Fallbacks only help if they are configured cleanly and have working auth. The dedicated fallback commands make this much easier than editing the model object by hand.
Quick answer
Use openclaw models fallbacks add <provider/model>, then verify the chain with openclaw models fallbacks list and openclaw models status --probe.
Command line steps
1. Add the fallback model
Pick a model that is cheaper or more available than your primary, not one that fails for the same reason.
openclaw models fallbacks add <provider/model>
2. Confirm the chain order
List the fallbacks immediately so you know what OpenClaw will try after the primary.
openclaw models fallbacks list
openclaw models status
3. Probe the provider before you trust it
A configured fallback with missing auth is still a broken fallback.
openclaw models status --probe
What to check if it still fails
- If probe reports missing auth, fix the provider credentials before you count the fallback as real coverage.
- If the model does not show up in
/model, review the allowlist withopenclaw configure --section model. - If failover feels random, simplify the chain first and test with one primary plus one known-good backup.
Frequently asked questions
What will I learn from How to Add a Fallback Model in OpenClaw?
Add fallbacks from the CLI, list the chain, and probe auth so the backup model is actually usable when the primary fails. The guidance is written for practical self-hosted OpenClaw environments, but commands and configuration can change between releases.
Should I back up OpenClaw before changing the configuration?
Yes. Back up the active configuration, credentials references, channel settings, and service definitions before repairs, upgrades, or model changes, then validate the gateway and important workflows after the change.
Can Milan help with OpenClaw setup or troubleshooting?
Yes. Milan provides OpenClaw setup, model and channel configuration, secure remote access, workflow automation, troubleshooting, and deployment support.
