evernote-unofficial-api

Unofficial Evernote API client & MCP server, reverse-engineered from the Evernote Web App (Ion v11.10.1). Full CRUD, semantic search, AI features, and 22 MCP tools for Claude Desktop / Claude Code.

MCP Server — Quick Start

# 1. Install
npm install

# 2. Authenticate (opens browser — only needed once)
npx tsx src/mcp-auth.ts

# 3. Start the MCP server
npx tsx src/mcp-server.ts

Claude Desktop Integration

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "evernote": {
      "command": "npx",
      "args": ["tsx", "src/mcp-server.ts"],
      "cwd": "/path/to/evernote-api"
    }
  }
}

Claude Code Integration

claude mcp add evernote -- npx tsx src/mcp-server.ts

22 MCP Tools Available

Related MCP server: MCP Notepad Server

TypeScript API — Quick Start

# Install dependencies
npm install

# Run the example (opens browser for OAuth2 login)
npx tsx src/example.ts

On first run, your browser will open for login via Evernote's official OAuth2 PKCE flow. After authenticating, tokens are cached at ~/.evernote-api/tokens.json and reused automatically.

Usage

import { authenticate, EvernoteClient, textToENML, markdownToENML } from "./src/index.js";

// Authenticate (opens browser on first run, uses cached tokens after)
const tokens = await authenticate();
const client = new EvernoteClient(tokens);

// Get user info
const user = await client.getUser();

// Create a note from plain text
await client.createNote({
  title: "My Note",
  content: textToENML("Hello from the API!"),
});

// Create a note from Markdown
await client.createNote({
  title: "Meeting Notes",
  content: markdownToENML("# Heading\n\n- Item 1\n- Item 2"),
});

// Semantic (AI) search
const results = await client.searchSemantic({
  query: "meeting notes from last week",
  maxResults: 10,
});

// AI Summarize
const summary = await client.aiSummarize({
  content: "Long text...",
  style: "bullet",  // bullet | email | meeting | paragraph | twitter
});

// AI Rephrase
const rephrased = await client.aiRephrase({
  content: "Make this sound better",
  style: "formal",  // concise | formal | friendly | funny | engaging | empathetic
});

// Export notes
await client.exportNotes(["noteId1", "noteId2"]);

// Raw request to any endpoint
await client.raw("GET", "/v1/some/endpoint", undefined, { param: "value" });

Available Methods

Notes

• createNote(params) — Create a new note

• updateNote(params) — Update an existing note

• deleteNote(noteId) — Delete (trash) a note

• exportNotes(noteIds) — Export notes

• requestNoteAccess(noteId) — Request access to a shared note

• scheduleReminder(noteId, time) — Schedule a reminder

Notebooks

• createNotebook(params) — Create a notebook

• deleteNotebook(notebookId) — Delete a notebook

Tags

• createTag(params) — Create a tag

• updateTag(tagId, params) — Update a tag

Search

• searchSemantic(params) — AI-powered semantic search

• semanticAnswer(query) — Get AI answer from your notes

• relatedNotesOrAnswer(query) — Related notes or AI answer

AI Features

• aiSummarize(params) — Summarize text

• aiRephrase(params) — Rephrase text

• aiSuggestTags(noteGuid) — Get tag suggestions for a note

• aiSuggestTitle(noteGuid) — Get title suggestion for a note

• aiCopilot(prompt) — AI copilot interaction

• aiDetectText(content) — Detect AI-generated text

User & Account

• getUser() — Get current user profile

• getUsage() — Get usage statistics

• getDevices() — List connected devices

• getBilling() — Get billing info

• getPayments() — Get payment history

Shortcuts

• createShortcut(shortcut) — Create a shortcut

• updateShortcut(id, update) — Update a shortcut

• deleteShortcut(id) — Delete a shortcut

Utilities

• richLink(url) — Generate rich link preview

• ocrBusinessCard(imageData) — OCR a business card

• getThumbnailUrl(noteId) — Get note thumbnail URL

• getUserPhotoUrl(size) — Get user photo URL

ENML Helpers

import { textToENML, markdownToENML, wrapInENML, checklistToENML, enmlToText } from "./src/index.js";

// Plain text → ENML
textToENML("Hello\n\nWorld");

// Markdown → ENML
markdownToENML("# Title\n\n**Bold** and *italic*");

// Raw HTML → ENML
wrapInENML("<p>Custom <b>HTML</b></p>");

// Checklist → ENML
checklistToENML([
  { text: "Buy groceries", done: true },
  { text: "Clean house", done: false },
]);

// ENML → plain text
enmlToText("<en-note><p>Hello</p></en-note>");

Architecture

Your Code → EvernoteClient → api.evernote.com (REST /v1/*)
                           → api.evernote.com/command (NSync mutations)
                           → www.evernote.com/shard/{s}/notestore (Thrift, advanced)

Auth tokens are obtained via OAuth2 PKCE against accounts.evernote.com — the same flow the official web client uses.

Production / Dokploy (Docker)

Important: This MCP server uses stdio transport. To use it remotely, your agents must run on the same host/container or you must add a proxy layer (e.g., mcp‑proxy / stdio bridge). If your agents already run in the VPS, you can deploy this container and point them to the command inside it.

1) Build the image

docker build -t evernote-mcp-server:latest .

2) Prepare tokens (one‑time)

You need a valid token file. Run auth once on a machine with a browser:

# Option A: locally (recommended)
EVERNOTE_TOKEN_PATH=./tokens.json npx tsx src/mcp-auth.ts

# Then upload tokens.json to the server

Or run inside a container and mount /data:

docker run --rm -it -v $PWD/data:/data -e EVERNOTE_TOKEN_PATH=/data/tokens.json evernote-mcp-server:latest node dist/mcp-auth.js

3) Run in production

docker run -d \
  --name evernote-mcp-server \
  -e EVERNOTE_TOKEN_PATH=/data/tokens.json \
  -v /opt/evernote-mcp:/data \
  evernote-mcp-server:latest

Dokploy settings (suggested)

• Image: evernote-mcp-server:latest

• Env: EVERNOTE_TOKEN_PATH=/data/tokens.json

• Volume: /opt/evernote-mcp:/data

• Port: none (stdio server)

MCP Proxy (stdio → HTTP/SSE)

If you need remote access (HTTP/SSE), use the proxy. This exposes the stdio MCP server over HTTP.

Local run

# Build first
npm run build

# Run proxy (both SSE + streamable HTTP)
npm run proxy

# Or explicitly choose SSE / stream
npm run proxy:sse
npm run proxy:stream

Docker (proxy)

Build with the proxy Dockerfile:

docker build -f Dockerfile.proxy -t evernote-mcp-proxy:latest .

Run:

docker run -d \
  --name evernote-mcp-proxy \
  -e EVERNOTE_TOKEN_PATH=/data/tokens.json \
  -e MCP_PROXY_PORT=8080 \
  -e MCP_PROXY_API_KEY=YOUR_SECRET \
  -v /opt/evernote-mcp:/data \
  -p 8080:8080 \
  evernote-mcp-proxy:latest

Dokploy (proxy) settings

• Dockerfile: Dockerfile.proxy

• Env:

  • EVERNOTE_TOKEN_PATH=/data/tokens.json

  • MCP_PROXY_PORT=8080

  • MCP_PROXY_API_KEY=... (recommended)

• Volume: /opt/evernote-mcp:/data

• Port: 8080

Disclaimer

This is an unofficial client based on reverse engineering the Evernote web application. It is not affiliated with, endorsed by, or supported by Evernote Corporation. Use at your own risk. The internal API may change without notice.