Custom Server: setup and connection examples

Enter the connection details

In Settings → AI → Custom Server, choose Add. The same procedure applies to apps that include the shared settings; installed versions and supported AI features differ.

  1. Name: an optional label that helps you recognize this connection.
  2. Base URL: copy the URL from an example below. Keep path segments such as /api/v1.
  3. API format: select the format shown below, or use Auto for the listed direct endpoints.
  4. API Key: paste your own key into the app, not this website. These four hosted services require credentials even though the form says “optional” for keyless local servers.
  5. Model: copy a current text-chat model ID from your provider’s model catalog. Use the API ID, not a display name. Model access and supported request options vary by account and model.
  6. Choose Save & Verify. After verification, select the connection in your chat or in AI action provider.

Save & Verify sends a short chat request. Hosted services may charge for it. A model list alone does not prove that chat requests will work.

On failure, check the message and correct the connection. Saved-but-unverified settings can be edited; do not assume they are ready for AI actions.

OpenRouter

Base URL
https://openrouter.ai/api/v1
API format
OpenAI-compatible

Use an OpenRouter API key and the exact model ID from its catalog, including the provider prefix (for example, the vendor/model form). A Claude or Gemini model reached through OpenRouter still uses OpenAI-compatible. Requests pass through OpenRouter to the selected provider.

Find a model ID in the official catalog · Official API documentation

OpenAI API

Base URL
https://api.openai.com/v1
API format
OpenAI-compatible

Use an OpenAI API key and a model that supports Chat Completions. QuickApps uses /chat/completions for this format; a Responses-only model or feature is not covered. Do not paste a /responses endpoint into Base URL.

Find a model ID in the official catalog · Official API documentation

Anthropic API

Base URL
https://api.anthropic.com/v1
API format
Messages

Use a Claude API key and the exact Claude model ID. This example is for the direct Anthropic API. Cloud-platform variants may require different URLs, authentication or additional headers. Credentials requiring a workspace header cannot be fully configured in the current form.

Find a model ID in the official catalog · Official API documentation

Gemini API

Base URL
https://generativelanguage.googleapis.com/v1beta
API format
GenerateContent

Use a Gemini API key and a model ID that supports generateContent. Enter the model ID without the models/ prefix. This example is for the Gemini Developer API. Vertex AI and other Google Cloud authentication paths need separate compatibility checks.

Find a model ID in the official catalog · Official API documentation

Auto and custom gateways

Auto selects Messages for api.anthropic.com and GenerateContent for generativelanguage.googleapis.com (except its OpenAI-compatible path). Other hosts default to OpenAI-compatible. A custom domain or proxy may need an explicit format.

The server URL determines the transport, not the model’s brand. OpenAI-compatible means the Chat Completions interface here; it does not guarantee every vendor feature, model or request option. Additional authentication headers or specialized cloud authentication may need an adapter or a gateway you manage.

If verification fails

Unauthorized / 401
Check the key, its permissions and the intended account. Do not put keys in support screenshots.
Forbidden / 403
Check model access, account restrictions and the service’s supported regions. A gateway does not guarantee access to a service unavailable in your region.
Not found / 404
Check the full Base URL, API format and exact model ID.
Invalid request / 400
Check whether the model supports the requested API and options. Try a documented text-chat model; a newer or reasoning model may have different requirements.
Rate limit / 429
Check usage limits, quota and the provider’s status. Wait before retrying.
Connection refused / timeout
Check network access and that a local server is running. Another device needs a reachable network address, not localhost.

Keys, costs and data

QuickApps stores Custom Server keys in operating-system secure storage. This page has no API-key form, does not test a server and does not apply settings to an app. Copy only public connection values, then enter credentials in the app.

Provider API access has its own account, pricing and usage conditions. Confirm costs and supported countries with the provider. Before sending personal or confidential content, review the actual destination and data handling, including any onward provider used by a gateway.

Availability and capabilities vary by app, installed version, model, account and region. These are configuration examples, not a claim that every model or feature is supported.

About these examples

Documentation and QuickApps shared settings reviewed: 2026-11-01. Applicable to versions showing Custom Server with an API format selector. External paid API calls for these four examples have not been verified as part of this guide. Successful local-server tests do not establish compatibility with every hosted model.

Back to AI connection guide · Contact support