MCP server

The Conjointly MCP server lets AI assistants work with your Conjointly data. Once connected, you can ask about your experiments in plain language: list your studies, read reports and analysis results, check respondent quality, start analyses and request exports, without writing any code.

MCP (Model Context Protocol) is an open standard for connecting AI assistants to other services. The Conjointly MCP server works on top of the same Conjointly API, with the same permissions, but it is limited to reading data, running analyses and requesting exports. Analyses and exports produce new results and files, but the MCP server cannot create, change or delete your experiments, reports, settings or respondent data.

Server address

https://api.conjoint.ly/mcp

You sign in with your Conjointly account when you connect. There is no API token to copy.

Connect your assistant

Claude (web and desktop)

1. In Claude, open Settings, then Connectors.
2. Click Add custom connector.
3. Enter a name, such as Conjointly, and the server address https://api.conjoint.ly/mcp.
4. Click Connect. Sign in to Conjointly if asked, including two-factor authentication if your account uses it.
5. Review the request and click Authorise.

Claude Code

Run:

claude mcp add --transport http conjointly https://api.conjoint.ly/mcp

Then run /mcp inside Claude Code and choose conjointly to sign in.

ChatGPT

Custom MCP servers need ChatGPT’s developer mode, which is available on the web for Plus, Pro, Business, Enterprise and Education plans. On Business, Enterprise and Education plans, a workspace admin may need to allow developer mode or add the connector for the workspace first.

1. In ChatGPT, open Settings, then Apps (called Connectors on some plans), and open Advanced settings.
2. Turn on Developer mode.
3. Back in Apps, click Create.
4. Enter a name, such as Conjointly, a short description, such as Conjointly experiments and reports, and the server address https://api.conjoint.ly/mcp.
5. Set Authentication to OAuth, confirm that you trust the server, and click Create.
6. Sign in to Conjointly when asked and click Authorise.

To use it, start a new chat, click + next to the message box, and choose Conjointly under Developer mode. ChatGPT asks you to confirm each tool call before it runs.

Microsoft 365 Copilot

Microsoft 365 Copilot reaches MCP servers through agents built in Copilot Studio, so you need a Copilot Studio licence, and your Microsoft 365 admin may need to approve the agent before others can use it.

1. In Copilot Studio, create an agent or open an existing one.
2. Go to Tools, click Add a tool, then New tool, then Model Context Protocol.
3. Enter a name, such as Conjointly, a short description, and the server address https://api.conjoint.ly/mcp.
4. Set Authentication to OAuth 2.0, choose Dynamic discovery, and click Create.
5. Create a new connection, sign in to Conjointly when asked, and click Authorise. Then add the tool to the agent.
6. To use the agent in Microsoft 365 Copilot or Teams, open Channels, choose Teams and Microsoft 365 Copilot, and publish it.

Each person who uses the agent signs in with their own Conjointly account the first time it calls the server.

Gemini Enterprise

Gemini Enterprise is not supported yet, because of a limitation on Google’s side. It requires an OAuth client ID and secret to be set up for it on the Conjointly side in advance, and does not support dynamic client registration, which the other clients here use to set up their sign-in automatically.

Mistral Vibe

Custom MCP connectors are available on all Mistral Vibe plans. On Team and Enterprise plans, an admin adds the connector and members then turn it on in their chats.

1. In Mistral Vibe, open the connector marketplace.
2. Click Add a custom connector.
3. Enter a name, such as Conjointly, a short description, and the server address https://api.conjoint.ly/mcp.
4. Choose OAuth as the authentication method, then click Connect.
5. Sign in to Conjointly when asked and click Authorise.

To use it, turn on Conjointly from the tools menu under the message box in a chat.

Other MCP clients

Any client that supports remote MCP servers over the Streamable HTTP transport with OAuth sign-in can connect to the same address. Some clients, including Claude Code, label this transport simply http.

What the assistant can do

The assistant sees only what you can see in Conjointly, and every request goes through the same permission checks as the platform.

Read (marked as read-only, so Claude runs these without asking; other clients may still ask):

• Experiments: list and search experiments, read their design, settings, sources, folders, sharing and activity log.
• Reports: report data, segments, pivot tables, weighting, waves, simulations and analysis results.
• Respondents: search participants, read individual answers, URL variables and quality signals.
• Account: team members, licence, balance, invoices, own audiences and predefined panels.

Start work (marked as not read-only, so Claude asks for your approval before each one; other clients may handle this differently):

• Run report analyses, such as attribute importances or level preferences for conjoint questions.
• Request exports, such as respondent data in Excel or SPSS, or individual conjoint utilities.

These tools produce analysis results and export files only. The MCP server cannot create, change or delete experiments, reports, settings or respondent data.

Example prompts

• “List my experiments launched this year and how many responses each has.”
• “Show the attribute importances for my latest conjoint study.”
• “Which respondents in experiment 12345 look like low-quality responses, and why?”
• “Export the respondent data for experiment 12345 to Excel.”

Disconnecting

Removing the connector in your AI assistant stops it using Conjointly, but does not revoke its sign-in on our side. To cut off MCP access at once, a team leader can tick Disable MCP access in Team settings.

Feedback

The MCP server is new. We would like to hear how you use it and what you would like it to do.