> ## Documentation Index
> Fetch the complete documentation index at: https://to11.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Versions

> The Chat Prompts list, statuses, immutable versioning, the latest label, and the permissions that gate each action.

Authoring builds a draft; this page covers what happens after: freezing a draft into an immutable version, which moves the prompt's one label, `latest`, so your application serves it.

## The Chat Prompts list

Open a project and select **Chat Prompts**. The list shows every chat prompt with its status, a summary tile counting the prompts with how many are published and archived, and controls to search, filter by status, and sort. Skills have their own **Skills** page; see [Skills](https://to11.ai/docs/skills/concepts.md).

A prompt's status is one of:

| Status | Meaning |
| - | - |
| **Active** | Has a published version; `latest` points at the newest one. |
| **Draft** | Has unpublished edits; no published version yet, or a new draft in progress. |
| **Archived** | Hidden from the default list view; restorable. |

### Creating a prompt

From the Chat Prompts list, choose **New prompt**. The wizard asks for the prompt's details, then its content:

* **Display name**: the human-readable name (at least 2 characters), for example `Customer support router`.
* **Slug**: the lowercase, hyphenated identifier (letters, numbers, and hyphens; at least 2 characters), derived from the name until you edit it. Your application renders by this slug.

Name and slug must be unique within the project. Creating a prompt needs a **developer**, **admin**, or **owner** role. **Save** creates the prompt and its first version. See [Authoring](https://to11.ai/docs/platform/prompts/authoring.md).

## Versioning and publishing

Editing always happens on a **draft**. Publishing freezes it:

1. Choose **Publish** in the editor.
2. The confirmation explains the effect: the current version `vN` becomes an immutable, read-only version, and a fresh `vN+1` draft is started from the same blocks, so you can keep editing without changing what's published.
3. Confirm to publish.

A published version can never be edited; that's what makes a version reproducible. To change a published prompt, edit the new draft and publish again. Publishing an unchanged draft is a no-op; there's nothing new to freeze.

Publishing needs a **developer**, **admin**, or **owner** role. The version rail beside the editor lists the newest 50 versions and offers **Load more** for older ones. The other editor tabs work with this history:

* **Versions**: the catalog of every version with its status. Its table pages over the versions loaded so far and offers **Load more** while older versions remain.
* **Diff**: a block-by-block comparison between two published versions, defaulting to the latest against the previous one.

## The `latest` label

A prompt with at least one version carries one label, `latest`, which always points at its newest version. Publishing a version is what moves it; nothing else does. There is no environment to promote a version to and no release to roll back: `latest` only moves forward, so to serve older content again you publish it as a new version. The **Versions** tab and the version rail mark the newest version with a `latest` chip.

Your application renders `latest` by default, so publishing a version changes the served prompt with no application redeploy. See [Rendering prompts](https://to11.ai/docs/platform/prompts/rendering.md#resolution-and-fallbacks).

### Storing a skill

Storing a **[skill](https://to11.ai/docs/skills/concepts.md)** version is the same act. The newest stored version is what `latest` points at, and every machine whose configuration follows `latest` receives it on its next sync; a machine pinned to a version keeps that version until the pin is changed. See [Publishing skills](https://to11.ai/docs/skills/publishing-skills.md).

## Maintaining prompts

From each row's actions menu in the Chat Prompts list:

* **Edit** opens the prompt in the editor.
* **Duplicate** copies the prompt and its latest version into a new draft, named `<name> (copy)` with a `-copy` slug.
* **Archive** hides the prompt from the default view without destroying it.
* **Restore** brings an archived prompt back to active.

You can also change a prompt's **status** and **tags** at any time, and select multiple prompts to act on them together from the bulk bar: set status, archive, restore, or duplicate the selection in one action.

> [!NOTE]
>
> The dashboard archives prompts rather than deleting them, so a prompt's history stays intact and reversible. Prompts are removed permanently only when their project is deleted.

## Permissions

Members don't hold permissions individually; they hold a **[role](https://to11.ai/docs/platform/roles-and-permissions.md)**, and the role carries the permissions. Prompts are project-scoped, so the role that applies is the one a member holds on the project. A workspace or organization role carries the same access down to every project beneath it.

| Role | Prompt access |
| - | - |
| **Owner** / **Admin** / **Developer** | Everything: read, create, edit, publish versions, change status and tags, and archive. |
| **Viewer** | Read-only. Browse prompts and versions; the editor opens read-only. This is the read-only project role. |
| **Member** | No prompt access. This is the base organization and workspace role. |

Each action maps to a permission, enforced server-side:

| Permission | Allows | Roles that hold it |
| - | - | - |
| `prompt:read` | Browse prompts and versions, and open a prompt read-only. | Viewer, Developer, Admin, Owner |
| `prompt:create` | Create prompts, edit blocks, variables, and tools, publish versions, and import a [skill](https://to11.ai/docs/skills/concepts.md) from a folder. | Developer, Admin, Owner |
| `prompt:update` | Change a prompt's status and tags. | Developer, Admin, Owner |
| `prompt:archive` | Archive and restore prompts. | Developer, Admin, Owner |

See [Roles & permissions](https://to11.ai/docs/platform/roles-and-permissions.md) for the full role model.

## Next steps

> [card] [Rendering prompts](https://to11.ai/docs/platform/prompts/rendering.md)
>
> Render a prompt's newest version from your application.

> [card] [Authoring prompts](https://to11.ai/docs/platform/prompts/authoring.md)
>
> Build the blocks, variables, tools, and config a version holds.

> [card] [Publishing skills](https://to11.ai/docs/skills/publishing-skills.md)
>
> Store a skill version from a folder; followers of `latest` receive it.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.