What is OpenHands?
OpenHands (by All Hands AI) is an open-source platform for autonomous AI software developers. An OpenHands agent can write code, run commands, browse the web and call APIs — doing what a human developer does, in a sandboxed runtime.
- Runs as a containerised agent with a web UI and a sandboxed execution runtime.
- Uses LiteLLM under the hood, so any OpenAI-compatible endpoint works via an
openai/model name. - Plans and executes multi-step tasks end to end, then reports back.
- Configurable from the UI (Advanced LLM settings) or
config.toml.
Why route OpenHands through Tokonomix?
OpenHands is fully autonomous — it can run for many steps unattended. That makes both model quality and a spend ceiling matter, and makes a consensus check on the plan genuinely valuable. Tokonomix gives you all three on one key.
- Drive the agent with any of 130+ models; switch with a one-line model change.
- Use
openai/tokonomix-consensuswhen correctness beats latency on a hard task. - Per-call spend caps stop an unattended run from running away with your budget.
- Region / origin routing keeps a code-streaming agent compliant.
Connect Tokonomix to OpenHands
In the OpenHands Settings → LLM tab, enable Advanced options, then set a custom model with the openai/ prefix and the Tokonomix Base URL.
# OpenHands Settings → LLM → Advanced
Custom Model : openai/anthropic/claude-sonnet-4-6 # or openai/<any-slug>
Base URL : https://tokonomix.ai/api/v1
API Key : tok_live_your-key
# Opt-in (planning/review only, no tool-calling): openai/tokonomix-consensusOpenHands routes through LiteLLM, so the model must be openai/<slug> — e.g. openai/tokonomix-consensus or openai/anthropic/claude-sonnet-4-6 — for the OpenAI-compatible Base URL to be used.
Verify it works
- Save the LLM settings and start a new conversation with a small task.
- Confirm OpenHands plans and executes, and that the model used is your Tokonomix slug.
- Check the request and token usage appear in your Tokonomix dashboard.