{"id":"f01ababe-b910-4f74-a192-796b0f629bcf","slug":"pjain-openclaw-mlx-skill","name":"mlx","description":"Run local Apple MLX-based models for image generation (mflux — Flux, Z-Image, FLUX.2, Qwen), audio/TTS (mlx-audio), vision/multimodal analysis (mlx-vlm), and LLM text inference (mlx-lm). Handles environment discovery, silent dependency installation, model downloads, and execution. Use whenever the user asks to generate images, audio, text, or analyze images locally with MLX.","canonicalUrl":"https://www.xpersona.co/skill/pjain-openclaw-mlx-skill","sourceUrl":"https://github.com/pjain/OpenClaw-MLX-Skill","homepage":null,"source":"GITHUB_OPENCLEW","vendor":{"slug":"pjain","label":"Pjain","url":"https://github.com/pjain/OpenClaw-MLX-Skill"},"protocols":["OPENCLEW"],"capabilities":["distributed","multi"],"trustScore":null,"trustConfidence":"unknown","artifactCount":0,"benchmarkCount":0,"lastRelease":null,"freshnessAt":"2026-02-25T01:46:31.732Z","freshnessLabel":"Feb 25, 2026","securityReviewed":true,"openapiReady":false,"stats":[{"label":"Trust score","value":"Unknown"},{"label":"Compatibility","value":"OpenClaw"},{"label":"Freshness","value":"Feb 25, 2026"},{"label":"Vendor","value":"Pjain"},{"label":"Artifacts","value":"0"},{"label":"Benchmarks","value":"0"},{"label":"Last release","value":"Unpublished"}],"factsPreview":[{"factKey":"docs_crawl","category":"integration","label":"Crawlable docs","value":"6 indexed pages on the official domain","href":"https://github.com/login?return_to=https%3A%2F%2Fgithub.com%2Fopenclaw%2Fskills%2Ftree%2Fmain%2Fskills%2Fasleep123%2Fcaldav-calendar","sourceUrl":"https://github.com/login?return_to=https%3A%2F%2Fgithub.com%2Fopenclaw%2Fskills%2Ftree%2Fmain%2Fskills%2Fasleep123%2Fcaldav-calendar","sourceType":"search_document","confidence":"medium","observedAt":"2026-04-15T05:03:46.393Z","isPublic":true},{"factKey":"vendor","category":"vendor","label":"Vendor","value":"Pjain","href":"https://github.com/pjain/OpenClaw-MLX-Skill","sourceUrl":"https://github.com/pjain/OpenClaw-MLX-Skill","sourceType":"profile","confidence":"medium","observedAt":"2026-02-25T01:46:32.932Z","isPublic":true},{"factKey":"protocols","category":"compatibility","label":"Protocol compatibility","value":"OpenClaw","href":"https://www.xpersona.co/api/v1/agents/pjain-openclaw-mlx-skill/contract","sourceUrl":"https://www.xpersona.co/api/v1/agents/pjain-openclaw-mlx-skill/contract","sourceType":"contract","confidence":"medium","observedAt":"2026-02-25T01:46:32.932Z","isPublic":true},{"factKey":"handshake_status","category":"security","label":"Handshake status","value":"UNKNOWN","href":"https://www.xpersona.co/api/v1/agents/pjain-openclaw-mlx-skill/trust","sourceUrl":"https://www.xpersona.co/api/v1/agents/pjain-openclaw-mlx-skill/trust","sourceType":"trust","confidence":"medium","observedAt":null,"isPublic":true}],"highlights":["Trust evidence available"],"agentCard":{"name":"mlx","description":"Run local Apple MLX-based models for image generation (mflux — Flux, Z-Image, FLUX.2, Qwen), audio/TTS (mlx-audio), vision/multimodal analysis (mlx-vlm), and LLM text inference (mlx-lm). Handles environment discovery, silent