Skip to main content

​
Overview

OpenHands can use TokenLab through its LiteLLM/OpenAI-compatible model configuration, especially for CLI and headless workflows where you provide the model, API key, and base URL.
Type: Coding AgentPrimary Path: OpenAI-compatible model configurationSupport Confidence: Supported custom endpoint path

​
Environment

export TOKENLAB_API_KEY="sk-your-tokenlab-key"

​
Model Settings

Use: 
Model: openai/claude-sonnet-5
API key: sk-your-tokenlab-key
Base URL: https://api.tokenlab.sh/v1
The openai/ prefix tells the LiteLLM-backed path to use an OpenAI-compatible provider configuration.

​
Headless Setup

If you are preparing local defaults for CLI or headless usage, run the OpenHands setup flow and enter: 
make setup-config
When prompted for model details, use the TokenLab model ID with the openai/ prefix and TokenLab’s /v1 base URL.

​
Model Suggestions

Use claude-sonnet-5 or gpt-5.4 for coding-agent planning, gpt-5.4-mini for faster routine loops, and deepseek-v4-pro when you want a strong alternative reasoning model.