Skip to main content
Connectors are how your production system gets into Pavo. Each one gives Pavo read-only, scoped access to a source - a warehouse, a code repository, a dashboard, an experimentation platform - which it indexes into typed, cited facts. Connect the source closest to your system first; you can add the rest over time.

​
What you can connect

Warehouses

Databricks · BigQuery · Snowflake · ClickHouse: tables, schemas, and query history.

Semantic layer

dbt: metric definitions, model structure, and lineage from your semantic layer.

Code

GitHub · Bitbucket: how the system is actually built.

Dashboards & BI

Tableau · Power BI · Superset · QuickSight · Redash: the plots leaders watch and the queries behind them.

E-commerce

Shopify: catalog, collections, discounts, and monthly sales.

Experimentation

Eppo · Statsig · Intelligems: experiments, metrics, feature flags, and on-site tests.

Docs & knowledge

Confluence · Google Docs: the written knowledge you already have.

Gateways

MintMCP: the tools your MCP gateway already exposes, read live under one agent identity.

​
What each source contributes

Different sources answer different questions about your system. Pavo reads all of them into the same fact index, with one exception: a gateway is not indexed, Pavo queries it live when a task or book needs it. Either way, it is worth knowing what each one is best at.

​
Access and freshness

  • Read-only and scoped. Every connector is granted read access, limited to the catalogs, repositories, spaces, or projects you choose. What you do not grant, Pavo does not see.
  • Your keys stay in your environment. Credentials live inside your Pavo instance and are not shared out. See Security.
  • Re-scan to stay current. Sources change. Re-scanning refreshes the index and picks up what is new. See Coverage and freshness.

​
When you can’t grant a live connection

Some environments cannot open a direct connection to a source, for security or policy reasons. Two options keep you moving:
  • Upload instead of connect. Provide anonymized files or an export rather than a live credential. Pavo indexes uploads the same way it indexes a connection.
  • Scope at the task level. When you do not want an entire system exposed for one piece of work, provide just the specific files a task needs, rather than a standing connection.
Not every source has to be integrated on day one. Teams often start with a warehouse and uploads, then add live connectors as trust builds.

​
Setting up a connector

Each connector page follows the same shape: what access it needs, the step-by-step setup, the connection details to share, and a short list of questions to help scope what gets indexed. Pick your source to begin:

Databricks

Grant read access to your Databricks workspace.

Snowflake

Create a read-only role and warehouse for Pavo.

GitHub

Connect your organization with a fine-grained token.

Statsig

Share a read-only Console key for experiments and metrics.

MintMCP

Create one agent identity on your MCP gateway for Pavo.