()


This page edits the model preferences in ~\.openclaw\openclaw.json.

Section 1 — Models that OpenClaw can use are registered within openclaw.json > agents.defaults.models.

Section 2 — Default model preferences set the default primary model and the model fallback order used by all agents.

Section 3 — Overrides enable you to specify per-agent model preferences.  You can specify, agent by agent, what primary model and fallback list to use.

Section 4 — Model-selection routing rules let ClawNanny's Orchestrator automatically select a model for each request based on the prompt content.

1. Registered models that OpenClaw can use  

These are the models registered in openclaw.json > agents.defaults.models.

To add models to or remove models from openclaw.json > agents.defaults.models,
use LanguageModel_list to edit rows in tblLanguageModel,
then click the [ Write from tblLanguageModel to agents.defaults.models ] button below so the new or changed row(s) will be usable for preferences below.


This updates openclaw.json > agents.defaults.models, which automatically restarts OpenClaw to apply the changes.

Model Key Display Name Perf Cost API key for this provider?

✗ Test failures

2. Default model preferences

Fallback order

OpenClaw tries these models in order if the primary is unavailable.

3. Overrides enable you to specify per-agent model preferences.

Checking the checkbox for an agent means that agent uses its own custom primary model and fallback list that you set in Section 3, completely ignoring whatever you set in Section 2 as the default model. 

Unchecking the checkbox means the agent goes back to inheriting Section 2's settings.

You can also edit the primary model and fallback list for an agent by clicking the agent's name.

Loading ...

4. Model-selection routing rules let ClawNanny's Orchestrator automatically select a model for each request based on the prompt content.

Model-selection routing rules let ClawNanny's Orchestrator automatically select a model for each request based on the prompt content. Rules are evaluated in priority order (lower number = higher priority). The first matching rule wins. Use always-type rules as catch-all defaults.

Priority Name Match type Target model Enabled Actions
Loading…
Simulate: which rule fires for this prompt?