Integrations

BeastLab AI exposes an OpenAI-compatible API, which means any tool that lets you configure a custom base URL and API key can talk to Beast models. This page walks through the specifics for the most common coding harnesses.

Before you start

1. Generate an API key from the BeastLab Portal.
2. Store it in an environment variable so your tools can pick it up:

   export BEASTLAB_API_KEY="sk-beast-..."

   Copy
3. Note the base URL you’ll need in every tool: https://api.beastlab.ai/v1

Cursor

Cursor lets you override its OpenAI provider to point at any compatible endpoint.

1. Open File → Preferences → Cursor Settings → Models.
2. Expand the API Keys section.
3. Toggle OpenAI Key on and paste your BeastLab API key.
4. Toggle Override OpenAI Base URL on and paste: https://api.beastlab.ai/v1
5. Scroll up to the model list and click View All Models.
6. Click Add Custom Model and add each Beast model you want to use: beast-max, beast-mini, beast-nano.

You can now select any Beast model from Cursor’s model picker. See a screenshot of the Cursor IDE showing Models and where those settings live. While the OpenAI base URL override is on, OpenAI models in Cursor won’t work; turn the override off to use them again. Other providers and models are unaffected.

Known limitations: Cursor does not appear to expose a way to configure or increase the response timeout for custom models, so longer Beast responses may time out before completing. It also does not appear to surface controls for setting the reasoning effort or search level on Beast models. Once we figure out how to configure these, we’ll update this page.

Kilo Code CLI

Kilo Code CLI is configured via ~/.config/kilo/config.json. Add a beastlab provider block like this:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "beastlab": {
      "name": "BEAST (beastlab.ai)",
      "options": {
        "baseURL": "https://api.beastlab.ai/v1",
        "apiKey": "{env:BEASTLAB_API_KEY}",
        "timeout": 1800000,
        "chunkTimeout": 600000
      },
      "models": {
        "beast-nano": {
          "reasoning_effort": "high"
        },
        "beast-mini": {
          "reasoning_effort": "high"
        },
        "beast-max": {
          "reasoning_effort": "high"
        }
      }
    }
  }
}

Beast models can run longer than typical API defaults allow. The snippet sets timeout to 1,800,000 ms (30 minutes) for the whole request and chunkTimeout to 600,000 ms (10 minutes) between streamed chunks, so slow or large completions are not aborted early.

The {env:BEASTLAB_API_KEY} reference reads your API key from the environment, so you’ll need to make BEASTLAB_API_KEY available to Kilo Code. You have two options:

• Persistent. Add the following line to your ~/.bashrc (or ~/.zshrc) so it’s set for every shell session:

  export BEASTLAB_API_KEY="sk-fXXXXXXXXXXXXX"

  Copy
• Per-invocation. Prefix the command each time you run Kilo Code:

  BEASTLAB_API_KEY=sk-fXXXXXXXXXXXXX kilo

  Copy

OpenCode

Step-by-step instructions coming soon. OpenCode uses the same configuration schema as Kilo Code CLI (note the $schema URL in the snippet above), so the provider block should drop into ~/.config/opencode/opencode.json with minimal changes.

Other OpenAI-compatible tools

For any other tool, look for a setting labelled OpenAI base URL, OpenAI endpoint, or similar, and point it at https://api.beastlab.ai/v1. Use your BeastLab API key in place of an OpenAI key. Model IDs are beast-max, beast-mini, and beast-nano.

Need help with a tool that isn’t listed here? Get in touch and we’ll add it.