> For the complete documentation index, see [llms.txt](https://docs.flora.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.flora.ai/mcp/tools/models-and-loras.md).

# Model and LoRA tools

Reference for the FLORA MCP tools that list generation models and their parameters, and train and check LoRA styles.

These [FLORA MCP](https://flora.ai/mcp) tools list the models your account can generate with and train LoRA styles from your own images. [`flora_list_models`](#flora_list_models) returns each model's `model_id`, its estimated cost in US dollars and, for one model, the exact `params` names that [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations) accepts. To browse models by name, see the [image](https://docs.flora.ai/models/image-models.md), [video](https://docs.flora.ai/models/video-models.md), [audio](https://docs.flora.ai/models/audio-models.md), [text](https://docs.flora.ai/models/text-models.md) and [3D](https://docs.flora.ai/models/3d-models.md) model catalogs.

| Tool                                                    | What it does                                                                                             | Labels                                   |
| ------------------------------------------------------- | -------------------------------------------------------------------------------------------------------- | ---------------------------------------- |
| [`flora_list_models`](#flora_list_models)               | List the generation models this account can use, or read one model's full parameter list.                | read-only                                |
| [`flora_list_lora_trainers`](#flora_list_lora_trainers) | List LoRA trainer families with their training parameters, defaults, image limits and compatible models. | read-only                                |
| [`flora_train_lora`](#flora_train_lora)                 | Train and save a LoRA style from your own images, the same way the FLORA web Train dialog does.          | writes, destructive, spends money, async |
| [`flora_get_lora`](#flora_get_lora)                     | Check whether a saved LoRA style is ready, with its base model and trigger details.                      | read-only                                |

## flora\_list\_models <a href="#flora_list_models" id="flora_list_models"></a>

List the generation models this account can use, or read one model's full parameter list.

**Labels:** read-only

**Shown in clients as:** List models

**Use when:** Before generating: to pick a `model_id`, see estimated cost and time, or get one model's exact `params` names (such as `image_url` or `image_urls`).

**Don't use when:** The default model for the job is fine: [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations) picks one when you leave out `model`.

{% hint style="warning" %}
The default page isn't the full catalog. Without `limit` you get 25 models (the argument's description says 20; the server returns 25), and `limit` stops at 100. There's no cursor, so check `total` and `truncated`, then narrow with `type` and `query`. `model_id` looks up one model across the whole catalog.
{% endhint %}

**Arguments**

| Argument   | Type   | Required | Notes                                                                           |
| ---------- | ------ | -------- | ------------------------------------------------------------------------------- |
| `type`     | string | no       | Filter by output modality. One of `image`, `video`, `audio`, `text`, `model3d`. |
| `model_id` | string | no       | Return this one model in full, including its parameters.                        |
| `query`    | string | no       | Filter by model id or name (case-insensitive; '.', '\_' and spaces match '-').  |
| `limit`    | number | no       | Max items to return (default 25).                                               |

**Returns:** Featured models first under `featured`, grouped by type, then the rest under `models`; a featured model isn't repeated in `models`, and `limit` caps both together. Each has a `model_id`, name, capabilities (such as `text-to-image`), `estimated_cost_usd` and `estimated_seconds`. `total` counts every model that matched your filters, and `truncated` appears when the page shows fewer. With `model_id`, that model's full `params` list with options and defaults.

**Cost:** Free. This tool doesn't spend money.

**Usually comes before:** [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations)

**Example prompt:** "Which FLORA video models can animate a photo?"

**Example call:**

```json
{
  "type": "video",
  "query": "seedance",
  "limit": 10
}
```

<details>

<summary>Full description from the server</summary>

{% code overflow="wrap" %}

```
List the generation models available to this account. When FLORA curates featured models they come first under `featured`, grouped by type (featured models are not repeated in `models`, and `limit` caps the total across both). Returns a summary per model with its capabilities, such as text-to-image or image-to-video; pass model_id to get one model's full parameter list for flora_create_generations.
```

{% endcode %}

</details>

## flora\_list\_lora\_trainers <a href="#flora_list_lora_trainers" id="flora_list_lora_trainers"></a>

List LoRA trainer families with their training parameters, defaults, image limits and compatible models.

