OpenClaw keeps forcing Anthropic even after switching provider to OpenAI
I'm trying to switch my OpenClaw setup to use OpenAI instead of Anthropic, but the system keeps trying to call Anthropic and throwing errors even though my provider is set to OpenAI.
Originally I was getting this error:
Agent failed before reply: No API key found for provider "anthropic".
Auth store: .../agents/main/agent/auth-profiles.json
So I removed the Anthropic provider entirely and configured OpenAI as the only model provider using openai/gpt-4.1-mini.
Now the provider config is valid and OpenAI is set up, but the agent still doesn't run.
In the Agents → main → Overview page I noticed:
Primary Model shows -
There is no dropdown to select a model.
When I try to manually enter:
openai/gpt-4.1-mini
or
["openai/gpt-4.1-mini"]
into the Primary model (default) field, the dashboard won't let me save and throws a comma / formatting error.
I also can't configure fallbacks because there is no model dropdown there either.
So right now the agent has:
Primary model: none
Fallbacks: none
But the system still tries to call Anthropic internally.
Errors I'm seeing now include:
GatewayRequestError: Sentinel value "OPENCLAW_REDACTED" in key models.providers.*.apiKey is not valid as real data
and sometimes:
Agent failed before reply: No API key found for provider "anthropic"
So it seems like the agent is still defaulting to Anthropic somewhere even though the provider config is OpenAI.
Main questions:
Where is the agent model fallback actually defined if not in the provider config?
Why does the Agents UI not show a dropdown for available models?
Is there another config file that defines the agent model separate from models.providers?
Right now I basically cannot set a primary model for the agent, so the agent never runs.
Originally I was getting this error:
Agent failed before reply: No API key found for provider "anthropic".
Auth store: .../agents/main/agent/auth-profiles.json
So I removed the Anthropic provider entirely and configured OpenAI as the only model provider using openai/gpt-4.1-mini.
Now the provider config is valid and OpenAI is set up, but the agent still doesn't run.
In the Agents → main → Overview page I noticed:
Primary Model shows -
There is no dropdown to select a model.
When I try to manually enter:
openai/gpt-4.1-mini
or
["openai/gpt-4.1-mini"]
into the Primary model (default) field, the dashboard won't let me save and throws a comma / formatting error.
I also can't configure fallbacks because there is no model dropdown there either.
So right now the agent has:
Primary model: none
Fallbacks: none
But the system still tries to call Anthropic internally.
Errors I'm seeing now include:
GatewayRequestError: Sentinel value "OPENCLAW_REDACTED" in key models.providers.*.apiKey is not valid as real data
and sometimes:
Agent failed before reply: No API key found for provider "anthropic"
So it seems like the agent is still defaulting to Anthropic somewhere even though the provider config is OpenAI.
Main questions:
Where is the agent model fallback actually defined if not in the provider config?
Why does the Agents UI not show a dropdown for available models?
Is there another config file that defines the agent model separate from models.providers?
Right now I basically cannot set a primary model for the agent, so the agent never runs.
OpenClaw: The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
146,573Members
Resources
Similar Threads
Was this page helpful?
