Add OpenClaw to local apps

[osolmaz]

Summary

• Add OpenClaw as a local app for tool-calling GGUF/MLX conversational models
• Reuse the existing local OpenAI-compatible server step shared by Pi and Hermes
• Configure OpenClaw via models.providers and set the selected local model as the default

Test plan

• pnpm --filter @huggingface/tasks test -- local-apps
• pnpm --filter @huggingface/tasks format:check
• pnpm --filter @huggingface/tasks check

Note: assumes the OpenClaw local-app logo asset is already available for the openclaw app key.

---

Note

Low Risk
Additive snippet registry and tests only; no runtime or security-sensitive logic changes.

Overview
Adds OpenClaw as a local app option on model pages for tool-calling GGUF and MLX conversational models (same visibility rule as Pi and Hermes).

Generated snippets walk users through starting the shared local OpenAI-compatible server (llama-server or mlx_lm.server), running openclaw onboard with custom provider settings pointed at http://127.0.0.1:8080/v1, and a sample openclaw agent command. Provider id switches between llama-cpp and mlx-lm based on model type.

Vitest coverage asserts the GGUF and MLX snippet shapes, including that onboarding uses --auth-choice custom-api-key without emitting a --custom-api-key flag.

^{Reviewed by Cursor Bugbot for commit faa3f61. Bugbot is set up for automated code reviews on this repo. Configure here.}

[pcuenca]

Haven't tested the commands, but conceptually looks good.

[osolmaz]

I am now going through this extensively and asking Peter about the long-term stability of CLI args, thank you for bearing with me. I will ping again once I am 100% confident the snippets fit the long term shape

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ccd307c317

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[osolmaz]

Current output of the snippets:

GGUF / llama.cpp:

# Install llama.cpp:
brew install llama.cpp

# Start a local OpenAI-compatible server:
llama-server -hf ggml-org/gemma-4-E4B-it-GGUF:Q4_K_M

# Install OpenClaw:
npm install -g openclaw@latest

# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "ggml-org/gemma-4-E4B-it-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

openclaw agent --local --agent main --message "Hello from Hugging Face"

MLX:

# Install MLX LM:
uv tool install mlx-lm

# Start a local OpenAI-compatible server:
mlx_lm.server --model "mlx-community/Qwen3-0.6B-4bit"

# Install OpenClaw:
npm install -g openclaw@latest

# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "mlx-community/Qwen3-0.6B-4bit" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

openclaw agent --local --agent main --message "Hello from Hugging Face"

[osolmaz]

AFAIU test failures are unrelated to the changes proposed in this PR

[pcuenca]

The llama.app website was released after this was first reviewed. I would now favor using the one-liner installer (instead of brew) and llama serve (instead of llama-server), but I suggest we merge this now and then update for both OpenClaw and Hermes in a new PR.

What do you think @gary149?

[osolmaz]

Sounds good to me either way. The reason I wasn't pushing earlier was to make sure openclaw logo and other assets are in the hub repo

I don't know what would happen if we merge this before assets, enum values, etc. are added there. It wouldn't get added to the list of applications right away, right?