Agents don't browse. They read. Your knowledge should be built for that.
Agents need to read books, not query catalogs — knowledge is a literacy problem, not a database problem. See how a curated library built for agents compares to the tools you already use.
CandleKeep vs Obsidian
| Dimension | Obsidian | CandleKeep |
|---|---|---|
| Designed reader | You (agents via plugins) | Agents read & write back |
| Knowledge | Your own notes, unvetted | Expert-curated, versioned books |
| Knowledge capture | You file every note by hand | Seamless — captured in your sessions |
| Reach | Local vault, one machine | Read from anywhere, any agent |
| Citations | Note titles, keyword search | Page-level, semantic |
| Shared library | No — one private vault | Shareable, with a marketplace |
Obsidian speaks MCP, so an agent can read your vault — but you still write and file every note by hand. CandleKeep rides inside your agents' sessions and turns the decisions made there into living books, seamlessly, so capture stops being your job.
Get Started FreeMost of these tools can already feed an agent — Obsidian, Notion, and Glean all speak MCP now. The difference isn't access; it's where the knowledge comes from. They infer it from raw data — everything you've written or connected, ranked by a machine. CandleKeep is curated by the experts who actually know: a human decides what's true, and an agent reads it cited by page. And because CandleKeep rides inside your agents' sessions, it sees the decisions as they're made — so capturing an expert's knowledge into a living book is seamless, not a second job. Curation you can trust, capture you don't have to babysit.