OpenClaw setup
OpenClaw doesn’t ship with a model. One command makes TokenMilo its model provider so your assistant runs on Claude and GPT.
Updated 2026-11-16
1. Get an API key
Sign up in the TokenMilo console, top up and create a key under API Keys.
2. Install OpenClaw
# macOS / Linux / WSL2
curl -fsSL https://openclaw.ai/install.sh | bash
# Windows (PowerShell)
iwr -useb https://openclaw.ai/install.ps1 | iexRequires Node.js 24+. The installer starts onboarding — choose the custom API key option and use the values below.
3. Option 1: one command (recommended)
openclaw onboard --non-interactive --accept-risk \
--auth-choice custom-api-key \
--custom-compatibility anthropic \
--custom-base-url "https://api.tokenmilo.com" \
--custom-model-id claude-sonnet-5-5 \
--custom-api-key "sk-your-tokenmilo-key"For GPT, use openai, base URL https://api.tokenmilo.com/v1 and model gpt-6.1-sol.
4. Option 2: edit the config file
Edit ~/.openclaw/openclaw.json to add both Claude and GPT. It reloads automatically.
{
agents: { defaults: { model: { primary: "tokenmilo-claude/claude-sonnet-5-5" } } },
models: {
mode: "merge",
providers: {
"tokenmilo-claude": {
baseUrl: "https://api.tokenmilo.com",
apiKey: "${TOKENMILO_API_KEY}",
api: "anthropic-messages",
models: [
{ id: "claude-sonnet-5-5", name: "Claude Sonnet 5.5", reasoning: true, input: ["text", "image"], contextWindow: 200000, maxTokens: 32000 },
{ id: "claude-opus-5-5", name: "Claude Opus 5.5", reasoning: true, input: ["text", "image"], contextWindow: 200000, maxTokens: 32000 },
],
},
"tokenmilo-gpt": {
baseUrl: "https://api.tokenmilo.com/v1",
apiKey: "${TOKENMILO_API_KEY}",
api: "openai-completions",
models: [{ id: "gpt-6.1-sol", name: "GPT-6.1 Sol", reasoning: true, input: ["text", "image"] }],
},
},
},
}Set the TOKENMILO_API_KEY environment variable, or paste your key into apiKey.
5. Switch models
openclaw models set tokenmilo-claude/claude-opus-5-5Model IDs are in the model list. Send a message and check Usage Logs in the console.
FAQ
Does the Claude URL need /v1?
No. OpenClaw’s Anthropic adapter adds /v1 itself, so use https://api.tokenmilo.com. The OpenAI-compatible provider uses https://api.tokenmilo.com/v1.
I added TokenMilo but OpenClaw still uses my old model
Adding a provider doesn’t change the default model. Run openclaw models set provider/model or edit agents.defaults.model.primary.
GPT models fail with a reasoning_effort error
Add compat: { supportsReasoningEffort: false } to that model entry.
Related guides
- CC Switch one-click setup (Claude Code & Codex) — The easiest way: import TokenMilo into Claude Code and Codex with one click.
- Codex CLI setup — Use OpenAI Codex CLI and the Codex IDE extension with GPT-6 via a custom provider.
- Claude Code setup — Run Claude Code with a TokenMilo key on macOS, Linux and Windows.
- DeepSeek Harness (dsh) setup with Claude & GPT — Add TokenMilo as a custom provider in DeepSeek’s open-source agent harness.
- OpenCode setup — Add TokenMilo as a custom provider in the OpenCode terminal agent.
- Aider setup — Pair-program in your terminal with Claude or GPT via TokenMilo.
Need an API key?
Sign up at TokenMilo, top up, and create a key in the console in under a minute.