Skip to content

External AI Client Requirements

What you need before connecting ChatGPT, Claude, Cursor, or another AI client to the Qualytics MCP server. The same requirements apply to every client, so check them once before you follow the setup page for your client.

An API Token

The client never stores your Qualytics token. When you connect, the client opens a Qualytics authorization page in your browser, and you paste a token there. The client then works with a separate MCP access token tied to it.

Token Can authorize a client The client acts as
Personal API Token You, with your role and team permissions.
Service token The service account the token belongs to, with its role and teams.
SCIM Administration Token Not applicable. The authorization page refuses it.

Every user with the Member role or higher can generate a Personal API Token from Settings > Tokens. See Tokens for the steps, and MCP Permissions for what each role allows.

When your deployment does not allow API tokens, no token can authorize a client, and the authorization page answers Invalid API token.

Do not send the token in a header

A Qualytics API token in an Authorization: Bearer header does not authenticate an MCP connection. Paste the token only into the Qualytics authorization page that your client opens. If a client was configured with that header, remove it and connect again.

Your Qualytics MCP URL

The MCP URL is the address of your Qualytics instance followed by /api/mcp/:

https://your-qualytics-instance.qualytics.io/api/mcp/

Use the instance address shown by your Qualytics administrator. The authorization is tied to that address, so a hostname alias that reaches the same instance can still fail to authorize.

Supported Clients

The client must support MCP over streamable HTTP and complete the authorization in a browser. Every client below does, and each company's page has the setup steps for its clients. If your client cannot complete the browser authorization, use the CLI MCP Server instead.

No. Company Clients
1. Amazon Amazon Amazon Q Developer in the IDE, Kiro (IDE and CLI)
2. Anthropic Anthropic Claude Code, Claude Desktop
3. Cursor Cursor Cursor CLI, Cursor editor
4. Google Google Gemini, Gemini CLI
5. Microsoft Microsoft VS Code with GitHub Copilot
6. OpenAI OpenAI ChatGPT, Codex (app, CLI, and IDE extension)
7. Windsurf Windsurf Devin CLI, Windsurf editor, Windsurf plugins
8. xAI xAI Grok
9. MCP Other Clients Any MCP client with streamable HTTP and browser authorization

What You Do Not Need

  • An AI provider in Qualytics. Each client brings its own model, so the AI provider that AgentQ uses is not involved. See Add Integration only if you also want AgentQ's built-in chat.
  • Administrator setup. Connecting a client needs no MCP-specific permission and nothing to turn on in Qualytics.

See Also

  • Examples

    Real scenarios built on the prompts the Qualytics MCP server ships with, from explaining an anomaly to judging whether data can be trusted.

    Examples

  • Best Practices

    Keep client connections secure and get reliable results from them.

    Best Practices

  • Permissions

    The user roles, tokens, and team permissions behind what a client can do.

    Permissions

  • Introduction

    What MCP is, what the Qualytics MCP server offers, and why it matters for data quality.

    Introduction

  • How It Works

    How a client, the Qualytics MCP server, and the model work together on a request, and what the model sees.

    How It Works