dependency installation, model downloads, and execution. Use whenever the user asks to generate images, audio, text, or analyze images locally with MLX.","source":"GITHUB_OPENCLEW","sourceId":"github:1162128263","repository":"https://github.com/pjain/OpenClaw-MLX-Skill","documentation":"https://www.xpersona.co/skill/pjain-openclaw-mlx-skill/agent/pjain-openclaw-mlx-skill","protocols":["OPENCLEW"],"capabilities":["distributed","multi"],"languages":["typescript"],"install":{"command":"git clone https://github.com/pjain/OpenClaw-MLX-Skill.git","ecosystem":"git"},"examples":[{"kind":"example","language":"python","snippet":"import subprocess, sys, platform, json, os\n\ninfo = {}\ninfo[\"platform\"] = platform.platform()\ninfo[\"machine\"] = platform.machine()\n\nis_apple_silicon = platform.machine() == \"arm64\" and platform.system() == \"Darwin\"\ninfo[\"apple_silicon\"] = is_apple_silicon\ninfo[\"mlx_viable\"] = is_apple_silicon\n\ntry:\n    v = platform.mac_ver()[0]\n    info[\"macos_version\"] = v\n    parts = list(map(int, v.split(\".\")))\n    info[\"macos_ok\"] = parts[0] > 13 or (parts[0] == 13 and parts[1] >= 5)\nexcept:\n    info[\"macos_version\"] = \"unknown\"\n    info[\"macos_ok\"] = False\n\ntry:\n    mem = subprocess.check_output([\"sysctl\", \"-n\", \"hw.memsize\"]).decode().strip()\n    info[\"ram_gb\"] = round(int(mem) / 1e9, 1)\nexcept:\n    info[\"ram_gb\"] = \"unknown\"\n\ntry:\n    st = os.statvfs(os.path.expanduser(\"~\"))\n    info[\"free_disk_gb\"] = round(st.f_bavail * st.f_frsize / 1e9, 1)\nexcept:\n    info[\"free_disk_gb\"] = \"unknown\"\n\npackages_to_check = [\"mlx\", \"mflux\", \"mlx_audio\", \"mlx_vlm\", \"mlx_lm\",\n                     \"huggingface_hub\", \"numpy\", \"pillow\", \"soundfile\"]\ninstalled = {}\nfor pkg in packages_to_check:\n    try:\n        mod = __import__(pkg)\n        installed[pkg] = getattr(mod, \"__version__\", \"installed\")\n    except ImportError:\n        installed[pkg] = None\ninfo[\"installed_packages\"] = installed\n\n# mflux cache (moved to ~/Library/Caches/mflux/ in v0.6+)\nmflux_cache = os.path.expanduser(\"~/Library/Caches/mflux\")\ninfo[\"mflux_cache\"] = mflux_cache\ninfo[\"mflux_cache_exists\"] = os.path.exists(mflux_cache)\n\nhf_cache = os.environ.get(\"HF_HOME\", os.path.expanduser(\"~/.cache/huggingface\"))\ninfo[\"hf_cache\"] = hf_cache\nmodels_dir = os.path.join(hf_cache, \"hub\")\ndownloaded = []\nif os.path.exists(models_dir):\n    for item in os.listdir(models_dir):\n        if item.startswith(\"models--\"):\n            downloaded.append(item.replace(\"models--\", \"\").replace(\"--\", \"/\"))\ninfo[\"downloaded_models\"] = downloaded\n\ntry:\n    from huggingface_hub import HfApi\n    info[\"hf_user\"] = HfApi().whoami()[\"name\"]\nexcept:\n    info[\"hf_user\""},{"kind":"example","language":"text","snippet":"8 GB   → Z-Image Turbo q8 (best choice) · flux-schnell q4 · FLUX.2 klein q8\n16 GB  → Z-Image Turbo q8 · flux-schnell q8 · flux-dev q8 · FLUX.2 klein q8\n32 GB  → flux-dev q8 (comfortable) · Qwen Image q6 · FLUX.2 dev q8\n64 GB  → flux-dev full · Qwen Image q8 · FLUX.2 dev q6\n96 GB+ → Any model at full precision"}]}}