Use cases

One memory for every AI tool.

At the end, Claude Code and ChatGPT have both worked in the same project, and your team can see all of it in one place.

Filmed on a sample company, Quillmere. Its people and data are invented.Watch on YouTube

What you need

  • Any plan, including Free. You bring the AI tools. Here they are Claude Code and ChatGPT, but any tool that connects to Stable Baseline will do.
  • Each tool connected to your Stable Baseline account. See the Claude Code and ChatGPT recipes.
  • A project to work in. Quickstart shows how to make one.

Steps

  1. Connect both tools

    In Claude Code, run this once in your repo:

    bash
    claude mcp add --transport http sb https://api.stablebaseline.io/functions/v1/cloud-serve/mcp

    Then type /mcp, choose sb, and authenticate. In ChatGPT, open the Stable Baseline app and sign in. Each tool asks what it may reach. Sign in as the same person in both, so both reach the same project.

  2. Ask Claude Code to write the docs

    In your repo, ask Claude Code to read your code and write a document into your project. For example:

    prompt
    Read our database schema and routes. Add a document called Order data and integrations to my Checkout service project, with a diagram of every table.

    Open the document in Stable Baseline. It fills in as Claude Code works.

    Claude Code's document, with a diagram of every table.
  3. Ask ChatGPT for the research

    In a new ChatGPT chat, type @ and pick Stable Baseline. Ask for a research note in the same project. For example:

    prompt
    Research why shoppers leave at checkout. Add a note called Why shoppers leave at checkout to a Research folder in my Checkout service project, with charts.

    ChatGPT asks you to confirm changes to your work. Confirm, then open the note.

    The research note, with its chart.
  4. Put the numbers in a table

    Ask ChatGPT to turn the numbers into a table. For example:

    prompt
    Put the numbers behind that note in a table called Checkout drop-off, in a database called Checkout insights.

    Open the database from your project's sidebar. The rows arrive as ChatGPT writes them.

    The table, filling row by row.
  5. Ask for a dashboard

    Ask for a dashboard on the same table. For example:

    prompt
    Build a dashboard from the Checkout drop-off table, with totals, a weekly trend and a filter for region.

    Open the database. The dashboard is a tab at the top of it.

    A dashboard from the same table.
  6. Pick it up as a person

    Use the filter at the top of the dashboard. In the film it is Region: press it, choose New Zealand, and every number follows. Press Reset to see everyone's view again.
  7. See who did what

    Open the document Claude Code wrote, and type a line of your own at the end. Then press the three dots, More actions, at the top right of the document, and choose Version history.

    Each version names who saved it: you, ChatGPT or Claude Code. A tool's changes show under its own name, with yours beside it.

    One record: a person, ChatGPT and Claude Code on the same document.

What each tool can reach

A connected tool only reaches what you allowed when you signed in, and never more than you can open yourself. See Permissions.

What to try next

Ask either tool for more, in plain words:

“Which tasks in the Checkout relaunch plan are overdue?”

“Add a risk to the work register: the payment provider's rate limits.”