Get Notes

This MCP server lets you query multiple Get Notes knowledge bases through an intelligent interface. It can search and recall notes from different knowledge bases, automatically selecting the most relevant KB for your question, while protecting you with rate limits and automatic retries for transient failures.

How to use

You connect with an MCP client to the Get Notes MCP Server to search and recall knowledge across multiple knowledge bases. Install the MCP server locally, then run it and cite the server to your client. You can either run the server with a one-off command via npx or configure a Claude Desktop instance to manage multiple knowledge bases automatically.

With the standard setup, Claude will automatically pick the most relevant knowledge base when you ask a question. You can also explicitly specify which knowledge base to search or recall from when needed.

How to install

Clone the project repository to your machine

Install dependencies:

npm install

Create a local environment configuration from the example:

cp .env.example .env

Edit your environment file to include your API key and topic IDs, for example:

GET_API_KEY=your_api_key_here GET_NOTE_TOPIC_ID=your_topic_id_1

Run the MCP server locally (example using npx):

npx get_notebook_mcp_server

For development or testing, you can start the server directly with Node if you have an index file:

node index.js

Optionally, test interactively with the MCP Inspector:

npx @modelcontextprotocol/inspector node index.js

## Additional content

Configuration examples show how to expose multiple knowledge bases under a single API key and how Claude Desktop can expand grouped configurations into separate KB entries. The server supports two main MCP connection styles: a direct HTTP endpoint (not shown in this guide since a remote URL isn’t provided in the examples) and a local stdio setup using npx to run the server. Use the stdio configuration when you want to run the MCP server on your machine or within your development environment.

Environment variables used by the MCP runtime are shown in the configuration examples. You’ll typically set API keys and topic IDs to authorize access to the Get Notes data and to map knowledge bases to topics.

Two practical configuration examples are provided below. The first config runs a single knowledge base. The second config demonstrates grouping multiple knowledge bases under one API key, which Claude Desktop expands into individual bases automatically.

"mcpServers": { "get-notes": { "command": "npx", "args": ["-y", "get_notebook_mcp_server"], "env": { "GET_API_KEY": "your_api_key_here", "GET_NOTE_TOPIC_ID": "your_topic_id_1" } } }

"mcpServers": { "get-notes-multi": { "command": "npx", "args": ["-y", "get_notebook_mcp_server"], "env": { "GET_KNOWLEDGE_BASES": "[{"id":"medical_group","name":"Medical Knowledge Bases","config":{"api_key":"your_api_key_here","topics":[{"id":"kb_pancreatic_trials","name":"Tumor Clinical Trials KB","topic_id":"your_topic_id_1"},{"id":"kb_patient_experience","name":"Patient Experience KB","topic_id":"your_topic_id_2"}]}}]" } } }

## Notes on usage and troubleshooting

For testing, you can run the MCP server locally and use the MCP Inspector to verify endpoints and behavior. If you encounter rate-limit messages, ensure your requests stay within the configured QPS and total request limits. If you see transient network errors, the server’s retry mechanism will automatically retry problematic requests.

## Available tools

### list\_knowledge\_bases

List all knowledge bases with IDs, names, and descriptions so you can reference them in searches.

### search\_knowledge

AI-processed search across a knowledge base, returning synthesized results and references.

### recall\_knowledge

Raw recall from a knowledge base without AI synthesis, returning raw notes or files.