openclaw - 💡(How to fix) Fix [Feature]: Allow onboarding to configure multiple providers and models

Summary

Currently, openclaw onboard appears to only allow configuring a single default provider/model during the onboarding flow. It would be very useful to support adding multiple model providers and multiple models directly from onboarding.

Problem to solve

During initial setup, users often need more than one provider/model, for example:

• one primary chat/coding model
• one fallback model
• one vision-capable model
• one cheaper or faster model for lightweight tasks
• one local provider plus one cloud provider
• separate providers for text, image, embeddings, or specialized routing

Today, onboarding guides the user toward a single default provider. After onboarding, adding more providers/models requires manual configuration or additional steps, which makes first-time setup harder and increases the chance of misconfigured routing.

Proposed solution

Extend openclaw onboard so users can add multiple providers and models in the same flow.

Suggested behavior:

1. After configuring the first provider/model, ask whether the user wants to add another provider or model.
2. Allow repeating the provider/model setup flow until the user is done.
3. Let users choose which configured model should be the default/primary model.
4. Optionally let users assign roles during onboarding, such as:
   • primary chat/coding model
   • fallback model
   • image/vision model
   • embedding/memory model
   • local model
5. Persist all configured providers/models consistently in the relevant OpenClaw config files.
6. Show a final summary of configured providers/models before finishing onboarding.

Expected behavior

A user should be able to run openclaw onboard once and configure several providers/models without leaving the onboarding flow.

Example desired flow:

Configure provider #1: OpenAI / GPT-5.x
Add another provider? yes
Configure provider #2: Anthropic / Claude
Add another provider? yes
Configure provider #3: local vLLM / custom model
Add another provider? no
Choose default model: OpenAI / GPT-5.x
Choose fallback model: Anthropic / Claude
Choose vision model: OpenAI / GPT-5.x vision-capable model
Finish onboarding

Actual behavior

Onboarding currently only configures one default provider/model, so users who need multiple providers must complete onboarding first and then configure the rest separately or manually.

Impact

This would improve the onboarding experience for users who rely on multiple providers, fallback chains, local + cloud setups, vision models, or separate embedding/memory models.

It would also reduce manual JSON editing and make the first-run setup closer to real-world OpenClaw usage, where a single default provider is often not enough.

Related context

This is related in spirit to provider configuration UX issues such as #38657, but this request is specifically about supporting multiple providers/models during the initial openclaw onboard flow rather than only editing an existing provider.