**Labels:** read-only

**Shown in clients as:** List LoRA trainers

**Use when:** Before [`flora_train_lora`](#flora_train_lora), so `base_model` and `trainer_params` come from the live catalog.

**Don't use when:** You want to generate with a LoRA that's already trained: check it with [`flora_get_lora`](#flora_get_lora), then pass `params.lora_id` to [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations).

**Arguments**

| Argument       | Type   | Required | Notes                                         |
| -------------- | ------ | -------- | --------------------------------------------- |
| `workspace_id` | string | yes      | Workspace for training and inference (ws\_…). |

**Returns:** Trainer families, each with its training parameters and defaults, image count limits and formats, and the inference model IDs that accept the result, with strength defaults and ranges.

**Cost:** Free. This tool doesn't spend money.

**Usually comes after:** [`flora_list_workspaces`](https://docs.flora.ai/mcp/tools/workspaces-and-projects.md#flora_list_workspaces)

**Usually comes before:** [`flora_train_lora`](#flora_train_lora)

**Example prompt:** "What LoRA trainers can I use in FLORA?"

**Example call:**

```json
{
  "workspace_id": "ws_123"
}
```

<details>

<summary>Full description from the server</summary>

{% code overflow="wrap" %}

```
Discover available LoRA trainer families, supported training parameters and effective web defaults, image-count limits/formats, and compatible inference model IDs with strength defaults/ranges. Call before flora_train_lora so base_model and trainer_params come from the live catalog. The resulting Style is attached to inference with params.lora_id. Defaults describe a dataset without manual labels; supplying labels selects manual captioning unless an explicit automatic mode conflicts.
```

{% endcode %}

</details>

## flora\_train\_lora <a href="#flora_train_lora" id="flora_train_lora"></a>

Train and save a LoRA style from your own images, the same way the FLORA web Train dialog does.

**Labels:** writes · destructive · spends money · async

**Shown in clients as:** Train LoRA

**Use when:** The user wants a reusable style or subject trained from their images.

**Don't use when:** You'd pass a trainer model to [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations): that skips the saved style. Upload the images with [`flora_create_asset`](https://docs.flora.ai/mcp/tools/assets.md#flora_create_asset) first and pass their FLORA URLs.

**Arguments**

| Argument         | Type                             | Required | Notes                                                                                                                                                                                                                                                                                                                                                                                                               |
| ---------------- | -------------------------------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `workspace_id`   | string                           | yes      | Workspace to authorize and bill (ws\_…).                                                                                                                                                                                                                                                                                                                                                                            |
| `project_id`     | string                           | yes      | Project that owns the training run (prj\_…). Use project\_id from flora\_list\_projects or flora\_create\_project. An API-issued canvas URL's /projects/\<id> segment becomes prj\_\<id>.                                                                                                                                                                                                                           |
| `name`           | string                           | yes      | Name for the saved Style. Can't be empty.                                                                                                                                                                                                                                                                                                                                                                           |
| `base_model`     | string                           | yes      | Catalog LoRA family key, such as flux-2 or krea-2. Can't be empty.                                                                                                                                                                                                                                                                                                                                                  |
| `image_urls`     | array of string (uri)            | yes      | Ordered training-image URLs uploaded to FLORA; trainer-specific image-count limits also apply. 1 to 100 items.                                                                                                                                                                                                                                                                                                      |
| `captions`       | array of (string or null)        | no       | Optional flat array of per-image captions (string or null), aligned index-by-index with image\_urls — exactly one entry per image. Null or blank entries use the shared fallback. Omit to keep web captioning defaults. Up to 100 items. Each item: up to 1024 characters.                                                                                                                                          |
| `trainer_params` | map of string, number or boolean | no       | Optional flat map of trainer parameters: scalar values only (string, number or boolean), keyed by the exact parameter names flora\_list\_lora\_trainers declares — default\_caption and trigger\_phrase are flat trainer\_params keys, not nested objects or top-level args. Omitted values use web Train defaults, including captioning mode; explicit false, zero and empty strings are preserved for validation. |
| `client_token`   | string                           | yes      | Persistent lowercase UUID for this logical training request. Reuse the same token and inputs after a lost response. Matches `^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$`.                                                                                                                                                                                                                       |

**Returns:** `style_id` and `run_id`. Poll the run with [`flora_list_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_list_generations), check readiness with [`flora_get_lora`](#flora_get_lora), then generate with `params.lora_id`.

