this month.
  • POST /v1/invoices/product/sent
  • GET /v1/invoices/product/sent/:invoiceId
  • GET /v1/invoices/product/sent/list

Reconciliation via MCP. The agent runs the rule set and explains the result.

It triggers the reconciliation between two datasets, reads the pairs with their decision and score, and sets aside what needs review.

Explore Nexus
Reconcile May's transactions from the Itaú account with the invoices issued.
  • GET /v1/items/{itemId}/accounts/{accountId}/transactions
  • POST /v1/rule-sets/{ruleSetId}/run
  • GET /v1/matches

Connect in minutes. One configuration, three products.

  1. Request access

    You get the API Key and the MCP server address for the products you subscribed to.

  2. Add the server

    Paste the configuration into your client: Claude, ChatGPT, Cursor or any other that speaks MCP.

  3. Ask in plain English

    The agent reads the documentation, picks the operation and calls the same API your code uses.

Request accessllms.txt
{
  "mcpServers": {
    "openi": {
      "url": "<MCP server URL>",
      "headers": { "Authorization": "Bearer <your API Key>" }
    }
  }
}