MCP Server for KeywordsPeopleUse Keyword Research
Accelerate keyword research with an MCP server integrated with KeywordsPeopleUse, delivering real-time keyword insights, analysis and exportable results.
npx -y @data-skunks/kpu-mcpOverview
This MCP (Model Context Protocol) server connects the KeywordsPeopleUse keyword research API to MCP-compatible clients (for example Claude Desktop, Cursor or any other MCP client). It exposes keyword research capabilities — autocomplete suggestions, People Also Ask questions, social Q&A (Reddit/Quora) and semantic keyword suggestions — in real time so you can query and surface structured search intent signals directly inside your editor or assistant.
The server runs on Node.js and acts as a bridge between MCP clients and the KeywordsPeopleUse API. Use it locally as a command-based MCP server or remotely via the mcp-remote proxy. Results are returned in structured formats so they can be displayed, analyzed or exported (CSV/JSON) by the consuming tool.
Features
- People Also Ask (PAA) question extraction from Google
- Google Autocomplete suggestions
- Reddit and Quora question harvesting
- Semantic keyword suggestions (related and intent-driven keywords)
- Real-time responses via MCP (SSE) for interactive client integrations
- Exportable results (JSON/CSV) for downstream analysis or reporting
Installation / Configuration
Prerequisites:
- Node.js and npm installed
- KeywordsPeopleUse API key (available on the KeywordsPeopleUse Standard plan or higher)
- Clone the repo
- Create a .env file with your API key
KPU_API_KEY=sk_01234567890123456789012345678901
- Install dependencies
- Run the server directly (example)
Notes:
- The server expects the KPU API key in process.env.KPU_API_KEY.
- If running as a command-based MCP server, provide the absolute path to index.js in your MCP client configuration (examples below).
Remote connection (mcp-remote)
To expose a remote MCP server that MCP clients can access locally, use the mcp-remote proxy:
Example: connect to a hosted MCP endpoint using npx
To configure a client like Claude Desktop to connect via mcp-remote, add an entry to the client config (replace YOUR_API_KEY):
Restart the client after updating configuration.
Available Tools / Resources
The MCP server exposes a set of tools (MCP actions) for keyword research. Typical tool names and behaviors:
| Tool name | Description |
|---|---|
| peopleAlsoAsk | Returns People Also Ask question nodes for a seed keyword |
| autocomplete | Returns Google autocomplete suggestions for a term or prefix |
| redditQuestions | Fetches Reddit threads or questions related to a keyword |
| quoraQuestions | Fetches Quora questions matching a topic |
| semanticKeywords | Returns related/semantic keywords and intent signals |
| exportResults | Returns results in exportable JSON/CSV format |
Each tool returns structured JSON suitable for display in MCP-aware clients. Check the repository for the exact request/response schema and example payloads.
Integration Examples
Cursor (command-based MCP server)
- Mac/Linux
- Windows
Claude Desktop (local command)
- Mac/Linux example for claude_desktop_config.json:
Use Cases
Keyword discovery: Supply a seed keyword and retrieve autocomplete suggestions plus semantic keywords to expand keyword lists for SEO or content planning.
- Example: Query “vegan protein powder” and combine autocomplete + semanticKeywords to generate content headings.
Content ideation with PAA: Pull People Also Ask nodes for a topic to identify common user questions to answer in articles or FAQs.
- Example: Use peopleAlsoAsk for “home coffee roaster” to create an FAQ section.
Competitive research and social intent: Gather Reddit and Quora questions to understand real user pain points and intent signals not visible in search alone.
- Example: Use redditQuestions to find threads where users discuss frustrations or desired features for a product category.
Export workflows: Aggregate results and export as CSV/JSON to feed spreadsheets, keyword tracking tools or editorial backlogs.
Links & Resources
- Repository: https://github.com/data-skunks/kpu-mcp
- KeywordsPeopleUse: https://keywordspeopleuse.com
- MCP Quickstart: https://modelcontextprotocol.io/quickstart/user
- mcp-remote (npm): https://www.npmjs.com/package/mcp-remote
If you need examples of request/response payloads or want to extend the server, check the code in the repository for the concrete MCP tool implementations and schema definitions.