Free Ride - Unlimited free AI

• You want to use OpenClaw without paying for an LLM API subscription and are comfortable with free-tier rate limits.
• You're hitting rate limits on free models and want automatic fallback switching so your sessions aren't interrupted.
• You want to quickly experiment with different free models (e.g. qwen3-coder, nvidia/nemotron) without manually editing config files.
• You want a "set it and forget it" setup where OpenRouter's smart router auto-selects the best available free model per request.

• You need guaranteed SLA, low latency, or consistent throughput — free models have shared capacity and unpredictable availability.
• You are not using OpenClaw; FreeRide writes directly to openclaw.json and is tightly coupled to the OpenClaw ecosystem.

Running freeride auto fetches the current list of free OpenRouter models, ranks them by quality, sets the best as your primary, and populates a ranked fallback list — all in a single step.

FreeRide configures up to 5 fallback models by default (configurable with -c). The first fallback is always openrouter/free, OpenRouter's own smart router, ensuring requests are rerouted automatically when a model hits its rate limit.

Only three keys in ~/.openclaw/openclaw.json are modified: agents.defaults.model.primary, agents.defaults.model.fallbacks, and agents.defaults.models. Everything else — gateway, channels, plugins, env vars, and custom instructions — is preserved.

Use freeride switch <model> to pin a specific free model as your primary, or freeride switch <model> -f to add it as a fallback only, without changing your current primary.

The freeride-watcher daemon continuously monitors for rate-limit events and rotates models automatically, or you can trigger a manual rotation with freeride-watcher --rotate.

Run freeride auto followed by openclaw gateway restart to instantly switch OpenClaw to a free coding-focused model like qwen3-coder. Continue using OpenClaw normally — rate limits trigger silent fallbacks with no session disruption.

Use freeride list -n 30 to see all available free models ranked by quality, then freeride switch <model> to try a specific one. Check active model status anytime with /status inside OpenClaw.

If you've already configured a preferred model, run freeride auto -f to leave your primary model untouched while populating the fallback list with the best current free options.

Enable freeride-watcher --daemon before a long task so the watcher monitors rate-limit events in the background and rotates through fallback models automatically, keeping your session alive without manual intervention.