OpenClaw

OpenClaw supports custom model providers. This guide uses its OpenAI Chat Completions client against /v1/chat/completions.

With the CLI

npx @aiml/cli setup openclaw

By hand

Merge this into ~/.openclaw/openclaw.json (or $OPENCLAW_CONFIG_PATH):

{
  "models": {
    "mode": "merge",
    "providers": {
      "aiml": {
        "baseUrl": "https://api.ai.ml/v1",
        "apiKey": "${AIML_API_KEY}",
        "api": "openai-completions",
        "models": [
          {
            "id": "anthropic/claude-sonnet-4.6",
            "name": "anthropic/claude-sonnet-4.6"
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "aiml/anthropic/claude-sonnet-4.6"
      }
    }
  }
}

Put the key in ~/.openclaw/.env. The config refers to it as ${AIML_API_KEY}, so it never sits in the config file itself:

AIML_API_KEY="aiml-live-…"

• OpenClaw validates its config strictly, and a key it does not know stops its gateway. Run openclaw config validate after editing.
• models.mode: "merge" keeps OpenClaw's built-in providers alongside this one.
• Model references are aiml/<model id>, so aiml/anthropic/claude-sonnet-4.6. Add more entries under models, with contextWindow and maxTokens taken from GET /v1/models.
• If your config file has comments, aiml setup leaves it alone and prints the block to add by hand. Rewriting the file would lose the comments.

Verified with OpenClaw 2026.9.7.