Content
# Magic Insights MCP server
Connect any [MCP](https://modelcontextprotocol.io) client to your [Magic Insights](https://magicinsights.ai) workspace and chat with your customer-feedback **studies** — list studies, search and filter feedback, and read AI-generated themes, topics, suggestions, tags, and sentiment trends.
This is a **remote, hosted** MCP server. There's nothing to install — point your client at the URL below and authenticate with a workspace API key.
- **Server URL:** `https://app.magicinsights.ai/api/mcp`
- **Transport:** Streamable HTTP
- **Auth:** API key (bearer token)
- **Official MCP registry:** [`ai.magicinsights/insights`](https://registry.modelcontextprotocol.io/v0/servers?search=ai.magicinsights)
## Get an API key
Create one in the app at **Settings → API keys** (`https://app.magicinsights.ai` → your workspace → Settings → API keys), then send it as the `Authorization: Bearer <key>` header.
## Connect
### Claude Code
```bash
claude mcp add --transport http magic-insights https://app.magicinsights.ai/api/mcp \
--header "Authorization: Bearer YOUR_API_KEY"
```
### Cursor
Add to `~/.cursor/mcp.json` (all projects) or `.cursor/mcp.json` (one project), then enable the server in Cursor settings:
```json
{
"mcpServers": {
"magic-insights": {
"url": "https://app.magicinsights.ai/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
```
### Other clients (Claude Desktop, Windsurf, MCP Inspector, …)
Most clients use the same `mcpServers` block above — check your client's docs for where its config file lives. The Claude API MCP connector accepts the same URL with an `authorization_token`.
## Tools
All tools are **read-only** and scoped to the workspace that owns the API key.
| Tool | What it does |
| --- | --- |
| `get_workspace` | One-call overview: plan, study count, connected platforms, tags, total feedback, date range, overall sentiment. |
| `list_studies` | List all studies with description and date range. |
| `get_study` | A study's summary, generated report, and top themes (with sentiment, ids, and example feedback). |
| `list_topics` | A study's topics and the themes grouped under each. |
| `get_theme_feedback` | All feedback under one theme, most relevant first. |
| `list_suggestions` | A study's AI-synthesized recommended actions (problem, solution, priority). |
| `get_suggestion_feedback` | The individual feedback behind a single suggestion. |
| `search_feedback` | Search and filter feedback by keyword, sentiment, platform, tag, or date range. |
| `get_feedback` | One feedback item in full — content, translation, AI summary, and source metadata. |
| `get_stats` | Aggregate stats: total count, sentiment breakdown, date range. |
| `get_trends` | Feedback volume and sentiment over time, bucketed by week or month. |
| `list_tags` | The workspace's custom tags and how many items carry each. |
| `list_platforms` | The source platforms feeding the workspace, each with its feedback count. |
## About
Magic Insights is an AI-driven customer-feedback analytics platform. Learn more at [magicinsights.ai](https://magicinsights.ai).
The server's metadata is published in the [official MCP registry](https://registry.modelcontextprotocol.io/v0/servers?search=ai.magicinsights); this repository hosts the public `server.json` and connection docs. The application source is closed.
