Connect AI and code

Your data in Mycorr is not locked in the app. AI assistants, scripts and notebooks can read it, and build on it, under exactly the permissions you have in the app.

Which one to use

You want to…Use
Ask questions about your data in plain language, or have an assistant build and fill tablesMCP server, signing in through the browser
Pull tables into a notebook or a data pipelinePython SDK with a token
Run an AI agent or a client unattended, without a browserMCP server with a token
Hand a table to someone without codeExport to CSV, Parquet or Google Sheets

How access works

The same rules hold whichever way you connect:

  • Your permissions, nothing more. A connection or token can read and change only what your account can, and a change to your access applies on its next request.
  • One organization at a time. A token belongs to the organization you were working in when you created it, and an assistant's connection to the organization you pick when you approve it. Through a company organization's connection, your personal models are readable too, but never writable.
  • Read-only unless you say otherwise. Tokens and connections are read-only unless you choose Read & write.
  • Releases outside your models. Where you can edit a table you read its latest version; anywhere else you read its release.
  • Usage counts. Data that leaves Mycorr counts toward your organization's egress allowance, and each AI tool call toward its monthly tool-call allowance. See Usage and limits.

Where things live

Everything you need is on the Access page, at the foot of the sidebar: AI clients with step-by-step instructions for each assistant, and Personal access tokens with Use a token with Python.

IDs that the tools take are one click away in the app: Copy model ID on a model's menu, and Copy table ID on a table's.