An LLM wiki is a set of linked pages that an AI agent reads, updates and cites across sessions. In Trove it is a Manuscript: pages with a table of contents and version history that your agent keeps current, so decisions and lessons outlast the chat.
Install Trove
An LLM wiki turns what you learn into connected pages. In Trove, it lives as a Manuscript: a source your agent can browse, update and cite alongside the rest of your library.
You find a tricky edge case. Make a decision. Discover why the obvious approach failed. The next session needs what you learned, but the conversation alone is a hard place to find it.
A place for the reasoning, not just the result.
Keep what will be useful again. Give it a clear home. Let your next session start with what you already worked to uncover.
A Manuscript uses the same building blocks as the rest of your library. You can follow what was written and see how it changes.
An index helps your agent find the relevant section before opening a page.
Concepts, decisions and examples can point to one another.
Inspect earlier versions and restore a page when an edit needs undoing.
Read the page behind an answer, with the surrounding pages still intact.
Follow the install guide for the tool you use. Your library becomes available in your existing workflow.
Decide what belongs there: project conventions, decisions, or lessons from the work. Clear instructions make a more useful book.
Ask your agent to record useful knowledge. Review what it writes and have it consult the book when it matters again.
From the library
A Manuscript is one part of an agent workflow. Go deeper into how agents share knowledge and work together with a practical guide from the library.
From the library
A companion for thinking about how knowledge moves between agents, tasks and sessions.
15 pages
An LLM wiki is a collection of connected pages that an AI agent writes and maintains: concepts, decisions, an index and a changelog. In Trove, this takes the form of a Manuscript your agent can read and update.
A Manuscript keeps useful knowledge in your Trove library beyond a single chat. Your agent can turn what you learn into written pages, then return to those pages when they matter again. You can inspect the content and its version history.
It browses the table of contents, opens relevant pages and cites what it reads. The index and links between pages help it navigate the material without treating the whole book as one long prompt.
What gets written depends on the book's instructions and your agent workflow. Review-first and automatic workflows serve different needs. Set clear guidance about what belongs in the book, and inspect the result.
A Manuscript is an evolving document you can inspect and edit. Trove gives the agent access to its contents and pages. The focus is a maintained source of knowledge, with citations and history, rather than a set of isolated search results.
For the next time it matters
Bring your sources and what you learn into one library your agent can return to.
Get started with Trove