feat: Extract and expose PDF attachment content via MCP

[autojack-bot]

Problem

Currently, the Evernote MCP can detect that a note has PDF attachments but cannot read their contents. Claude reports it can only see that PDFs exist, not what's inside them.

Evernote itself supports full-text search inside PDFs and its native AI can read them — but this capability isn't surfaced through the MCP.

Expected Behavior

When a note has a PDF attachment, the MCP should be able to extract and return the text content of that PDF so Claude (or any MCP client) can reason over it.

Proposed Solution

• When fetching a note with PDF attachments, download the resource binary from the Evernote API
• Extract text content using a PDF parsing library (e.g. pdf-parse, pdfjs-dist)
• Include extracted text in the note response alongside or inside the attachment metadata

Acceptance Criteria

• PDF text content is extracted and returned when reading a note with PDF attachments
• Falls back gracefully if extraction fails (e.g. scanned PDFs with no text layer)
• Existing behavior for notes without PDFs is unchanged
• Works with the existing auth approach (clipper-sso token)

Context

Reported by a community user in the AutoJack Discord. Related to the broader Evernote-as-knowledge-source use case where users want Claude to reason over their full Evernote library including attached documents.

cc @jack-arturo