Fivetran Skills Private Preview

Fivetran Skills are open-source analytical packages that help AI agents turn data in your destinations into actionable insights. Each skill provides AI agents with the domain-specific context and instructions they need to analyze data in your destination. You can ask questions about your data, identify trends and anomalies, and generate reports without manually identifying relevant tables, writing SQL queries, or understanding your destination schema. With a compatible AI agent, Fivetran Skills identify the appropriate data, generate and run queries against a supported destination, and return the results with relevant context.

How skills work

Each skill is distributed as a plugin containing the instructions, metadata, query patterns, and setup information an AI agent needs to analyze a specific data domain. Each plugin includes the following:

• A SKILL.md file with instructions that the AI agent loads when the skill is invoked
• Metadata that describes the skill and its capabilities
• Query patterns and guidance for answering common business questions
• Supporting scripts for setup, readiness checks, and local configuration

When setting up the skill initially, it uses the Fivetran API key to discover your destinations and connections. Once the setup is complete, it queries the data in your destination using the destination credentials you configured.

All skills follow the same workflow to prepare and analyze your data:

1. Verify prerequisites: The skill performs a readiness check to verify that the required connectors, Quickstart data models, destination tables, and destination CLI are available. The skill stores the readiness results in your local environment, so it skips the check when you invoke the same skill again.
2. Prepare the analysis: The skill identifies destinations that are active, selects the relevant schemas and tables, and confirms that the destination contains updated data.
3. Analyze your data: When you ask a question, the skill determines the required metrics, dimensions, filters, and time period, and then runs the required SQL query against your destination. It can also perform further analysis, such as detailed breakdowns, anomaly detection, and comparisons with previous time frames.
4. Return the results: The skill returns the analysis in the form of a Markdown table with a factual summary, anomaly callouts, and suggested follow-up questions. It can also generate a self-contained HTML report with charts that opens locally in your browser.

How skills access your data

Skills use your destination's CLI to query data in your destination. They use the credentials you configure for your destination and guide you through installing the CLI if it is not already installed. They do not have separate credentials or independent access to your destination or data source. Skills do not query your data source directly. They only access and analyze data that is already synced into your destination.

How skills work with Quickstart data models

Fivetran Skills and Quickstart data models serve different purposes but work together to support AI-powered data analysis. Quickstart data models transform raw data from Fivetran connections into analytics-ready tables, while Fivetran Skills provide an analytical layer that helps AI agents understand and query that data. Together, they provide AI agents with standardized, analytics-ready data and the domain knowledge needed to answer business questions.

Setup guide

Follow our step-by-step Fivetran Skills setup guide to configure a skill and analyze data in your destination.

Supported AI agents

Fivetran Skills are compatible with the following AI agents:

• Claude Code
• Claude Cowork
• Codex
• Gemini CLI

Supported destinations

Fivetran Skills can query data from the following destinations:

• BigQuery
• Databricks
• Snowflake

Skills catalog

Fivetran offers the following skills, organized by category to help you find the skill most suitable for your business requirements: