Read-only mirror. This repo is automatically published from an internal repository. Do not open pull requests here — all development and review happens internally.
The official repository of Fivetran skills and plugins for AI Agents, as a Claude Plugin Marketplace.
Register the marketplace:
/plugin marketplace add fivetran/skills
Install a specific plugin from the marketplace (see Plugins below):
/plugin install <plugin>@fivetran-skills # replace <plugin> with a named from the list plugin
Reload your plugins to load the installed skill in your current session:
/reload-plugins
- Click Customize in the left nav and click the + next to Personal Plugins
- Click + Create Plugin → Add Marketplace
- Enter
fivetran/skills - Navigate to + Browser Plugins → Personal → skills
- Add the specific skill you need by clicking the + next to the skill
Skills can also be installed individually by referencing them in the skills/ subdirectory.
npx skills add fivetran/skills
See Vercel's Skills docs for flags like --global, --skill, --agent, and --list.
| Plugin | Description |
|---|---|
| base | Fivetran MCP and general skills |
| ad-performance-analysis | Cross-channel ad performance analysis via BigQuery, Snowflake, or Databricks |
| customer-support-analysis | First reply, resolution time, backlog, SLA, and CSAT analysis for Zendesk |
| marketing-automation-analysis | Funnel velocity, nurture, and email engagement analysis for Marketo (and other marketing automation tools) |
| sales-pipeline-analysis | HubSpot sales pipeline funnel and rep performance analysis |
| store-performance-analysis | E-commerce store performance analysis from raw Shopify connector data |
| Skill | Description |
|---|---|
| fivetran-account-info | Get a quick overview of the connected Fivetran account |
| lakehouse-explorer | Query Fivetran-managed Iceberg tables via DuckDB and Polaris, cost-efficiently |
The bundled Fivetran MCP server uses the published uvx launcher from
fivetran/fivetran-mcp:
uvx --from git+https://github.com/fivetran/fivetran-mcp fivetran-mcpWhen the base plugin is enabled, Claude Code prompts for the Fivetran
API key and API secret via userConfig and injects them into the MCP server
configuration automatically.
These skills are provided as-is. Fivetran makes no guarantees about their fitness for any particular purpose and accepts no liability for issues arising from their use.
When a skill is invoked (whether it succeeds or fails), the plugin sends a usage event to Fivetran containing: an anonymous per-machine identifier, your Fivetran account and user ID (once authenticated), the skill name, invocation status (success or failure), the model in use, and a session identifier. The receiving server also records standard request metadata, including your IP address, as part of normal HTTP logging.
Users can learn more about these data practices in Fivetran's Privacy Policy.
Set FIVETRAN_SKILLS_NO_TELEMETRY=1 in the environment your agent runs in.
Values of 0, false, FALSE, no, or NO keep telemetry on; any other value disables it.