Connect an AI Provider

Connect an AI Provider

Feluda needs access to at least one AI model before you can start a conversation in Workbench or use AI inside a workflow.

You can connect:

• a cloud AI provider, which runs models through an online service; or
• a local AI provider, which runs compatible models on your own computer.

You only need one working provider to begin. You can add more providers and models later.

Before you begin

Make sure Feluda is installed and opens to the Dashboard.

Then decide which type of provider you want to use.

A cloud provider is often the quickest way to access a wide choice of capable models.

A local provider may be a better choice when you want more of the work to remain on your own computer or when you need to work without an internet connection.

Open AI Providers

In Feluda:

1. Open the sidebar.
2. Select AI Providers.
3. Review the available provider options.
4. Choose the provider you want to connect.

Feluda includes options for supported cloud services and compatible local model applications.

The information requested on the setup screen depends on the provider you choose.

Connect a cloud AI provider

Cloud providers make AI models available through an online account.

Supported options may include services such as OpenAI, Anthropic, Mistral, and Google.

The exact models available to you depend on the provider, your account, and the provider's current model access.

Get an API key

Sign in to the website of the cloud provider you want to use.

Create or copy an API key from the provider's account or developer area. Follow the provider's own instructions because each service manages keys differently.

An API key is a private value that allows Feluda to use your provider account.

Treat it like a password:

• do not share it with other people;
• do not paste it into an ordinary AI conversation;
• do not place it inside a prompt or document;
• do not include it in a public screenshot; and
• remove or replace it immediately if you believe it has been exposed.

Add the provider in Feluda

After you have the key:

1. Open AI Providers.
2. Select the matching cloud provider.
3. Enter the requested provider details.
4. Add the API key in the protected field.
5. Save the provider.
6. Wait for Feluda to show the models available through that provider.

Some providers may request additional information. Enter only the details shown in the provider's own account and Feluda setup screen.

Check the provider account

A valid API key does not always mean that every model is available.

Your provider account may also need:

• access to the selected model;
• an active billing method or provider balance;
• permission to use the provider's API; or
• approval for a particular model or region.

These requirements are controlled by the provider, not by Feluda.

Review the provider's account page if the connection is saved but no models appear.

Connect a local AI provider

A local provider lets Feluda use models running on your own computer.

Feluda supports compatible local model applications such as Ollama and LM Studio.

Before adding a local provider, you need to:

1. install the local model application;
2. download at least one model;
3. start the local model service; and
4. keep it running while Feluda uses the model.

Local models do not normally require an API key.

Connect Ollama

To use Ollama:

1. Install Ollama for your operating system.
2. Download a model through Ollama.
3. Make sure Ollama is running.
4. Open AI Providers in Feluda.
5. Select Ollama.
6. Confirm the local address shown in the setup.
7. Ask Feluda to load the available models.
8. Save the provider.

Ollama commonly uses this local address:

http://localhost:11434

Keep the default address unless you changed the Ollama setup yourself.

Connect LM Studio

To use LM Studio:

1. Install LM Studio.
2. Download a model that your computer can run.
3. Load the model in LM Studio.
4. Start the local server in LM Studio.
5. Open AI Providers in Feluda.
6. Select or add the matching local provider.
7. Enter the local address shown by LM Studio.
8. Load the available models and save the provider.

LM Studio commonly uses this local address:

http://localhost:1234

Use the address shown in LM Studio when it differs from the example above.

Choose a model

A provider may offer several models.

Models can differ in:

• writing quality;
• reasoning ability;
• speed;
• supported input types;
• tool support;
• context size;
• provider cost; and
• computer requirements for local use.

For your first test, choose a general-purpose chat model.

When using a local provider, begin with a smaller model if you are unsure what your computer can run comfortably. You can try a larger model after confirming that the setup works.

You do not need to choose one model permanently. Feluda lets you select a suitable provider and model for each conversation or workflow.

Test the connection in Workbench

Saving a provider is only the first part of the setup. Confirm that the selected model can answer a message.

1. Open Workbench.
2. Select the provider you just added.
3. Select one of its available models.
4. Enter a short message.
5. Send it.
6. Review the response.

Try this message:

Reply with one sentence confirming that the AI model is connected.

If the model responds, the provider is ready to use.

You can now continue working in Workbench or select the same provider inside a workflow.

Add more than one provider

You can connect more than one provider to Feluda.

This allows you to choose different models for different tasks.

For example, you may prefer:

• a local model for private or offline work;
• a fast cloud model for simple tasks;
• a stronger model for detailed analysis; or
• different models for different steps in one workflow.

Add one provider at a time and test each connection in Workbench before relying on it in a workflow.

Use clear provider names when Feluda allows you to name a connection. This makes it easier to choose the correct option later.

Cloud and local privacy

The provider you choose affects where AI processing takes place.

With a cloud provider, the information sent to the model is processed by that provider. Review the provider's own privacy terms before using confidential or sensitive information.

With a local provider, model processing can remain on your computer. The task may still use the internet when you enable an external tool, website, or service.

Before sending sensitive information, check:

• which provider and model are selected;
• whether any tools are enabled;
• whether the task connects to an external service; and
• whether all details in the input are necessary.

Use non-sensitive examples while testing a new provider.

Manage provider keys safely

API keys should only be placed in the protected provider or Secrets areas intended for private values.

Never ask an AI model to remember, rewrite, check, or display an API key.

When a key is no longer needed, remove it from Feluda and revoke it through the provider's account page.

If you believe a key has been exposed:

1. revoke the key with the provider;
2. create a replacement;
3. update the provider in Feluda; and
4. test the new connection in Workbench.

If no models appear

If a provider is saved but its models do not appear, check the following.

For a cloud provider

Confirm that:

• the API key was entered correctly;
• the key is still active;
• your internet connection is working;
• the provider account has API access;
• the account can use the requested models; and
• any provider-side account requirements have been completed.

Save the provider again after correcting the details, then return to Workbench.

For a local provider

Confirm that:

• the local model application is open;
• its local server is running;
• at least one model has been downloaded;
• the local address and port are correct; and
• your firewall is not blocking the local connection.

Restart the local model application if needed, then ask Feluda to load the models again.

If the provider returns an error

The wording of an error can help you find the cause.

Provider messages and options can change over time. Review the provider's own status or account page when Feluda's setup appears correct.

Change or remove a provider

You can return to AI Providers whenever your setup changes.

From there, you can update connection details, replace an API key, or remove a provider you no longer use.

Before removing a provider, check whether any saved workflows use one of its models. Those workflows may need another provider and model before they can run again.

Continue with Workbench

Your provider setup is complete when:

• the provider is saved;
• at least one model is available; and
• the model successfully answers a message in Workbench.

Continue with What is Feluda Workbench? to learn how to write useful instructions, select tools, and review model activity.