Skip to content

Testable MCP for your AI app

You do not need to learn the interface. Tell ChatGPT, Claude or Grok what you need and it does the work in your Testable account.

You ask
Build a Flanker task in Testable: 8 practice and 96 test trials, keys F and J.
Example
You getFlanker tasktrial fileValidated
Example trial file rows
Rowtypestim1keycondition1
1instructions
2practice>>>>>jcongruent
3practice>><>>fincongruent
rows 4 to 9
10test<<><<jincongruent
rows 11 to 105

105 rowsinstructions, 8 practice, 96 test

What is MCP?MCP (Model Context Protocol) is the open standard that lets an AI app work inside another service. Testable Magic is the AI inside Testable. Testable MCP connects your own AI app to all your projects.

Three steps, once per app. No code.

  1. Add Testable

    Open chatgpt.com/plugins, select +, then Add custom MCP server. Fill in the form.

    Name
    Testable
    Server URL
    https://testable.org/mcp
    Authentication
    OAuth

  2. Install and sign in

    Select Create as a plugin, then install it from your personal plugins. ChatGPT opens Testable. Sign in and click Connect.

  3. Ask

    In a new chat, type @ and pick Testable.

    List my Testable projects.

ChatGPT shows a notice for every custom server. Select I understand and want to continue. Menus look different? See the OpenAI guide to custom MCP servers.

When your app lists your projects, you are connected. Any other app that supports remote MCP uses the same address.

Say it as you would to a colleague. Name Testable and your project.

  • BuildCreate a Testable project with a Stroop task: 4 practice and 40 test trials, keyboard responses.
  • EditCopy my Testable project "Go/No-go v2" and add 4 practice trials with feedback at the start.
  • AnalyzeIn my Testable project "Stroop online", how many participants finished, and what are mean RT and accuracy by condition?
  • RecruitHow many respondents has my Testable Minds study for "Stroop online" collected so far?

Classic tasks import ready-made from the Testable library. For anything else, your AI app writes the task as an HTML page.

  1. Your AI app writes it and uploads the page as a stimulus.

  2. Participants see it as a trial, like any other.

  3. The Testable SDK saves it in your results, with your extra columns.

Example: a mouse tracking task. Custom HTML stimuli reference
  • Your AI app acts as youIt sees only the projects you own or collaborate on.

  • Technical identifiers stay withheldIP address, browser session id and completion code never reach your AI app.

  • Live studies with results change only after you confirmUntil you do, Testable refuses the change.

  • Testable Minds stays read-onlyYour AI app reads budget, studies, sessions and respondents.

For your ethics application: the responses you ask about go to the AI app you chose, under its terms. IP addresses, session ids and completion codes never do. Testable MCP connector lists every field the tools can return. Disconnect an app at any time under Account, then Testable MCP.

Claude Code, Codex and CursorSetup prompt, commands and tokens

Paste this into your coding agent. It asks you before it changes a config file.

Connect Testable (testable.org) to this agent. Read https://testable.org/docs/ai-agents/setup.md and follow the steps for the agent you are. Ask me before you change any config file. When it is done, run the "whoami" tool to check the connection.

Claude Code

Terminal window
claude mcp add --transport http testable https://testable.org/mcp

Then run /mcp, choose testable and sign in through the browser.

Codex

Terminal window
codex mcp add testable --url https://testable.org/mcp

Sign in when Codex asks. If it does not, run codex mcp login testable.

Cursor

{
"mcpServers": {
"testable": {
"url": "https://testable.org/mcp"
}
}
}

Add it to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for every project.

For a script with no browser, create a token under Account, then AI agents and send it as Authorization: Bearer <token>.

Open this page in the AI app you use.

Message box empty? Copy the prompt

Read https://testable.org/docs/getting-started/use-testable-with-ai-agents/ and explain it to me as a researcher who runs online experiments on Testable: what Testable MCP lets my AI app do in my Testable account, how I connect the AI app I use, and how my projects and my participants' data stay protected. End with the first prompt I should try.