Data Sources
Connect your databases, warehouses, storage, and SaaS apps to MarcoPolo.
MarcoPolo connects to your data wherever it lives. Once connected, your AI can query, browse, correlate, and analyze across any source in a single conversation.
Each connected system becomes a connection in your workspace at connections/<name>/, with its own README, RULES, SYNTAX, queries, and metadata snapshots. See Workspace for the layout.
Supported sources
| Category | Platforms |
|---|---|
| Databases | PostgreSQL, MySQL, Microsoft SQL Server, ClickHouse, MongoDB, Microsoft Fabric |
| Cloud Warehouses | Snowflake, BigQuery, Redshift, Athena |
| Object & File Storage | AWS S3, Azure Blob, Google Drive, OneDrive, SharePoint, SFTP |
| SaaS Apps | Salesforce, HubSpot, Jira, Google Sheets (through Google Drive), ClickUp, Slack, GitHub, NetSuite, QuickBooks |
| Observability | Grafana Loki, Grafana Tempo, Grafana Mimir, Prometheus, Amazon CloudWatch, Sentry |
50+ connectors and growing. Need something not listed? Contact [email protected].
Try it without credentials
Demo Connections — seven hosted, no-credentials connections (Snowflake, Salesforce, S3, BigQuery, MongoDB Atlas, SFTP, large-Parquet) — let you run the use-case walkthroughs as written before bringing your own data. Install one with:
You: "Install the Snowflake demo."Adding a connection
A connection in MarcoPolo is separate from your account. Your account establishes who you are; connections give your AI access to specific systems.
Add and manage connections at mcp.marcopolo.dev/app/connections. The flow depends on the source — OAuth for most SaaS apps, username/password for databases, key pairs or service account keys for warehouses, access keys or service principals for storage. The setup UI guides you through the right method. Credentials are stored encrypted and never exposed to the AI.
Network requirements
MarcoPolo connects to your data sources over the public internet.
IP allowlisting. If your source requires a firewall allowlist (database firewalls, MongoDB Atlas, etc.), add MarcoPolo's NAT gateway:
34.208.3.240/32SSH tunnels. PostgreSQL, MySQL, ClickHouse, MongoDB, Redshift, and CockroachDB connections can reach a private database through an SSH bastion host. MarcoPolo generates a key pair for the connection; you install the public key on the bastion.
Local databases. MarcoPolo can't connect to databases running only on your laptop. Expose them via a tool like ngrok or another secure tunnel.
Private networks. For VPC peering, PrivateLink, or BYOC deployments, contact [email protected].
Sharing with your team
- Open Connections in the web UI.
- Select a connection and click Share.
- Share with specific people or your entire company (same email domain).
Shared connections appear automatically in teammates' workspaces. Teammates query through the connection without ever seeing its credentials. See Workspace Isolation for details.
Uploading files
You can upload CSV, Excel, and other files directly through the web UI or by asking your AI. They land in data/uploads/ inside your workspace where your AI can read, parse, and analyze them alongside connected sources.
Detailed guides
Each source type has different connection patterns and best practices:
- Demo Connections: Hosted, no-credentials demos for evaluation and walkthroughs.
- Databases: PostgreSQL, MySQL, Microsoft SQL Server, ClickHouse, MongoDB, Microsoft Fabric.
- Cloud Warehouses: Snowflake, BigQuery, Redshift, Athena.
- Object Storage: S3, Azure Blob, Google Drive, OneDrive, SharePoint, SFTP.
- SaaS Applications: Salesforce, HubSpot, Jira, Google Sheets, ClickUp.