Skip to main content

Documentation Index

Fetch the complete documentation index at: https://nango.dev/docs/llms.txt

Use this file to discover all available pages before exploring further.

AI agents become useful when they can act in the systems your users already use: HubSpot, Gmail, Slack, Linear, Notion, Salesforce, and the rest. The hard part is not just calling an API. It is OAuth, token refresh, per-user permissions, rate limits, logging, and keeping credentials out of the model loop. Nango handles that plumbing. Users authorize integrations through Nango, you expose selected action functions as tools, and agents call those action functions through the Nango API or the built-in MCP server.

​
When this fits

Use Nango for AI tool calling when:
  • Your agent needs to fetch or change data in external APIs.
  • Tool calls must run with each user’s own credentials.
  • You want logs, retries, rate-limit handling, and a deployable runtime.
  • You need template action functions now, with room for custom action functions later.
  • You want one MCP server URL for Nango-hosted tools.
For RAG-style workflows where data should be replicated before the agent needs it, use Sync external API data.

​
How Nango fits in

  1. Your user authorizes an integration with Nango.
  2. You enable template action functions or deploy custom action functions for allowed operations.
  3. Your agent lists or receives those action functions as tools.
  4. The agent asks Nango to execute an action function for the right connection.
  5. Nango runs the external API call with the user’s credentials and logs the execution.
This keeps provider credentials in Nango and gives you a stable tool surface across frameworks.

​
Technical setup

Nango supports both direct action function execution and MCP-based tool discovery. Use the technical Tool calling & MCP guide for headers, transports, tool discovery, action execution, and client setup.