> ## Documentation Index
> Fetch the complete documentation index at: https://bruno-a6972042-reform-ai-docs.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.
# Connect a Custom AI Provider
> Connect an OpenAI-compatible hosted, proxied, or local AI endpoint to Bruno AI.
Bruno AI is available from **v4.0.0** and higher. Turn it on under
**Preferences → AI → Configuration** before configuring a provider.
Bruno can connect to any provider that exposes an **OpenAI-compatible API**. This includes third-party hosted providers, company gateways and proxies, self-hosted inference servers, and local model applications such as Ollama or LM Studio.
**Bruno never stores your API keys.**
The keys stay on your machine and are not written to collections or Git. Bruno saves them locally in the same application-data folder as `preferences.json`, and they never leave your machine except when Bruno talks to the provider you configured.
* macOS: `~/Library/Application Support/bruno/`
* Windows: `%APPDATA%\bruno\`
* Linux: `~/.config/bruno/`
## What OpenAI-compatible means
An OpenAI-compatible provider implements API routes and request formats based on the OpenAI API. Bruno needs:
* A **Base URL** for the compatible API
* An optional **API Key**
* At least one exact **Model ID** accepted by the endpoint
Compatibility refers to the API format. The model itself does not need to be
made by OpenAI.
## Before you begin
Start the provider or local model server and find its Base URL and model ID in its documentation or interface.
Common local examples:
| Provider | Base URL example | Model ID example |
| --------- | --------------------------- | ------------------------------------- |
| Ollama | `http://localhost:11434/v1` | `llama3.1:8b` |
| LM Studio | `http://localhost:1234/v1` | The identifier displayed by LM Studio |
The exact values depend on your provider and configuration.
## Add an endpoint
1. Open **Preferences** from the bottom-left corner of Bruno.
2. Go to **AI → Configuration**.
3. Turn on **AI Features** if it is not already enabled.
4. Under **OpenAI-Compatible Endpoints**, click **+ Add endpoint**.
5. Enter a **Name** that identifies the provider, such as `Ollama local` or `Company AI gateway`.
6. Enter the provider's **Base URL**, including any required API prefix such as `/v1`.
7. Add an **API Key** if the endpoint requires authentication.
The API key is optional. Local endpoints commonly do not require one, while hosted providers and company gateways usually do.
## Add models
Under **Models**:
1. Enter the **Model ID** expected by the provider.
2. Optionally enter a friendlier display **Label**.
3. Click **Add**.
4. Enable the model using its checkbox.
5. Toggle the custom provider on.
Model IDs must match the provider exactly, including punctuation, versions, or tags. For example, Ollama may require `llama3.1:8b`, not only `llama3.1`.
You can add multiple model IDs to one endpoint and configure multiple custom endpoints.
## Test the connection
After setting the Base URL and enabling the provider, click **Test connection** (the bolt icon).
* **Connection successful** means Bruno can reach the endpoint and use its configuration.
* If the test fails, review the error and check the endpoint, authentication, and model configuration.
## Edit or remove an endpoint
Expand the endpoint to:
* Change its **Name** or **Base URL**
* Add, replace, or remove its API key
* Add, rename, disable, or remove models
* Select **Remove endpoint** to delete the endpoint from Bruno
## Local provider examples
### Ollama
1. Install and start Ollama.
2. Pull a model:
```shell theme={null}
ollama pull llama3.1:8b
```
3. Confirm the OpenAI-compatible API is running:
```shell theme={null}
curl http://localhost:11434/v1/models
```
4. In Bruno, use:
* Base URL: `http://localhost:11434/v1`
* API Key: leave empty
* Model ID: the exact name shown by `ollama list`
### LM Studio
1. Download and load a model in LM Studio.
2. Open **Developer** (called **Local Server** in some versions).
3. Start the local server.
4. Confirm the API is running:
```shell theme={null}
curl http://localhost:1234/v1/models
```
5. In Bruno, use:
* Base URL: `http://localhost:1234/v1`
* API Key: leave empty unless authentication is enabled
* Model ID: the identifier shown by LM Studio
Bruno cannot connect to a local endpoint unless its server is running.
Starting or loading a model in the provider application may be a separate step
from starting its API server.
## Troubleshooting
### Connection refused
The endpoint is not running, the port is incorrect, or Bruno cannot reach the host. Test the provider's models endpoint from the same machine before testing it in Bruno.
### 404 Not Found
Check the Base URL. Many compatible providers require an API prefix such as `/v1`.
### Unauthorized or forbidden
Confirm that the endpoint requires an API key and that the key has the necessary permissions. Replace the saved key if it has expired or been revoked.
### Model not found
Copy the model ID exactly from the provider. Include required tags or versions.
### Models do not appear in Bruno
Make sure:
1. The endpoint has a Base URL.
2. The custom provider is enabled.
3. At least one model has been added and selected.
4. The main **AI Features** toggle is enabled.
## Security considerations
A custom endpoint receives the AI context that Bruno sends to it. Review the endpoint owner's data-handling policy before connecting a hosted provider.
Bruno's AI security protections redact sensitive headers, body fields, response values, and secret variables by default. Review them under **Preferences → AI → Security**.
## Next steps
* [Connect OpenAI](/ai/bruno-ai/openai)
* [Connect Anthropic](/ai/bruno-ai/anthropic)
* [Configure autocomplete and security](/ai/bruno-ai/configuration)
* [Use Bruno AI](/ai/bruno-ai/usage)