--- title: "Plan mode and auto mode" description: "Use when you want to change the setting for Trove's automatic library checks, or to understand why your agent did or did not check." url: https://docs.heytrove.ai/concepts/plan-and-auto-mode updated: 2026-10-06 --- # Plan mode and auto mode > Load this page when: your agent consults Trove too often, too rarely, or at the wrong moment ## For agents - Read the mode with `trove config get skill.planMode`. Values are `always`, `smart`, `ask` and `off`. - In plan mode, run Trove once after exploration and before writing the plan. Do not run the enricher. - In auto mode, run at most one librarian per task, before the first edit, and skip renames, typos and one-line fixes. One setting decides when your agent checks your [library](https://docs.heytrove.ai/concepts/library-and-books.md) without being asked. It covers two moments: while planning, and before a non-trivial coding task. Changes apply at the next session start. > **Do not rely on this yet.** We are improving these automatic checks and working to make the [librarian](https://docs.heytrove.ai/subagents/librarian.md) start automatically and reliably. Today, when the answer matters, ask explicitly. For example: "Search my Trove library for how we handle retries." That starts the librarian, and the librarian starts the [readers](https://docs.heytrove.ai/subagents/item-reader.md). ## Choose when Trove checks your library Open Settings → Answers → Library activation. You can also ask your agent, for example: "Set Trove library activation to When it likely helps." | Label | Value | Behavior | | --- | --- | --- | | On relevant tasks | `always` | Checks on relevant tasks. This is the default. | | When it likely helps | `smart` | Checks only when it likely helps. | | Ask me first | `ask` | Asks before each check. | | Never automatically | `off` | Never checks on its own. Asking explicitly still works. | ## Plan mode Trove runs once, after your agent has explored the code and before it writes the plan. The librarian runs in the background, then one item reader runs per angle. The [book enricher](https://docs.heytrove.ai/subagents/book-enricher.md) does not run in plan mode. ## Auto mode Auto mode works in Claude Code through the session hook, and it is active only when the setting is not `off`. It runs before any non-trivial coding or design task, before the first edit, and never at the end of a session. It skips renames, typos, one-line fixes and questions with no code. Your agent names the sub-domain of the task, then checks the project notebook. A hit means one reader on that book. A miss or a decline within three sessions means it proceeds without the library, unless a new [marketplace book](https://docs.heytrove.ai/concepts/marketplace.md) arrived. Otherwise one librarian runs, and never more than one per task. ## The project notebook The notebook is a file at `~/.trove/projects//explored.json`. It records which sub-domains your agent has already looked up in a project. It lives under your home folder and is never written into the repo. ## Other agents Cursor, GitHub Copilot, Gemini CLI and Grok have no session hook. They read `./.trove/shelf.json` themselves. See [Shelves](https://docs.heytrove.ai/concepts/shelves.md) for how to share it. ## Common mistakes - Agent did not check: ask it explicitly to search your Trove library. Then ask your agent: "Check that Trove Auto mode is on." If the setting is `off`, ask your agent to set it to `always`, then start a new session. - Setting changed but nothing happened: start a new session. The setting applies at session start. ## Best practice Ask explicitly when the answer matters: "Search my Trove library for how we handle retries." Use plain words. The librarian decides how many books to use and how deep to read, so do not set a book limit yourself. The setting is a helpful extra, not a guarantee. See [Get answers you can trust](https://docs.heytrove.ai/best-practices/answers-you-can-trust.md) and [Plan a feature with your library](https://docs.heytrove.ai/usage/plan-a-feature.md). Want your agent to follow these practices? Ask it: "Read How to Use Trove in my Trove library and follow it." ## 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. - Read the setting: `trove config get skill.planMode` - Change the setting: `trove config set skill.planMode ` (`always`, `smart`, `ask` or `off`) See the [CLI reference](https://docs.heytrove.ai/reference/cli.md). ## Related - [How Trove works](https://docs.heytrove.ai/how-trove-works.md) shows the full pipeline. - [Get answers you can trust](https://docs.heytrove.ai/best-practices/answers-you-can-trust.md) covers asking in plain words. - [Plan a feature with your library](https://docs.heytrove.ai/usage/plan-a-feature.md) shows the steps. - [Librarian](https://docs.heytrove.ai/subagents/librarian.md) explains the reading list.