--- title: "Talk to your data - Atlan MCP Cookbook" url: "https://atlan.com/mcp-cookbook/talk-to-your-data/" description: "Pair Atlan with your warehouse or SQL-tool MCP to find the right certified data in plain language, then go straight to it." --- > Atlan is hosting Context Conference, bringing together the leaders and builders at the frontier of giving AI the context it needs to understand their business. It runs online on October 28, 2026, from 11:00 AM to 2:00 PM ET. Atlan co-founder Prukalpa Sankar opens and closes the day. Leaders from AstraZeneca, BNY and Verizon share why they invest in context and what they get from it. Registrants get early access to The AI Context Gap, a new study from MIT Technology Review Insights. Register: https://atlan.com/context-conference/ Recipe 17 in [The Atlan MCP Cookbook](https://atlan.com/mcp-cookbook/). Pair Atlan with your warehouse or SQL-tool MCP: Atlan finds the right certified table and hands over the governed definition, so you query the source correctly - not just quickly. Use case: Find & understand data / Action: Look up / Persona: Data/Business User, Data/BI Analyst, AI Builder / Effort: Quick / Runs in: any AI tool Requires Atlan MCP connected to your AI tool: [set up Atlan MCP](https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/remote-mcp-overview). ## 1. The task **Scenario.** Someone needs revenue by customer for the last two years. They don't know which table to trust or how revenue is actually supposed to be calculated - and getting either wrong produces a number the business won't stand behind. This is a two-MCP move: the Atlan MCP finds the right certified table and, just as importantly, hands over its governed context - the owner, the certification, and the definition and calculation of the metric. A warehouse or SQL-tool MCP (Snowflake, your SQL client, and the like) then runs the query, using that definition so the result matches how the company actually defines revenue. The ask: - The right certified table for the data you need (say, order and revenue information) - The governed definition and calculation of the metric (how revenue is actually computed) - A query against the source that applies that definition, so the number is right Hero prompt: "I need revenue by customer for the last two years. Find the certified table with the order and revenue data, pull the governed definition and calculation of revenue from Atlan, then query the source using that definition so the number matches how we define it." Step by step: 1. Find the right certified asset. Prompt: "Find the certified table with [order/revenue] data, and show its owner and certification." 2. Pull the governed context - the definition and calculation. Prompt: "Get the governed definition and calculation of [revenue] from Atlan for that table." 3. Query the source with that context. Prompt: "Using that definition, query the table in [Snowflake] to return [revenue by customer for the last two years]." ## 2. What you'll give it Required context: - What you need, in plain language (the metric and any time range) - Two MCPs connected: the Atlan MCP (to find the asset and pull its governed context - definition, calculation, owner, certification) and a data-access MCP for your warehouse or SQL tool, e.g. Snowflake, to run the query Optional context: - Your team or domain, to prefer sources you already have access to - Whether you want only certified assets, or everything that matches ## 3. What Atlan creates Outputs: - The certified table to use, with its owner and certification - The governed definition and calculation of the metric, so the query computes it the company's way - A query run against the source that returns the data, calculated correctly - not just a table name **Summary.** Atlan supplies the context - the right table and the governed definition - so the query against the source is pointed at the trusted asset and calculated the way the business defines it. You get the number you can stand behind, not just a list of tables. **Read-only.** Nothing is written. Atlan reads the asset and its governed context, and the query against the source is read-only - so anyone can run it safely. ## 4. Follow-up prompts - **Compare the options:** "Between the top two tables, which is more current, and which do more teams use?" - **Get to the data:** "For the certified one, who do I ask for access, and is there a sample or data dictionary?" - **Refine the need:** "I actually need this by region. Which of these has a region column?" ## 5. Tips & troubleshooting **Cook's note.** Atlan hands you the definition, not just the table - so the query comes back right, not just runnable. - Match the tool to the work. Use whatever AI tool you already have open; the recipe reads, so there is nothing to set up and nothing to break. - Automate or chain it. Wire it into a Slack assistant so "where do I find X" questions get answered from Atlan without anyone filing a ticket. - It only reads. No writes, so ask freely and follow up as many times as you need to find the right source. ## Related recipes - [Route data requests to right assets](https://atlan.com/mcp-cookbook/route-data-requests-to-right-assets/): Surface certified, well-documented assets in response to a business request. - [Gamify your context](https://atlan.com/mcp-cookbook/gamify-your-context/): Generate per-domain enrichment scorecards to motivate metadata contributions. - [Trace a broken dashboard to its root cause](https://atlan.com/mcp-cookbook/trace-a-broken-dashboard-to-its-root-cause/): Trace lineage upstream from a broken dashboard to find the source of wrong numbers. [Browse all recipes](https://atlan.com/mcp-cookbook/) / [Set up Atlan MCP](https://docs.atlan.com/product/capabilities/atlan-ai/how-tos/remote-mcp-overview)