Building custom integrations for Nango is much faster when using AI tools. This guide shows you how to leverage AI tooling to build high-quality integrations quickly.
Install the Nango function builder skill
Install Nango’s skill pack into your development environment:
npx skills add NangoHQ/skills
This adds nango-function-builder, which includes guidance for building both Actions and Syncs.
If your coding agent is already running, restart it after installing the skill so it can load the new instructions.
Use it with your coding agent
The nango-function-builder skill is loaded automatically when using popular coding agents/tools (Claude Code, Cursor, Gemini CLI, Codex CLI, VS Code with Copilot, Opencode, and more).
Tell your agent what you want to build. You can start with a simple prompt to build an Action:
I want to build a Nango action that will get specified Slack channel information.
However, the more information you give the agent up-front, the better the initial version of the Action is going to be:
I want to build a Nango action that will get Slack channel information.
Integration ID: slack
Connection ID: my-slack-connection
Inputs: channel_id
Outputs: id, name, is_private
API Reference: https://api.slack.com/methods/conversations.info
Similarly, a prompt to build a Sync could look like this:
I want to build a Nango sync that will store all Figma projects.
Integration ID: figma
Connection ID: my-figma-connection
Frequency: every hour
Inputs: team_id
Outputs: Project model with fields id, name, last_modified
API Reference: https://www.figma.com/developers/api#projects-endpoints
What Gets Generated
Once you provide your input, your agent typically:
- Generates the function code - Creates a complete action or sync implementation
- Creates a full test suite - Writes comprehensive tests
- Validates with dry run - Runs
nango dryrun to verify against your connection
- Iterates on failures - Automatically debugs and fixes issues
You’ll end up with a fully tested, working function ready to deploy.
The more detail you provide, the better the generated function will be. Include API documentation links when available.