Set Up LiteLLM & Web Search
for Your AI Tools

This process uses two different web addresses with completely different purposes. This is the most common source of mistakes, so please note them first.

Ask an administrator to invite you

Your account must be created via an invitation from the administrator, Rao. Please contact her and let her know you need a LiteLLM account. You will then receive an invitation email.

Open the link in the email and register

Find the invitation email in your inbox, click the registration link inside, and fill in the requested information to complete sign-up.

Log in and create a key

1. Open the portal in your browser: https://litellm.feedmob.ai
2. Log in with the account you just registered.
3. Find the Virtual Keys page in the menu.
4. Click the Create New Key button.
5. In Key Name, enter a name you'll recognize, e.g. my-key. Leave the other options at their defaults.
6. Click create. The page will show a key starting with sk-.

Below are the setup instructions for three common tools. You only need to configure the one you actually use — you don't have to do all three. The model name is always minimax/minimax-m3.

Kilo is an AI assistant inside VSCode. We recommend adding the service through its graphical interface first, then adjusting settings as needed.

Part 1 — Add a custom provider in the Kilo UI

• Open Kilo settings and choose to add a custom provider.
• In the API Key field, paste your sk-xxxxxxxx key.
• In the service address (base URL) field, enter the value below:

https://litellm-api.feedmob.ai/v1

Part 2 (optional) — Add capabilities such as image recognition

If you need options not available in the UI (such as image recognition), edit the config file ~/.config/kilo/kilo.jsonc using the example below:

{
  "model": "litellm/minimax-m3",
  "provider": {
    "litellm": {
      "name": "LiteLLM",
      "npm": "@ai-sdk/openai",
      "options": {
        "baseURL": "https://litellm-api.feedmob.ai/v1",
        "timeout": 300000
      },
      "models": {
        "minimax-m3": {
          "id": "minimax/minimax-m3",
          "name": "MiniMax M3",
          "tool_call": true,
          "reasoning": true,
          "attachment": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] }
        }
      }
    }
  }
}

Edit the config file ~/.codex/config.toml and add the following:

model = "minimax/minimax-m3"
model_provider = "litellm"

[model_providers.litellm]
name = "litellm"
base_url = "https://litellm-api.feedmob.ai"
wire_api = "responses"
requires_openai_auth = true
env_key = "LL_API_KEY"

Then set your key as an environment variable (we recommend adding it to ~/.bashrc or ~/.zshrc so it persists):

export LL_API_KEY="sk-xxxxxxxx"

Connect by setting three environment variables (again, we recommend adding them to ~/.bashrc or ~/.zshrc):

export ANTHROPIC_BASE_URL="https://litellm-api.feedmob.ai"
export ANTHROPIC_AUTH_TOKEN="sk-xxxxxxxx"
export ANTHROPIC_MODEL="minimax/minimax-m3"

In whichever tool you configured, ask a question that needs current information, for example:

Search for the latest international oil prices (Brent / WTI).

• If it returns recent information with source links, web search is working.
• If it can't search, first check: whether the service address was set to address 2 (not address 1), whether the key is correct, and whether the model name is minimax/minimax-m3.