Standalone PKM app — Cursor SDK as harness layer (2026-05-06, 14:21)
Short follow-up to 2026-05-06-Standalone PKM app - cons of Obsidian setup, sync, UI focus (~18 min later). Notes Cursor SDK as a possible harness/AI layer, but flags that it leaves the sync and cloud routines problems untouched — those are the two challenges that survive the SDK choice.
Re obsidian app
I guess, one viable idea would be to use cursor sdk for the harness/ai layer? Still doesnt solve the sync and routines issue tho.
Related
- 2026-05-06-Standalone PKM app - cons of Obsidian setup, sync, UI focus — parent memo (challenges #2 sync, #6 cloud routines are the two this SDK choice doesn’t solve).
- 2026-05-06-Standalone PKM app - yjs CRDT sync stack — same-day follow-up (15:06) that does address the sync side, with Claude Code SDK rather than Cursor SDK as the harness.
- OpenClaw for PKM — parent concept page.
- Agent-Native Software Architecture — harness/SDK choice sits at the orchestrator layer of this pattern.
- 2026-05-19-Distill app review - AI-native PKM with local-only limits — 13 days later: Distill answers the harness layer differently again — built-in model + BYO endpoint with named size tiers (small/medium/large), rather than picking a Cursor/Claude SDK as harness.