MCP 服务器

Minds MCP Server

Minds MCP enables AI assistants to run synthetic market research end to end using grounded synthetic customer panels for concept testing, message testing, and segment comparison, with durable server-side studies exporting to PDF, CSV, XLSX, JSON, and Markdown.

README

Minds MCP Server

Run AI market research from your assistant: synthetic customer panels for concept testing, message testing, and segment comparison.

Endpoint: https://getminds.ai/mcp (streamable HTTP) Docs: https://getminds.ai/mcp/setup Quick-start guide: https://minds-ai-co.github.io/minds-mcp/ Technical reference: https://minds-mcp.readthedocs.io/en/latest/ GitBook: https://minds-1.gitbook.io/minds-mcp/ Client compatibility: https://minds-mcp-compatibility.pages.dev/ Setup generator: https://minds-mcp-setup-generator.netlify.app/ Public API explorer: https://minds-api-explorer.vercel.app/?utm_source=content&utm_medium=content&utm_campaign=content-seo-mcp-reference&utm_content=mcp-repository-api-explorer API + MCP integration lab: https://minds-api-mcp-integration-lab-bba685.gitlab.io/?utm_source=content&utm_medium=content&utm_campaign=content-seo-mcp-reference&utm_content=mcp-repository-integration-lab Archived release: https://doi.org/10.5281/zenodo.21711429 Auth: OAuth 2.1 with PKCE, or a Minds API key

This repository is documentation for a hosted MCP server. There is no code to install or run. The server is operated by Minds and the implementation lives in our application, not here.

Documentation maintenance

reference/ is the single source for maintained integration content. Read the Docs and GitBook publish the same Markdown through repository sync. GitHub Pages, Cloudflare Pages, and the Netlify setup generator are intentionally thin, platform-specific surfaces; Zenodo preserves versioned release snapshots. See DOCUMENTATION.md for the publishing contract.

What it does

Minds is a synthetic market research platform. This MCP server lets ChatGPT, Claude, Cursor, and any other MCP client run customer research end to end without leaving the assistant.

Describe an audience in a brief ("German Gen Z grocery shoppers", "enterprise IT buyers in fintech") and the server runs deep web research to ground that audience in government statistics, peer-reviewed studies, and industry reports rather than in generic model priors. Then ask the panel a question, run a structured study, and export the results.

Typical jobs:

Concept testing and message testing before creative production
Ad pretesting, landing page audits, and packaging tests with per-segment reactions
Buyer persona validation and objection discovery for B2B go to market
Segment comparison and positioning checks across markets
Screening research hypotheses before commissioning fieldwork with real respondents

Studies run durably server side, so a long study survives the chat session that started it. Results export to PDF, CSV, XLSX, JSON, and Markdown.

Scope. Minds does not replace representative human fieldwork. It replaces the slow first pass: sharpening the question, surfacing objections, and deciding which assumptions deserve real-respondent validation.

Setup

ChatGPT

Settings, then Connected Apps, then add https://getminds.ai/mcp and authorize via OAuth. Panel results render as interactive widgets inline.

Claude Desktop and claude.ai

Customize, then Connectors, then add https://getminds.ai/mcp as a remote connector and authorize via OAuth.

Claude Code

claude mcp add --transport http mindsai https://getminds.ai/mcp \
  --header "Authorization: Bearer minds_YOUR_API_KEY"

Cursor

Settings, then MCP, then add https://getminds.ai/mcp and authorize via OAuth. The repository also contains a Marketplace-ready Cursor plugin manifest at .cursor-plugin/plugin.json; its MCP configuration is intentionally kept in cursor.mcp.json so the shared Agent Plugins configuration can retain the portable streamable-http transport name.

Google Antigravity

agy plugins install https://github.com/minds-ai-co/minds-mcp

The root plugin.json and mcp_config.json make this repository an Antigravity plugin. Antigravity discovers OAuth from the hosted endpoint and prompts you to authenticate. You can also add Minds directly in the Antigravity MCP manager with the server URL https://getminds.ai/mcp.

Gemini CLI remains supported for Google Cloud enterprise and API-key users. Its legacy extension manifest stays available as gemini-extension.json.

ChatGPT and Codex Plugins

The repository includes the shared plugin package at .codex-plugin/plugin.json and its remote server configuration in .mcp.json. The public directory listing uses the same hosted endpoint and maintained reference content.

VS Code (Copilot), Windsurf, Langdock, Open WebUI

Add https://getminds.ai/mcp as a remote MCP server. Full per-client instructions, including API key setup, are at https://getminds.ai/mcp/setup.

API keys

Generate one in the Minds app under Settings, then API Keys. Keys start with minds_ and are passed as Authorization: Bearer minds_....

Tools

Tool	What it does
`list_groups`	List your saved Audiences, segments, and persona collections.
`create_group_from_brief`	Build a grounded synthetic Audience from a population brief via deep web research.
`ask_group`	Ask, survey, or poll a single Audience and get per-respondent answers.
`list_panels`	List your panels, studies, surveys, and focus groups.
`create_panel`	Create a panel or study from a research goal.
`ask_panel`	Ask one straightforward question across a whole panel.
`get_panel_status`	Check panel composition and live progress on in-flight questions.
`export_panel`	Export panel results as PDF, CSV, XLSX, or JSON.
`plan_panel_study`	Draft a structured research plan for multi-question studies, asset audits, and MaxDiff.
`run_panel_study`	Execute a confirmed study plan as a durable server-side run.
`get_panel_study`	Check durable progress, artifacts, and method calculations for a running study.
`list_research_methods`	List supported research methods and which of them can actually execute.
`get_panel_summary`	Return a whole-study summary with semantic blocks and heatmaps.
`list_study_drafts`	List reusable study-plan drafts in the connected workspace.
`save_study_draft`	Save a study plan as a reusable draft before execution.

Research methods: MaxDiff runs through a deterministic adapter. Conjoint is planned and cannot execute yet. Call list_research_methods for the current state rather than assuming.

Discovery

Server card: https://getminds.ai/.well-known/mcp/server-card.json
OAuth protected resource metadata: https://getminds.ai/.well-known/oauth-protected-resource
Registry name: ai.getminds/minds