Skip to content
Action StationsAction Stations Docs

Other hosted services

Besides OpenAI, Google Gemini and Anthropic, the app can use these hosted services for writing. Each one is a cloud service: prompts, retrieved excerpts of your documents and the scenario so far are sent to that company for every stage routed to it, and you pay the company directly. Before anything is sent, the app shows the request and asks for your approval, and it replaces names, phone numbers, emails and addresses with placeholders by default: see What goes to cloud models.

Service Type in the form What it is
OpenRouter OpenRouter One key for models from many makers, billed by OpenRouter
Mistral Mistral Models from Mistral AI
Groq Groq Open models run on Groq’s fast hardware
DeepSeek DeepSeek Models from DeepSeek
xAI xAI Grok models
Together AI Together AI Many open models, billed by Together
Azure OpenAI Azure OpenAI Models you deploy in your own Microsoft Azure resource

The generic OpenAI-compatible type stays available for any other server that speaks the same protocol.

Set up

Section titled “Set up”
  1. Create an API key on the service’s website and make sure the account can be billed.
  2. Providers → Add provider and choose the service from Type. The address and a suggested model are filled in. The suggestion is only a starting point: change it to any model the service offers.
  3. Paste the key into API key. It is required for these services.
  4. Add provider, then Test. The test asks the service for its model list, not for a generated answer, so it costs nothing. After the provider is saved, the model field offers the models your key can use.
  5. In Routing, assign the tasks you want this service to run.

Azure OpenAI

Section titled “Azure OpenAI”

Azure runs models in a resource you create in your own Azure account, so there is no shared address.

  1. In Azure, create or open an Azure OpenAI (Foundry) resource and deploy a model. Note the resource name, the deployment name and one of the resource’s keys.
  2. In the form, choose Azure OpenAI and replace YOUR-RESOURCE in the base URL with your resource name. The address looks like https://my-resource.openai.azure.com/openai/v1/.
  3. Enter the deployment name as the model, not the model family.
  4. Paste the key. Azure keys are sent in the api-key header, which the app does automatically.
  5. If your organization uses a deployment address (.../openai/deployments/<name>) instead, put that in the base URL and open Advanced to enter the API version your Azure resource expects, such as 2024-10-21. The version is added to every request. Leave it empty for the current address shown above.

OpenRouter

Section titled “OpenRouter”

The app sends the name “Action Stations” and its website address with each request so OpenRouter can show where traffic comes from. These are not secret and carry no project data. Model names look like maker/model; use Load models in the model field to see what is available.

What it costs

Section titled “What it costs”

You pay each service directly, usually per token. The app adds no cost of its own and does not know current prices. Open Advanced in the provider form to enter the Input price and Output price in US dollars per million tokens from the service’s price list. With both set, each stage shows its estimated cost; without them it shows tokens only and says the price is not set. OpenRouter and Together publish prices per model, so enter the price of the model you chose.

Embeddings

Section titled “Embeddings”

Services that offer an embeddings endpoint can index your documents too. In Routing, give the Embeddings row the provider and one of its embedding models, then choose Configured provider under Settings → Embeddings. Your document text is sent to that service while it is indexed, and changing the embedding provider means the index must be rebuilt (Rebuild index on the project’s Documents page). Not every service has embeddings: if a request fails with “not found”, choose a different provider for that row.

Troubleshooting

Section titled “Troubleshooting”
Symptom Fix
Test says the key is not valid The key is wrong or revoked, or belongs to another service. Edit → Replace the key.
“Model not found” or “deployment not found” The model name is wrong. For Azure it must be the deployment name. Choose one from the model list.
Out of credit or quota errors Add credit on the service’s website.
Rate limit errors The app waits for the time the service gives and retries. If they persist, route fewer tasks to the service.
Structured answers keep failing Open Advanced and set JSON mode to JSON object or Ask in the prompt only.