Blogger Theo (t3.gg) has presented a way to use third-party models, such as GPT-5.6 and Kimi k3, inside Anthropic's official Claude Code CLI tool using a proxy layer called CLIProxyAPI.
What Happened
Using the CLIProxyAPI tool, it has become possible to redirect Claude Code requests to other APIs. This allows developers to leverage the capabilities of GPT-5.6 and Kimi k3 models directly in the terminal via the Claude Code interface, providing flexibility in choosing LLMs for agentic tasks.
Context
The technical novelty of this solution lies not in the models themselves, but in the proxying method—redirecting requests within closed CLI tools. CLIProxyAPI creates an abstraction layer that allows for a multi-model approach within a single workspace.
Why It Matters for the Industry
This case demonstrates a shift from closed ecosystems toward a modular architecture for agentic software. It highlights the growing need for proxy layers (middleware) to bypass proprietary restrictions and the desire of developers to use the most effective models for specific tasks.
Why It Matters for Users
Developers gain the ability to immediately improve task performance by using specialized or more powerful alternatives instead of standard Anthropic models. This accelerates experimentation with agentic tasks and allows for the optimization of cost and quality by switching models.
What Is Not Yet Known / Limitations
There is a fundamental conflict between the positions of Product Builders/Founders (focusing on flexibility and modularity) and Enterprise Architects/IP Counsel (focusing on security risks, compliance, and Terms of Service violations).
Sources
Author
Look at AI, Editorial Team