Liberate your OpenClaw

Summary

Anthropic is limiting access to Claude models in open agent platforms for Pro/Max subscribers, so the post describes ways to keep OpenClaw, Pi, or Open Code agents running with open models from Hugging Face. It presents Hugging Face Inference Providers as the quickest hosted route: users create a token, run openclaw onboard --auth-choice huggingface-api-key, and select a model such as GLM-5, or set a repo_id in OpenClaw’s config. For fully local inference, it recommends installing llama.cpp, starting a server with Qwen3.5-35B-A3B-GGUF, and connecting OpenClaw to the local OpenAI-compatible endpoint with custom configuration; the example model is described as suitable for 32GB of RAM. The hosted option prioritizes speed and model availability, while local inference offers privacy, zero API costs, full control, and no rate limits.

Takeaways

1. 1

   Hugging Face Inference Providers can route OpenClaw to open models; the post recommends GLM-5 for its Terminal Bench scores and allows model changes by setting the model’s repo_id in OpenClaw configuration.

2. 2

   The local example installs llama.cpp, serves unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL, and connects OpenClaw to http://127.0.0.1:8080/v1; the model is described as suitable for 32GB of RAM.

3. 3

   Hosted inference is presented as the fastest route, while llama.cpp provides local privacy, zero API costs, full control, and no rate limits; HF PRO subscribers receive $2 in monthly free credits for Inference Providers usage.