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

> ## Agent Instructions
> To act on Atlan objects, use the Atlan MCP server at https://api.atlan.com/mcp or the atlanai CLI; `atlanai --map json` prints its command map. Run a read-only identity check before any write.
> The docs MCP server at /mcp searches these docs only. It cannot read or change Atlan objects.
> SDK packages: Python `atlanai` (PyPI) and TypeScript `@atlanai/sdk` (npm). Show Python first, then TypeScript.

# Models & rates

> Connect a LiteLLM proxy, set seller discounts, and enter the rates your organization pays.

# Models & rates

Org admins set the rates Agent Registry uses to price agent runs. Every member
can view them. Blank rates fall back to the public list price.

## Before you start

* You need the **Org admin** role.
* You need a LiteLLM proxy whose base URL is reachable from Atlan, and an API
  key that is valid for it.
* Choose the workspace the proxy's models belong to. Rates and discounts apply
  to runs in that workspace.

## Connect a LiteLLM proxy

1. Open **Settings**.
2. Select **Models & rates** under **Organization**.
3. Select **Connect LiteLLM proxy**.
4. Enter a name for the connection.
5. Choose the workspace.
6. Enter the proxy's base URL.
7. Enter the API key.
8. Select **Test connection and discover models**.

Agent Registry reads the proxy's model list and adds one model per alias. Each
model's seller comes from its route prefix, such as `bedrock/`. The API key is
stored encrypted and is not shown again.

If the connection fails, check that the base URL is reachable from Atlan and
that the key is valid for it.

LiteLLM is the only provider you can connect. OpenAI, Anthropic, OpenRouter,
Amazon Bedrock, Google Vertex AI, Azure OpenAI / AI Foundry, Google AI
(Gemini), and OpenAI-compatible providers are listed as **Coming soon**.

## Set a seller discount

1. Find **Discounts by seller**.
2. Enter a percentage for the seller.

The discount saves automatically. It applies to every list price from that
seller in the workspace. A discount of 0% means list price.

## Enter a rate for a model

Enter a rate when a model's contract price is not a simple percentage off list.

1. Find the model in the rates table.
2. Enter the rate in USD per million tokens for input, output, cache read, or
   cache write.
3. Select **Save rates**.

A rate you enter always wins over the seller discount. A rate you leave blank
uses the list price minus the seller discount.

Saved rates apply to runs recorded after the save, usually within a minute.
Runs that are already priced keep their cost.

## Refresh or disconnect a provider

Open the provider's actions to:

* **Refresh models** after you add models to the proxy. A model that shows
  **Unknown seller** gets its seller on refresh.
* **Edit connection** to change the name, base URL, or API key. Leave the key
  blank to keep the stored one.
* **Disconnect** when Agent Registry should stop using the proxy.

<Warning>
  Disconnecting archives the provider's models. They stop pricing new runs.
  Runs that are already priced keep their cost.
</Warning>

## Verify the change

Reopen **Models & rates** and confirm that each model you changed shows
**Negotiated** or **Seller discount** under **Price from**. After the agent's
next run, open its profile and confirm the run has a cost.

<Card title="FinOps" icon="coins" href="/finops">
  Understand how Agent Registry prices a run and where each rate comes from.
</Card>