**Cost:** Spends money when training starts. There's no quote option, so tell the user training is billed and get approval first. Save one `client_token` and reuse it, with the same inputs, after a lost response so the training isn't billed twice.

**Usually comes after:** [`flora_list_lora_trainers`](#flora_list_lora_trainers), [`flora_create_asset`](https://docs.flora.ai/mcp/tools/assets.md#flora_create_asset)

**Usually comes before:** [`flora_list_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_list_generations), [`flora_get_lora`](#flora_get_lora)

**Example prompt:** "Train a LoRA called Product style from the 20 product photos I uploaded."

**Example call:**

```json
{
  "workspace_id": "ws_123",
  "project_id": "prj_abc",
  "name": "Product style",
  "base_model": "flux-2",
  "image_urls": [
    "https://media.flora.ai/example/product-01.jpg",
    "https://media.flora.ai/example/product-02.jpg"
  ],
  "client_token": "3f2b8c1e-7a4d-4e6b-9c2a-1d5e8f0a7b3c"
}
```

<details>

<summary>Full description from the server</summary>

{% code overflow="wrap" %}

```
Train and save a LoRA using the same image preparation, captions, defaults, billing and Style lifecycle as the FLORA web Train dialog. SPENDS USAGE. Upload source images with flora_create_asset first, then pass their FLORA URLs. Choose the base_model family (for example flux-2 or krea-2). Persist one lowercase UUID client_token before submitting; reuse it with the same inputs if the response is lost. Returns style_id and run_id. Poll flora_list_generations with run_ids: [run_id] and check flora_get_lora readiness, then use params.lora_id with flora_create_generations. Do not train by passing a trainer model to flora_create_generations: that bypasses the saved Style lifecycle.
```

{% endcode %}

</details>

## flora\_get\_lora <a href="#flora_get_lora" id="flora_get_lora"></a>

Check whether a saved LoRA style is ready, with its base model and trigger details.

**Labels:** read-only

**Shown in clients as:** Get LoRA

**Use when:** After training, to see whether the style is ready before generating with it.

**Don't use when:** Training failed and you want to know why: poll the training `run_id` with [`flora_list_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_list_generations).

**Arguments**

| Argument       | Type   | Required | Notes                                                           |
| -------------- | ------ | -------- | --------------------------------------------------------------- |
| `workspace_id` | string | yes      | Workspace in which the Style will be used (ws\_…).              |
| `style_id`     | string | yes      | Saved Style identifier returned by flora\_train\_lora (sty\_…). |

**Returns:** The style's readiness, `base_model` and saved trigger metadata. Missing, deleted and inaccessible styles return `404`. When it's ready, pass its ID as `params.lora_id` (not `params.style_id`) to [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations).

**Cost:** Free. This tool doesn't spend money.

**Usually comes after:** [`flora_train_lora`](#flora_train_lora)

**Usually comes before:** [`flora_create_generations`](https://docs.flora.ai/mcp/tools/generations.md#flora_create_generations)

**Example prompt:** "Is my Product style LoRA ready?"

**Example call:**

```json
{
  "workspace_id": "ws_123",
  "style_id": "sty_123"
}
```

<details>

<summary>Full description from the server</summary>

{% code overflow="wrap" %}

```
Read a saved LoRA Style's readiness, base_model and saved trigger metadata in a workspace. Pass the style_id from flora_train_lora. When ready, pass this ID as params.lora_id (not params.style_id) in an item of flora_create_generations with a compatible model. Missing, deleted and inaccessible Styles return 404; poll the training run_id with flora_list_generations using run_ids to diagnose a failed training.
```

{% endcode %}

</details>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.flora.ai/mcp/tools/models-and-loras.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
