Academy
CLI and MCP

Connect AI assistants via MCP

MCP (Model Context Protocol) is the open standard that connects AI assistants to external systems. With Dattos's MCP server, your assistant — Claude, ChatGPT, Cursor, and others — can now operate the platform in natural language: create and run reconciliation flows, query results, generate reports, and manage records, always with the permissions of the logged-in user.

What is available to the assistant

The MCP server exposes about 70 tools covering the full platform lifecycle:

  • Flow building: create analyses, add data sources, and assemble pipelines with the support of Dattos AI.

  • Execution: trigger and resume runs by reference date and monitor status.

  • Query: loads, data sets, reconciliation results, and summaries.

  • Export: generate and download reports.

  • Administration: users, roles, connectors, and folders.

This is not a read-only surface — the assistant performs real actions. That is why everything runs with the user's own API key: the assistant sees and does exactly what that user can, nothing more.

Current scope: the tools cover Process Automation and platform administration. The Account Reconciliation and Task Management modules are not yet available via MCP — coverage is planned as the product evolves.

Path 1 — Install in your assistant (local)

With the Dattos CLI installed and authenticated (see "Dattos CLI: installation and first steps"), a single command configures the MCP server in the tool:

dattos mcp install <ferramenta>

Supported tools: claude-code, claude-desktop, cursor, vscode, windsurf, gemini, chatgpt, and copilot-studio. The command writes the dattos server entry to the tool's configuration file (for cloud options, it shows the connection instructions). Without an argument, dattos mcp install lists the available targets.

After installing, restart the assistant and ask, for example: "list my reconciliation analyses in Dattos and run the sales one for 30/06".

Path 2 — Hosted connector (for cloud assistants)

For assistants that run in the cloud (Claude.ai connectors, ChatGPT, n8n), Dattos provides a hosted MCP gateway, with no local installation: the tool connects over HTTP by providing the gateway URL and two headers — the user's API key (Authorization: Bearer api-...) and your instance address (X-Dattos-Api-Url).

Enablement: making the hosted connector available for your environment is handled together with the Dattos team — contact your account manager or support.

Security

  • User permissions, always: the credential used is the personal API key — privacy roles, folder scope, and permissions apply fully to the assistant.

  • No credential custody: in the hosted gateway, the key travels in each request and is not stored — revoking the key in your profile cuts off the assistant's access.

  • Restricted sensitive operations: privileged actions and operations that depend on local files are unavailable in hosted mode.

  • Best practices: treat the API key like a password; prefer a user with minimum permissions for automations; revoke and regenerate the key periodically in My Profile > API Key.

  • Dattos CLI: installation and first steps

  • Dattos CLI: commands and capabilities

  • API keys: generate and revoke

  • Introduction to integrations

Was this helpful?