Demo Connections
Hosted, no-credentials connections for trying MarcoPolo with realistic datasets.
Demo connections let you evaluate MarcoPolo end-to-end without bringing your own data — no OAuth flow, no credentials, no network setup. They install through the install_demo_connection MCP tool, become first-class connections in your workspace, and behave the same way real ones do: connection list, connection query, connection browse, dashboards, schedules, joins through DuckDB.
They're also the fastest way to reproduce a workflow against a known dataset — a use-case walkthrough you can paste into any client and run as written.
How install works
You: "Install the Snowflake demo."AI calls install_demo_connection(demo_connection="snowflake-demo")
→ success: true
→ connection installedThe AI can then run connection list --json and the demo appears alongside any other connections in your workspace. If the requested name is ambiguous ("snowflake" matches more than one demo), the response carries available_demo_connections and the AI re-prompts with a specific id.
You can also install demos by intent:
You: "Set me up with something I can use to try a wine quality analysis."AI calls install_demo_connection(demo_connection="wine", intent_text="wine quality")
→ resolves to bigquery-winequality-demoA demo connection installed in your workspace is a real connection. You can query it from any client (the same workspace appears in Claude, Codex, Cursor, Replit, and the web app), share it with teammates, build dashboards on it, and schedule jobs against it.
What's available
| ID | Type | What it contains |
|---|---|---|
snowflake-demo | Snowflake | An ANALYTICS database with a small retail schema — CUSTOMERS, ORDERS, PRODUCT_METRICS. Good for warehouse SQL, joins, aggregations, and dashboard authoring. |
salesforce-demo | Salesforce | A populated demo Salesforce org — accounts, opportunities, contacts, leads. Good for SOQL, picklist handling, and CRM-style workflows. |
aws-s3-demo | S3 | The opsrus-lakehouse-demo bucket: a bronze/silver/gold lakehouse with date-partitioned event and support-ticket data. Good for browse/download, log analysis, and Parquet exploration. |
bigquery-winequality-demo | BigQuery | The UCI Wine Quality dataset, hosted in BigQuery. Good for warehouse SQL on a familiar public dataset. |
mongodb-atlas-sample-dbs-demo | MongoDB | The standard MongoDB Atlas sample databases (movies, Airbnb listings, training data, etc.). Good for document-style query patterns. |
large-parquet-on-s3-demo | S3 | A larger Parquet dataset on S3. Good for testing scale — pagination, partitioned reads, and DuckDB performance against remote files. |
sftp-demo | SFTP | A demo SFTP server. Good for testing the SFTP path end-to-end: browse, query, test, upload, download. |
The most up-to-date list comes from the platform itself:
You: "What demo connections can I install?"The AI calls install_demo_connection with no specific id; the response carries available_demo_connections with current ids, labels, and descriptions.
Pairing demos with use cases
Each use case walkthrough is grounded in one or two of the demos above:
- Data Exploration —
snowflake-demo - Cross-Source Analysis —
snowflake-demo+salesforce-demo - Triage & Break-Fix —
snowflake-demo+aws-s3-demo - Reporting —
snowflake-demo - Pipelines —
snowflake-demo+ DuckDB
Install the relevant demo first, then run the prompts in the walkthrough as written.
Going from demo to real
The demo workflow is the same workflow you'll use against your own systems. When you're ready, swap the demo for a credentialed connection through connection_setup and reuse the query files, dashboards, and schedules you built against the demo — they live in your workspace's connections/<name>/ directories and are yours to keep.