---
title: "Model configuration"
description: "Choose a provider, endpoint, credentials, and a model suitable for platform tool use."
---

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

# Model configuration

An Admin configures the model in AI settings. Supported provider choices include OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, Ollama, and OpenAI-compatible services.

## Provider settings

| Provider type | Endpoint | API key |
| --- | --- | --- |
| Built-in hosted providers | Provider-defined; custom endpoint is not accepted | Required |
| OpenAI-compatible | Required | Required |
| Ollama | Required | API-key field is not accepted |

Use the endpoint expected by the selected integration. Ensure the CDelivery server can reach it; a URL accessible only from your browser or another machine may fail from the platform.

Enter credentials, discover available models where supported, and select a model. Model discovery requires Admin access.

## Preserve or replace credentials

Stored keys are not returned in plaintext. An omitted key can reuse the existing one only while provider and endpoint scope remain the same. Changing either scope requires suitable new credentials.

Save the settings, then ask a read-only operational question through a configured channel. Confirm both the model response and any tool results.

## Troubleshoot

Check provider selection, endpoint, key validity, model identifier, outbound network, quota, and tool support.

External providers receive the context needed for the conversation and operations. Configure [AI on-call](/docs/guides/automation/duty/) separately, and use [messaging guides](/docs/guides/automation/im/) for delivery and identity problems.

Source: https://cdelivery.dev/docs/guides/automation/models/index.mdx