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

# AI Assets

> Catalog and govern large language models, AI agents, Model Context Protocol servers, and AI applications

# AI Assets

AI Assets extend the catalog to the systems that build, run, and govern artificial intelligence workloads. The application programming interface (API) models eight resources across five areas. They cover large language models (LLMs), AI agents, and Model Context Protocol (MCP) servers. They also cover AI applications and prompt templates.

<CardGroup cols={2}>
  <Card title="LLMs" icon="brain" href="/v2.0.x/api-reference/ai-assets/llm">
    Catalog LLM services and the models hosted by them.
  </Card>

  <Card title="AI Agents" icon="robot" href="/v2.0.x/api-reference/ai-assets/ai-agents">
    Model cataloged agents, invoke AI Studio agents, and record executions.
  </Card>

  <Card title="MCP" icon="server" href="/v2.0.x/api-reference/ai-assets/mcp">
    Catalog MCP services and servers, including their tools, resources, and prompts.
  </Card>

  <Card title="AI Applications" icon="window" href="/v2.0.x/api-reference/ai-assets/ai-applications">
    Govern chatbots, agents, and custom AI applications. Cover copilots, assistants, and retrieval-augmented generation (RAG) systems.
  </Card>

  <Card title="Prompt Templates" icon="message-lines" href="/v2.0.x/api-reference/ai-assets/prompt-templates">
    Version and govern reusable prompts, variables, examples, and evaluation metrics.
  </Card>
</CardGroup>

## Resource Model

| Concept        | Catalog Resource | Base Path                  | Relationship                                   |
| -------------- | ---------------- | -------------------------- | ---------------------------------------------- |
| LLM            | LLM Service      | `/v1/services/llmServices` | Parent service for models                      |
| LLM            | LLM Model        | `/v1/llmModels`            | Belongs to an LLM service                      |
| AI Application | AI Application   | `/v1/aiApplications`       | Global catalog entity                          |
| AI Agent       | Agent Execution  | `/v1/agentExecutions`      | Time-series execution record for an agent      |
| MCP            | MCP Service      | `/v1/services/mcpServices` | Parent service for servers                     |
| MCP            | MCP Server       | `/v1/mcpServers`           | Belongs to an MCP service                      |
| MCP            | MCP Execution    | `/v1/mcpExecutions`        | Time-series execution record for an MCP server |
| AI Application | Prompt Template  | `/v1/promptTemplates`      | Reusable prompt referenced by applications     |

Catalog an AI agent as an AI Application whose `applicationType` is `Agent` or `MultiAgent`. `AgentExecution` stores runtime observations; it isn't the agent definition. The separate [AI SDK](/v2.0.x/sdk/ai-sdk) manages executable AI Studio agents.

## SDK Coverage

The v2.0 core software development kits (SDKs) don't expose every AI resource through the same interface. The examples in this section use only confirmed client paths.

| Resource        | Python Core SDK               | Java Core SDK                                  |
| --------------- | ----------------------------- | ---------------------------------------------- |
| LLM Service     | REST only                     | `client.llmServices()`                         |
| LLM Model       | REST only                     | `client.llmModels()`                           |
| AI Application  | REST only                     | `client.aiApplications()` and `AIApplications` |
| Agent Execution | REST only                     | REST only                                      |
| MCP Service     | Generic `OpenMetadata` client | REST only                                      |
| MCP Server      | Generic `OpenMetadata` client | `client.mcpServers()` and `McpServers`         |
| MCP Execution   | REST only                     | REST only                                      |
| Prompt Template | REST only                     | `client.promptTemplates()`                     |

The separate AI SDK supports AI Studio agent management and invocation in Python, TypeScript, Java, and the command-line interface. It doesn't replace the catalog REST resources listed above.

## Common Behavior

LLM services, LLM models, MCP services, MCP servers, AI applications, and prompt templates support standard list, retrieve, create, create-or-update, patch, version, soft-delete, hard-delete, and restore operations. Agent and MCP executions use time-series APIs with create, retrieve, list, and delete operations.

Authenticate every request with a bearer token. For setup details, see [Authentication](/v2.0.x/api-reference/authentication).