model-switch

Name: model-switch
Availability: InStock
Author: Alibaba

Hot-patch openclaw.json to switch the Manager Agent's LLM model.

5.3k|652|Updated Feb 21, 2026

Search Tags:#llm #configuration #model management #ai gateway #hot-swap #manager agent

Ready to run

npx skills add https://github.com/alibaba/hiclaw --skill model-switch-alibaba

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: model-switch
Source: https://github.com/alibaba/hiclaw/tree/main/manager/agent/skills/model-switch
Command: npx skills add https://github.com/alibaba/hiclaw --skill model-switch-alibaba

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill allows for the dynamic and seamless switching of the Manager Agent's underlying Large Language Model (LLM) without requiring a full system restart.

Core Features & Use Cases

Hot-Swapping Models: Instantly change the Manager's LLM by patching its configuration file.
Model Connectivity Test: Verifies model accessibility via the AI Gateway before applying changes.
Reasoning Control: Option to enable or disable the model's reasoning capabilities.
Use Case: When performance issues arise with the current model or a new, more capable model becomes available, an administrator can use this Skill to switch to it immediately, ensuring continuous and optimized operation.

Quick Start

Use the model-switch skill to update the manager's model to claude-sonnet-4-6.

Frequently Asked Questions about model-switch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I hot-swap the Manager Agent's LLM model without restarting?▼

To hot-swap the LLM model, the Skill patches the openclaw.json configuration file to update the model ID, context window, and max tokens, allowing you to switch models without a full system restart.

How does model switching validate LLM reachability through the AI Gateway?▼

Model switching validates model reachability by testing connectivity through the AI Gateway before applying any configuration changes, ensuring the new LLM is accessible before the hot-patch is committed.

Can I disable reasoning capabilities for a specific LLM during a hot-swap?▼

Yes, you can disable reasoning for specific models. The Skill supports reasoning control, allowing you to explicitly enable or disable the model's reasoning capabilities when patching the configuration.

Does hot-swapping models support context window overrides for the Manager Agent?▼

Yes, hot-swapping supports context window overrides. The Skill patches the configuration to apply custom context window sizes and max token limits alongside the new model ID.

When do I need to switch the Manager Agent's underlying LLM model?▼

You need to switch the underlying LLM model when performance issues arise with the current model or when a new, more capable model becomes available, ensuring continuous and optimized operation.

What is the best way to update the Manager Agent to use a new LLM configuration?▼

The best way to update the Manager Agent is by using this Skill to hot-patch the openclaw.json file, which handles model validation, context window overrides, and reasoning control automatically.