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

# AI Hub

> Publish models, agents, and MCP servers to a public hub so developers can discover what is available.

## Overview

The **AI Hub** allows tenant administrators to expose selected **models**, **agents**, and **MCP servers** in a public-facing catalog. This makes it easier for developers to discover what capabilities are available while still enforcing access through virtual keys, teams, and guardrails.

The AI Hub is discovery-focused only. Listing an asset as public does **not** remove authentication or governance requirements.

***

## AI Hub URL

At the top of the page, a **Model Hub URL** is shown.\
This URL points to the public hub view where published assets are visible to developers.

***

## Link Management

The **Link Management** section controls which helpful links appear under **Useful Links** in the public hub.

### Add New Link

* **URL**\
  The destination link.
* **Display Name**\
  A human-friendly label shown in the hub.
* **Add Link**\
  Saves the link and makes it visible.

### Manage Existing Links

A table displays all configured links with options to remove or edit them.

Use this feature to surface:

* Documentation
* SDKs
* Internal runbooks
* Getting-started guides

***

## Hub Tabs

The AI Hub is organized into three tabs:

* **Model Hub**
* **Agent Hub**
* **MCP Hub**

Each tab allows selective publishing of assets.

***

## Model Hub

The **Model Hub** tab lists models that can be made visible in the public catalog.

### Filters & Search

* **Search Models** – filter by model name
* **Provider** – filter by provider
* **Mode** – filter by model mode
* **Features** – filter by supported capabilities

These filters help admins quickly find the models they want to expose.

***

### Model Table

Each row typically includes:

* **Public Model Name**
* **Provider**
* **Mode**
* **Token Information**
* **Feature Indicators**\
  Examples include vision support or function calling.
* **Public**\
  Indicates whether the model is currently published.
* **Details**\
  Opens a detailed view of the model.

***

### Make Models Public

Select **Select Models to Make Public** to open the publishing flow.

#### Step 1: Select Models

* Choose one or more models.
* Filters and search are available.
* A “Select All” option may be shown if multiple models are available.

> Even when a model is public, users must still use a valid virtual key to access it.

#### Step 2: Confirm

* Review selected models.
* Confirm to publish them to the public hub.

***

## Agent Hub

The **Agent Hub** tab controls visibility of agents configured in the tenant.

### Agent Table

The table may include:

* **Agent Name**
* **Description**
* **Version**
* **Protocol**
* **Skills**
* **Capabilities**
* **I/O Modes**
* **Public**
* **Details**

If no agents exist, the table displays an empty state.

***

### Make Agents Public

Select **Select Agents to Make Public** to open the publishing flow.

* Choose agents to include
* Confirm publication

> Public agents still enforce authentication, routing, and guardrail policies.

***

## MCP Hub

The **MCP Hub** tab manages visibility of MCP servers.

### MCP Server Table

Typical columns include:

* **Server Name**
* **Description**
* **URL**
* **Transport**
* **Auth Type**
* **Status**
* **Available Tools**
* **Created By**
* **Public**
* **Details**

If no MCP servers are configured, an empty state is shown.

***

### Make MCP Servers Public

Select **Select MCP Servers to Make Public** to open the publishing flow.

#### Step 1: Select Servers

* Choose MCP servers to expose in the hub.
* A “Select All” option may appear when applicable.

#### Step 2: Confirm

* Confirm selection to publish the servers.

> Publishing MCP servers enables discovery only. Requests still require valid keys and permissions.

***

## Governance Notes

* Publishing assets does **not** bypass authentication.
* Team, organization, and virtual key restrictions still apply.
* Guardrails and routing rules are enforced exactly as in non-public usage.
* Public visibility is intended for discovery and onboarding.

***

## Common Use Cases

* Provide developers with a clear catalog of available models
* Publish approved agents for internal or partner use
* Advertise MCP capabilities and supported tools
* Reduce onboarding friction without relaxing security

***

## Best Practices

* Publish only approved, stable assets
* Use descriptive names and metadata
* Keep Useful Links updated
* Review public listings periodically
* Pair AI Hub publishing with clear documentation links

***

## Related Sections

* **Models** – Configure and manage models
* **Agents** – Build and manage agents
* **MCP Servers** – Register server-backed tools
* **Virtual Keys** – Enforce access and limits
* **Guardrails** – Apply policy enforcement