---
url: /docs/kitfile/format.md
description: >-
  Dive deep into the Kitfile format. Learn the syntax, structure, and key
  components for defining ModelKits.
---

# Kitfile AI/ML Packaging Manifest Format Reference

The Kitfile manifest for AI/ML is a YAML file designed to encapsulate all the necessary information about the package, including code, datasets, model, and their metadata. This reference documentation outlines the structure and specifications of the manifest format.

## Overview

The manifest is structured into several key sections: `manifestVersion`, `package`, `code`, `datasets`, `docs`, `prompts`, `mcpServers`, and `model`. Each section serves a specific purpose in describing the AI/ML package components and requirements.

### `manifestVersion`

* **Description**: Specifies the manifest format version.
* **Type**: String
* **Example**: `1.0.0`

### `package`

This section provides general information about the AI/ML project.

#### `name`

* **Description**: The name of the AI/ML project.
* **Type**: String

#### `version`

* **Description**: The current version of the project.
* **Type**: String
* **Example**: `1.2.3`

#### `description`

* **Description**: A brief overview of the project's purpose and capabilities.
* **Type**: String

#### `authors`

* **Description**: A list of individuals or entities that have contributed to the project.
* **Type**: Array of Strings

### `code`

* **Description**: Information about the source code.
* **Type**: Object Array
  * `path`: Location of the source code files or directory relative to the context
  * `description`: Description of what the code does.
  * `license`: SPDX license identifier for the code.

### `datasets`

* **Description**: Information about the datasets used.
* **Type**: Object Array
  * `name`: Name of the dataset.
  * `path`: Location of the dataset file or directory relative to the context.
  * `description`: Overview of the dataset.
  * `license`: SPDX license identifier for the dataset.

### `docs`

* **Description**: Information about included documentation for the model
* **Type**: Object Array
  * `description`: Description of the documentation
  * `path`: Location of the documentation relative to the context

### `prompts`

* **Description**: Information about prompt files intended for use with LLMs
* **Type**: Object Array
  * `path`: Location of the prompt file or directory relative to the context
  * `description`: Description of the prompt

### `mcpServers`

* **Description**: Information about MCP servers packaged as MCP Bundle (`.mcpb`) archives. Each entry's path must point to a single `.mcpb` file: a ZIP archive containing a local MCP server and a `manifest.json` at its root. Each bundle is stored verbatim as its own raw, uncompressed layer (media type `application/vnd.kitops.modelkit.mcpb.v1.raw`), so the layer digest covers the exact bundle bytes and unpacking yields a byte-identical archive. The bundle's `manifest.json` remains the source of truth for the server's configuration; none of its fields are duplicated in the Kitfile.
* **Type**: Object Array
  * `name`: Name for the MCP server. Required, and must be unique within `mcpServers`.
  * `path`: Location of the `.mcpb` file relative to the context.
  * `description`: Description of the MCP server.

### `model`

* **Description**: Details of the trained models included in the package.
* **Type**: Object
  * `name`: Name of the model
  * `path`: Location of the model file or directory relative to the context
  * `framework`: AI/ML framework
  * `version`: Version of the model
  * `description`: Overview of the model
  * `license`: SPDX license identifier for the dataset.
  * `parts`: List of related files for the model (e.g. LoRA weights)
    * `name`: Identifier for the part
    * `path`: Location of the file or a directory relative to the context
    * `type`: The type of the part (e.g. LoRA weights)
  * `parameters`: An arbitrary section of yaml that can be used to store any additional data that may be relevant to the current model, with a few caveats. Only a json-compatible subset of yaml is supported. Strings will be serialized without flow parameters. Numbers will be converted to decimal representations (0xFF -> 255, 1.2e+3 -> 1200). Maps will be sorted alphabetically by key.

## Example

```yaml
manifestVersion: 1.0
package:
  name: AIProjectName
  version: 1.2.3
  description: >-
    A brief description of the AI/ML project.
  authors: [Author Name, Contributor Name]
code:
  - path: src/
    description: Source code for the AI models.
    license: Apache-2.0
datasets:
  - name: DatasetName
    path: data/dataset.csv
    description: Description of the dataset.
    license: CC-BY-4.0
prompts:
  - path: system.prompt.md
    description: System prompt for the LLM.
mcpServers:
  - name: filesystem
    path: filesystem.mcpb
    description: MCP server providing filesystem access.
model:
    name: ModelName
    path: models/model.h5
    framework: TensorFlow
    version: 1.0
    description: Model description.
    license: Apache-2.0
```