Inkdrop MCP Server

A Model Context Protocol server for the Inkdrop Local HTTP Server API.

Installation

Read the docs here.

Components

Tools

1. read-note: Retrieve the complete contents of the note by its ID from the database.
   • Required inputs:
     • noteId: The ID of the note to retrieve. It can be found as _id in the note docs. It always starts with note:.
   • Optional inputs:
     • includeTemplate: When true and the note has a sourceTemplateId, also fetch that template and embed its body (fill-out instructions and examples) as an extra text block. Default: false.
   • Note: If the note was created from a template, the response includes a resource link to its source template, whose body holds the fill-out instructions and examples.
2. read-backlinks: Retrieve backlinks for a note — the notes that link to it.
   • Required inputs:
     • noteId: The ID of the note to find backlinks for. It always starts with note:.
   • Note: Each result is returned as a resource link (inkdrop://note/<id>) pointing to the referring note. Backlinks are found by searching all note bodies for the note's link URI.
3. search-notes: List all notes that contain a given keyword.
   • Required inputs:
     • keyword: Keyword to search for.
   • Note: Results include truncated note bodies (200 characters). Use read-note to get full content.
   • Supports advanced search qualifiers like book:, tag:, status:, title:, etc.
4. list-notes: List all notes with specified conditions.
   • Required inputs:
     • bookId: The notebook ID. It always starts with 'book:'.
   • Optional inputs:
     • tagIds: An array of tag IDs to filter. Each starts with 'tag:'.
     • keyword: Keyword to filter notes.
     • sort: Sort field (updatedAt, createdAt, or title). Default: updatedAt.
     • descending: Reverse the order of output. Default: true.
   • Note: Results include truncated note bodies (200 characters). Use read-note to get full content.
5. create-note: Create a new note in the database.
   • Required inputs:
     • bookId: The notebook ID. Must start with 'book:' or be 'trash'.
     • title: The note title.
     • body: The content of the note in Markdown.
   • Optional inputs:
     • status: The note status (none, active, onHold, completed, dropped).
     • tags: An array of tag IDs to assign to the note. Each must start with 'tag:'.
6. update-note: Update an existing note in the database. Only the fields you provide will be updated; omitted fields remain unchanged.
   • Required inputs:
     • _id: The note ID. Must start with 'note:'.
     • _rev: The revision ID (CouchDB MVCC-token).
   • Optional inputs:
     • bookId: The notebook ID. Must start with 'book:' or be 'trash'.
     • title: The note title.
     • body: The content of the note in Markdown.
     • status: The note status (none, active, onHold, completed, dropped).
     • tags: An array of tag IDs to assign to the note. Each must start with 'tag:'.
7. patch-note: Update the body of an existing note by performing an exact string replacement. More efficient than update-note for small edits to large notes as it saves tokens. You must first read the note with read-note to get the current body.
   • Required inputs:
     • _id: The note ID. Must start with 'note:'.
     • _rev: The revision ID (CouchDB MVCC-token).
     • old_string: The exact text to find in the note body. Must match exactly one occurrence. Include enough surrounding context to ensure a unique match.
     • new_string: The text to replace old_string with. Use an empty string to delete the matched text.
8. list-notebooks: Retrieve a list of all notebooks.
9. read-book: Retrieve a single notebook by its ID.
   • Required inputs:
     • bookId: The notebook ID. Must start with 'book:'.
10. create-book: Create a new notebook in the database.
    • Required inputs:
      • name: The notebook name.
    • Optional inputs:
      • parentBookId: The ID of the parent notebook. Must start with 'book:'. Omit it to create the notebook at the root level.
11. update-book: Update an existing notebook in the database. Only the fields you provide will be updated; omitted fields remain unchanged, so you don't need to read the notebook first.
    • Required inputs:
      • _id: The notebook ID. Must start with 'book:'.
    • Optional inputs:
      • _rev: The revision ID (CouchDB MVCC-token). Only needed as an optimistic-concurrency guard — pass it to make the update fail on a conflicting concurrent edit.
      • name: The notebook name.
      • parentBookId: The ID of the parent notebook. Pass null to move the notebook to the root level.
12. list-tags: Retrieve a list of all tags.
13. read-tag: Retrieve a single tag by its ID.
    • Required inputs:
      • tagId: The tag ID. Must start with 'tag:'.
14. create-tag: Create a new tag in the database.
    • Required inputs:
      • name: The name of the tag.
    • Optional inputs:
      • color: The color type of the tag (default, red, orange, yellow, olive, green, teal, blue, violet, purple, pink, brown, grey, black). Default: default.
15. update-tag: Update an existing tag in the database. Only the fields you provide will be updated; omitted fields remain unchanged, so you don't need to read the tag first.
    • Required inputs:
      • _id: The tag ID. Must start with 'tag:'.
    • Optional inputs:
      • _rev: The revision ID (CouchDB MVCC-token). Only needed as an optimistic-concurrency guard — pass it to make the update fail on a conflicting concurrent edit.
      • name: The name of the tag.
      • color: The color type of the tag. Omit it to keep the tag's current color.
16. create-file: Create a new attachment file in the database from a local image file or from base64 image data. Returns the Markdown to embed the attachment in a note body.
    • Required inputs: exactly one of
      • filePath: Absolute path to a local image file to attach. Preferred — it keeps the image bytes out of the conversation.
      • data: Base64-encoded image data, without a data: URI prefix.
    • Optional inputs:
      • name: The file name. Defaults to the file name of filePath, or image.<ext> when data is given.
      • contentType: The MIME type (image/png, image/jpeg, image/jpg, image/svg+xml, image/gif, image/heic, image/heif). Inferred from the filePath extension when omitted, so it is required with data.
      • publicIn: An array of note IDs where the attachment is included. Each must start with 'note:'.
    • Note: Attachments are limited to 10MB.
17. read-file: Retrieve an attachment file by its ID. Returns the file metadata, and the image itself when it is a PNG, JPEG or GIF.
    • Required inputs:
      • fileId: The attachment ID. Must start with 'file:'.
    • Optional inputs:
      • outputPath: Absolute path to write the attachment to. When given, the file is saved there and its bytes are not returned inline. Required to get at SVG, HEIC and HEIF attachments, which cannot be displayed inline.
      • includeImage: Whether to return the image alongside the metadata. Default: true. Set to false to fetch only the metadata of a large attachment.

Debugging

Since MCP servers run over stdio, debugging can be challenging. For the best debugging experience, we strongly recommend using the MCP Inspector.

You can launch the MCP Inspector via npm with this command:

npx @modelcontextprotocol/inspector "./dist/index.js"

Be sure that environment variables are properly configured.

Upon launching, the Inspector will display a URL that you can access in your browser to begin debugging.

You can also watch the server logs with this command:

tail -n 20 -f ~/Library/Logs/Claude/mcp-server-inkdrop.log