---
title: "Manuscripts"
description: "Use when you want agents to record what they learn in a living note, or to control when they write to it."
url: https://docs.heytrove.ai/concepts/manuscripts
updated: 2026-10-06
---

# Manuscripts

> Load this page when: you are creating a manuscript, changing when agents write to one, or a manuscript is not receiving notes

## For agents

- Check manuscripts only at task completion, using the `log when:` line from session context.
- Default flow is a proposal box, then wait for "yes" or "skip". Write directly only when the manuscript is marked `[auto-update]`.
- Undo an edit with `trove items history <id>` and `trove items restore <id> --version <n>`.

A manuscript is a living note that agents add to as they work. You describe when a note belongs in it, and your agent offers to record what it learned at the end of a task.

## What to tell your agent

| Task | What to tell your agent |
| --- | --- |
| See your manuscripts | "List my Trove manuscripts." |
| See only active ones | "List my active Trove manuscripts." |
| Show one | "Show my Architecture decisions manuscript." |
| Create one | "Create a Trove manuscript called Architecture decisions. Log when we make a database schema decision." |
| Turn automatic writes on | "Turn on auto-update for my Architecture decisions manuscript." |
| Pause or resume | "Pause my Architecture decisions manuscript." |
| Archive | "Archive my Architecture decisions manuscript." |

## How an agent decides where a note belongs

Each manuscript has a field called "When should agents read it". It shows up as `log when:` in the session context. If you leave it blank, Trove uses the first acceptance criterion instead.

Your agent checks it only when a task is complete, never mid-task. A sharp line such as "log when we make a database schema decision" works better than a vague one.

## Ask first or write directly

By default, your agent shows a proposal box with the note it wants to add. You answer "yes" or "skip". Nothing is written until you say yes.

A manuscript with `[auto-update]` turned on is written directly, followed by a one-line 📝 notice. This is how an LLM wiki works. The seeded wiki also offers "always", which skips the proposal for the rest of the session.

## Create and manage

Create a manuscript in the web app at [Manuscripts](https://heytrove.ai/manuscripts), or ask your agent to create one. Personal includes a limited number of manuscripts. See [Plans and limits](https://docs.heytrove.ai/reference/plans-and-limits.md).

## Privacy

Manuscripts are private unless you share or publish them. Every edit is saved as a version. To review the edits, ask your agent: "Show the version history of my Architecture decisions manuscript." To undo an edit, ask your agent to restore the previous version.

## Best practice

Send every durable learning to a manuscript first. Each manuscript sets its own approval rule, and there is no global rule. Many people keep approval on while a [book](https://docs.heytrove.ai/concepts/library-and-books.md) is in progress. Later, move the best entries into chapters. See [Manuscripts that stay useful](https://docs.heytrove.ai/best-practices/manuscripts-that-stay-useful.md).

## Direct CLI

Your agent runs these through the `trove` command. You can run them yourself in a terminal, but asking your agent is the recommended way: it picks the right flags, waits for processing and checks the result.

| Task | Command |
| --- | --- |
| List manuscripts | `trove ms list` (add `--active` for active ones) |
| Show one | `trove ms show <id>` |
| Create | `trove ms create "<title>" --description "<text>"` |
| Turn automatic writes on | `trove ms update <id> --auto-update` |
| Pause or resume | `trove ms update <id> --pause` (or `--resume`) |
| Archive | `trove ms archive <id>` |
| Review and undo edits | `trove items history <id>`, then `trove items restore <id> --version <n>` |

See the [CLI reference](https://docs.heytrove.ai/reference/cli.md).

## Related

- [Library and books](https://docs.heytrove.ai/concepts/library-and-books.md) explains versions and history.
- [Log what you learn](https://docs.heytrove.ai/usage/log-what-you-learn.md) shows the steps.
- [Manuscripts that stay useful](https://docs.heytrove.ai/best-practices/manuscripts-that-stay-useful.md) covers approval and chapters.
- [Book writer](https://docs.heytrove.ai/subagents/book-writer.md) is the subagent that edits Markdown books.