Custom Server URL Guide

Connect Grambo to any local or remote AI service โ€” LM Studio, Ollama, LiteLLM, vLLM, or any OpenAI-compatible endpoint

Back to Help Center

๐Ÿ“– Overview

Grambo's Custom Server URL feature lets you connect to any AI service that exposes an OpenAI-compatible API. This includes local AI tools like LM Studio, Jan, and manually configured Ollama, as well as self-hosted or remote services like LiteLLM, vLLM, and AnythingLLM. You supply the server URL, model name, and an optional API key โ€” Grambo handles the rest.

๐Ÿ“ก OpenAI Format Required: The server you connect to must return responses in the OpenAI chat completions format. Most modern local AI tools support this out of the box.

โš™๏ธ How to Set Up Custom Server URL

Follow these steps to connect Grambo to your custom AI server:

  1. Open Grambo Settings: Launch Grambo on your Mac and click the "Settings" button.
  2. Choose "Bring Your Own Key (BYOK)" under "Service Type".
  3. Select "Custom Server URL" from the model/provider options.
  4. Enter your Server URL: Paste the base URL of your AI server, including the path (e.g. /v1):
    http://localhost:1234/v1
  5. Enter the Model Name: Type the exact model name your server is running (e.g. llama3, mistral, phi3):
    llama3
  6. Enter your API Key (if required by your server). If your server does not require authentication, enter none or leave it blank:
    none
  7. Save your settings and start using Grambo with your custom server.

โœ… You're all set! Grambo will now send grammar correction requests to your custom server. All processing happens through your chosen AI service.

๐Ÿ”Œ Compatible Services

Any service that implements the OpenAI chat completions API format will work. Here are some popular options with their default URLs:

Service Default URL API Key
LM Studio http://localhost:1234/v1 Optional
Ollama (manual) http://localhost:11434/v1 Not required
Jan http://localhost:1337/v1 Optional
LiteLLM http://localhost:4000/v1 Required
vLLM http://localhost:8000/v1 Optional
AnythingLLM http://localhost:3001/api/openai Required
Remote / Custom Server https://your-server.com/v1 Varies

๐Ÿ’ก Tip: Make sure your local AI server is running before saving your settings in Grambo. You can verify it's running by visiting the server URL in your browser.

๐Ÿ“ก OpenAI Response Format Requirement

Grambo expects the server to respond using the standard OpenAI chat completions format. This is the same format used by the official OpenAI API and is widely supported by local AI tools.

  • Endpoint: Your server must accept POST requests at /chat/completions (relative to the base URL you provide)
  • Request format: Grambo sends a standard messages array with a system prompt and user text
  • Response format: The server must return a JSON object with choices[0].message.content containing the corrected text
  • Model field: The model name you enter in settings is passed in the request body as the model parameter

โš ๏ธ Compatibility: If your server does not return responses in the OpenAI format, Grambo will not be able to parse the correction. Check your server's documentation to confirm it supports the OpenAI chat completions API.

โ“ Frequently Asked Questions

What is the difference between Custom Server URL and the built-in Local AI?

The built-in Local AI uses Grambo's auto-setup to install and manage Ollama for you. Custom Server URL is for advanced users who already have an AI server running โ€” whether locally or remotely โ€” and want to connect Grambo to it manually.

Do I need an API key?

It depends on your server. Most local servers (LM Studio, Ollama, Jan) do not require an API key. If your server requires authentication, enter the key in the API Key field. If not, you can enter "none" or leave it blank.

Can I connect to a remote server, not just localhost?

Yes! You can use any URL โ€” local (http://localhost:...) or remote (https://your-server.com/v1). Just make sure the server is accessible from your Mac and returns OpenAI-compatible responses.

What model name should I enter?

Enter the exact model identifier your server expects. For example, in LM Studio this is shown in the model dropdown. For Ollama, it's the model tag you pulled (e.g. llama3, mistral, qwen2). Check your server's interface or documentation for the correct name.

Why is Grambo not getting a response from my server?

Check that: (1) your server is running and accessible at the URL you entered, (2) the model name is correct, (3) the server returns OpenAI-compatible JSON responses, and (4) if an API key is required, it is entered correctly.

Can I use Custom Server URL alongside other BYOK providers?

Yes. You can switch between Custom Server URL, OpenAI, Anthropic, Google Gemini, and Local AI anytime in Grambo's settings. Each mode is independent.

๐Ÿ“ง Need Help?

If you have trouble connecting to your custom server or have questions about compatibility, feel free to reach out to us at support@gramboapp.com