mcp-evernote: v2.0.0

2.0.0 (2026-08-12)

⚠ BREAKING CHANGES

• get_resource now defaults to text extraction, not binary download; pass as:"binary" (or deprecated includeData:true) for base64. evernote_get_resource_text, evernote_get_resource_recognition, and evernote_list_note_resources are removed from the default surface (still callable as deprecated aliases; EVERNOTE_LEGACY_TOOLS=true re-lists them).
• notes: minimum supported Node.js is now 20.16.0 (was 18.18.0), required by the pdf-parse dependency.

Features

• api: rate-limit aware NoteStore transport (#58) (e58d66e)
• consolidate MCP tool surface (27 → 15) behind backward-compatible aliases (#62) (d5a04fa)
• inject live notebook names into tool descriptions and add graceful fallback (#45) (9d0a814)
• notes: batch get_note, format projection, and paged search content (#59) (1b9a974)
• notes: cache notebooks/tags, enrich get_note, surface rateLimitDuration (#54) (3a0272e)
• notes: extract OCR text from image attachments (#55) (7452a51)
• notes: extract PDF attachment text via lazy-loaded pdf-parse (#53) (00cdd0c)
• notes: USN-keyed note body cache (#60) (2c79e2b)

Bug Fixes

• declare the disjoint Node range pdf-parse actually supports (#71) (4976b71)
• exit orphaned stdio MCP servers on parent death (#67) (a11e33b)
• polling: use filtered Evernote sync chunks (#37) (46c4ec7)
• resolve noteStoreUrl via getUserUrls() instead of non-existent getNoteStoreUrl() (#42) (c696ecb)
• stdio: keep Evernote Thrift errors off MCP stdout (#64) (c3aeec9)