Get started with MCP
Connect your AI assistant to Querio and create your first analysis through MCP.
The first notebook uses example data. You don't need a database connection.
1. Connect Querio
Use an MCP client that supports remote Streamable HTTP connections, browser authorization and MCP resources.
Add Querio using this server URL:
https://app.querio.ai/mcpFollow your client's sign-in flow and approve the connection. If you don't have a Querio account, choose Sign up and verify your email.
Return to your assistant once connected. You need permission to create notebooks and run code. Your assistant uses your Querio access; approving the connection doesn't grant additional permissions.
2. Connect to your workspace
Ask your assistant:
Use Querio's workspaces_list tool to find my workspace.Starting a new team with no workspace yet? Ask it to create one:
Create a Querio workspace called "My team".If you're joining an existing team, follow its invitation instead.
Your assistant uses the returned workspace ID for subsequent operations. You don't need to copy it between messages.
3. Create your first analysis
Querio provides two guides through MCP: a workflow guide and a notebook-authoring reference. Your assistant should read them before creating the analysis.
Give it this task:
Use Querio MCP to create a notebook called "Revenue quickstart"
in my workspace.
First read these MCP resources:
querio://guides/skill
querio://guides/notebook-authoring
Use this example data:
month,revenue
2026-01,1200
2026-02,1500
2026-03,1800
Create notebook cells containing the data, a monthly revenue
chart, and the total revenue. Run the cells and check the results.
Confirm the notebook saved successfully. Then close only the
session and execution environment you created.
Show me the findings and the saved notebook ID.Your assistant writes the analysis, and Querio executes it. The example data and analysis remain in the saved notebook after its runtime closes.
4. Check the result
The total should be 4,500, with revenue increasing each month.
Ask your assistant to show the calculation and chart. To check that the work was saved, ask it to read the notebook using notebooks_get and inspect its cells.
If execution succeeds but saving fails, keep the session open and resolve the save error before closing it. Repeating the creation task could duplicate work.
5. Start working with your data
Ask your assistant to help connect a datasource. It can open a Querio setup link for you to enter credentials in your browser. Keep database passwords out of the conversation.
Give Querio the definitions and business rules behind your questions. Manage context in the app, through Git, or with the CLI. MCP can copy selected file contents with files_mkdir and files_write. Use the CLI for directory uploads, file reads and downloads; see MCP file tools.
Ask a question about your business
Start with something your team already knows how to answer. Compare the result with a trusted source, then build on the analysis.
For supported tools, connection troubleshooting and technical details, see the MCP reference.
Last updated on October 27, 2026