datatomcp
Sign inStart free

4 November 2026 · Killian, founder

datatomcp is now an AI-native data layer

datatomcp started as a way to connect one spreadsheet to one AI agent. Most teams now use several agents on the same data. Reaching the data is solved; the hard part now is making sure every agent means the same thing by "revenue". So datatomcp has grown into the layer between a team's data and its agents.

What "AI-native data layer" means here

One place where your team's sheets and files live, kept fresh and linked together, with the meaning of every number written down, and a record of every question an agent asked. Agents connect over MCP or REST. A database does the counting, so totals are exact. It does the work of a data warehouse and a semantic layer for teams that don't have a data team.

What's new

  • Definitions. When you add data, datatomcp drafts definitions from the columns. A person approves or edits each one. Agents only use approved definitions, every answer that uses one says who approved it, and a glossary explains your team's words.
  • Audit. Every agent question is kept: who asked, from which agent, on which dataset, the query, the time and any error. When two agents ask the same thing, you see whether they got the same answer. Export it as a spreadsheet.
  • Map. One page shows which agents asked, which definitions they used, which datasets those read, and how the datasets link.
  • Charts. Our own library of 52 chart types, from a query over the API and in the in-app assistant. Inside Claude and ChatGPT from 7 November.
  • Assistant. Ask about a workspace inside datatomcp, on your own AI key (Anthropic, OpenAI, Mistral and others).
  • AI cleaning, opt-in. AI suggests clean-up rules like "treat n/a as empty". A person approves each one, and it runs again on every refresh.
  • API reference. Every REST endpoint is documented on /docs/api, with an OpenAPI spec.

What stays the same

  • Connecting a Google Sheet, CSV or Excel file still takes about two minutes.
  • Agents get read-only access and never run raw SQL.
  • Every answer names the dataset version it came from and links to the rows behind it.
  • Your data is stored in the EU and never used for training. Anything that uses AI is off until you turn it on.
  • Free for 3 datasets and 2 people, with no card needed.

Who it's for

Teams that work from spreadsheets and want their agents to give the same, checkable answer, and the consultants and agencies who set this up for their clients. If you used datatomcp to connect one sheet to Claude, nothing changes for you: the same MCP URL keeps working, and the new pages are there when your team grows.

Questions or feedback: [email protected]. You get a reply from me.