Dimensions Analytics MCP
Connect Cursor, Claude Desktop, VS Code, and other MCP clients to Dimensions research data.
Dimensions Analytics MCP lets your AI assistant search publications, grants, researchers, reports, source titles, and more; find topic-similar documents from an abstract; run analytics and trends; look up DOIs and Dimensions IDs; and explore the live Dimensions schema — all through natural language in tools your client already supports.
After a short install, you configure your MCP client once, restart the app, and use dimensions like any other integration. Example prompts and workflows are in docs/USAGE.md.
Prerequisites
- Node.js 20+ — required for the guided installer and the local server; Claude Desktop supplies its own Node runtime for the extension, and the hosted server needs none
- Dimensions API key — from your Dimensions account (same key as the Dimensions DSL API)
Don’t have an API key yet? Request a demo or quote here.
Quick install
Claude Desktop (macOS or Windows): download the latest extension, then open Settings → Extensions → Advanced settings → Install Extension… and select the file. Enter your API key in the extension settings. See installation details.
Other MCP clients: use the guided installer below.
Hosted server (no install): clients that can send an Authorization header — Cursor, VS Code, Windsurf, Claude Code, Codex CLI — can connect straight to https://app.dimensions.ai/mcp with Authorization: Bearer <your API key>. The guided installer offers this too. See Hosted server.
Mac / Linux — paste in Terminal:
bash -c "$(curl -fsSL https://raw.githubusercontent.com/digital-science/dimensions-analytics-mcp/main/scripts/install.sh)"
Windows — paste in PowerShell:
irm https://raw.githubusercontent.com/digital-science/dimensions-analytics-mcp/main/scripts/install.ps1 | iex
The installer checks Node.js, downloads Dimensions Analytics MCP, asks for your API key and, if needed, a custom instance URL, and configures Claude Desktop, Cursor, VS Code (Copilot), or Windsurf. Full details: Installation.
No GitHub token is required. The installer downloads @digital-science-dsl/dimensions-analytics-mcp from npm.
Custom Dimensions instance: If you log in at a URL other than https://app.dimensions.ai (for example https://your-instance.dimensions.ai), you must set DIMENSIONS_BASE_URL to that URL. A key from a custom instance fails with 401 Unauthorized against the standard host. The installer asks for this; for manual config:
"env": {
"DIMENSIONS_API_KEY": "your-api-key",
"DIMENSIONS_BASE_URL": "https://your-instance.dimensions.ai"
}
When done, restart your AI app and look for dimensions in MCP settings. Guided prefix installs update themselves for same-major npm releases; other install methods log an upgrade warning. Details: Updating.
Using Dimensions in chat
Once connected, ask your assistant to search publications, grants, researchers, reports, journals, run analytics, or look up a DOI.
Tool reference: docs/USAGE.md.
More documentation
- Installation — guided installer, manual setup, and the hosted server
- Usage guide — workflows for AI assistants
- Reference — names, env vars, rate limits, schema cache
- MCP field aliases — field name shortcuts accepted by search tools
Package: @digital-science-dsl/dimensions-analytics-mcp on npm.
Privacy Policy
This extension runs on your computer. It sends your API key and your queries only to the Dimensions API at the instance you configure (app.dimensions.ai by default). With each request it also sends usage metadata: the tool name, the extension version, the MCP client name, and a random session ID created when the server starts. Digital Science uses this metadata to monitor and improve the service. The extension doesn't send data to any other service, and the Claude Desktop extension doesn't check npm for updates.
How Digital Science collects, uses, stores, shares, and retains this data, and how to contact us, is described in the Dimensions privacy policy. Questions: [email protected].
License
MIT