{"slug": "qwen-mm-plugins", "title": "Qwen-MM-Plugins", "summary": "Alibaba's Qwen team released Qwen-MM-Plugins, a suite of native multimodal plugins that add vision, video, and 3D capabilities to Qwen models and any agent harness. The plugins install as skills plus optional MCP servers, with a single script supporting Claude Code, Codex, Qoder, OpenClaw, Qwen Code, and Gemini CLI. The suite includes core vision tools, video memory and editing, Blender and FreeCAD integration, and an education agent, with installation via curl or harness-specific commands.", "body_md": "**English** · [中文](/QwenLM/Qwen-MM-Plugins/blob/main/README.zh.md)\n\nNative multimodal plugins for Qwen models. Make any agent harness multimodal-native.\n\n[🧩 Capabilities](#-capabilities)[🏗 Architecture](#-architecture)[📦 Installation](#-installation)[🔧 Dependencies](#-dependencies)[🔑 Configuration](#-configuration)[🚀 Quick Start](#-quick-start)[🧪 Development](#-development)\n\nEach capability is installed separately — a **skill** (so the model knows the toolset exists) plus an optional **MCP server** (the tools themselves).\n\nWe ship [ cookbooks](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks) of Qwen3.8-Max + these plugins in action — each capability's cookbook (linked in the table below) has its full tool listing, setup, and worked cases. Enjoy!\n\n| Capability | What it does | Install name | Cookbook |\n|---|---|---|---|\ncore |\nFoundational vision: dynamic-resolution reading of images / videos / documents / 3D models, plus OCR, grounding, segmentation, ASR, vision chat, and web search | `qwen-mm-plugins-core` |\n|\n\n**video-memory**`qwen-mm-plugins-video-memory`\n\n[TBD](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks/video-memory/usage.md)**video-edit**`qwen-mm-plugins-video-edit`\n\n[TBD](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks/video-edit/usage.md)**blender****running** Blender via Python (thin client, 22 tools) — modeling / materials / lighting / rendering`qwen-mm-plugins-blender`\n\n[TBD](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks/blender/usage.md)**freecad****running** FreeCAD (thin client, 14 tools) — modeling, property edits, STEP/STL import/export, FEM analysis`qwen-mm-plugins-freecad`\n\n[TBD](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks/freecad/usage.md)**edu-agent****skill-only**, no MCP server)`qwen-mm-plugins-edu-agent`\n\n[TBD](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks/edu-agent/usage.md)A capability = a **skill** (so the model knows the tools exist) + an optional **MCP server** (the tools themselves, launched on demand by `uvx`\n\n— needs [uv](https://docs.astral.sh/uv/), no manual pip).\n\nOne script handles **install · configure · verify · uninstall** across every harness it supports (Claude Code · Codex · Qoder · OpenClaw · Qwen Code · Gemini CLI). It drives each harness's own native install under the hood — nothing reinvented — and writes a single shared config file (`~/.qwen-mm-plugins/config`\n\n) that GUI and terminal harnesses both read, so you set things up once:\n\n```\ncurl -fsSL https://raw.githubusercontent.com/QwenLM/Qwen-MM-Plugins/main/install.sh | bash\n```\n\nOr run one action at a time — `bash install.sh install`\n\n/ `configure`\n\n/ `verify`\n\n/ `uninstall`\n\n(what `configure`\n\nand `verify`\n\ndo is detailed under [Configuration](#-configuration) and [Dependencies](#-dependencies)).\n\n**Windows x64:** use WSL2 (Ubuntu recommended) and clone the repository inside your WSL\nhome directory (for example `~/code`\n\n), rather than under a mounted Windows drive such as\n`/mnt/c`\n\n. Then run the same commands there. WSL2 is currently the only supported Windows\nenvironment; native Windows has not yet been validated. See the concise\n[Windows notes](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/installation.md#windows-wsl2).\n\nPrefer your harness's own commands — or you're on opencode / pi / QwenPaw, which the installer doesn't cover? Register the skill + MCP yourself.\n\n**Plugin-marketplace harnesses** (Claude Code · Qoder · Codex · OpenClaw · Qwen Code) — add the marketplace, then install a capability (replace `<cap>`\n\nwith `core`\n\n/ `video-memory`\n\n/ `video-edit`\n\n/ `blender`\n\n/ `freecad`\n\n):\n\n```\n# Claude Code\nclaude   plugin  marketplace add https://github.com/QwenLM/Qwen-MM-Plugins.git\nclaude   plugin  install       qwen-mm-plugins-<cap>@qwen-mm-plugins\n# Qoder\nqodercli plugins marketplace add https://github.com/QwenLM/Qwen-MM-Plugins.git\nqodercli plugins install       qwen-mm-plugins-<cap>@qwen-mm-plugins\n# Codex\ncodex    plugin  marketplace add https://github.com/QwenLM/Qwen-MM-Plugins.git\ncodex    plugin  add           qwen-mm-plugins-<cap>@qwen-mm-plugins\n# OpenClaw\nopenclaw plugins install       qwen-mm-plugins-<cap> --marketplace https://github.com/QwenLM/Qwen-MM-Plugins.git\n# Qwen Code\nqwen extensions install https://github.com/QwenLM/Qwen-MM-Plugins.git:qwen-mm-plugins-<cap> --consent\n```\n\n`marketplace add`\n\nalso accepts a local repo path; re-running is safe. On **codex**, `marketplace add`\n\ndoes **not** refresh an already-added marketplace, so run `codex plugin marketplace upgrade qwen-mm-plugins`\n\nbefore `plugin add`\n\nto pick up newly-published capabilities.\n\n**Other harnesses** (Gemini CLI · opencode · pi · QwenPaw · …) register the skill + MCP in their own config — exact per-harness blocks are in [ docs/en/installation.md](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/installation.md). Easiest of all:\n\n**just ask the agent**— \"install\n\n`qwen-mm-plugins-<cap>`\n\n\".`uvx`\n\ninstalls the Python dependencies for the chosen profile on first launch — no manual pip. The only things you install yourself are **system tools**: `ffmpeg`\n\n(video / audio), plus optional `libreoffice`\n\n/ `blender`\n\n/ `texlive`\n\n/ `chromium`\n\nfor `visualize`\n\n. Run `bash install.sh verify`\n\nto self-test what's installed — it confirms your API key and reports any missing system tools (fetching each capability's env and running `--check-system`\n\nunder the hood). Full system-tool table, the edu-agent (skill-only) setup, and the blender/freecad thin-client notes: see [ docs/en/installation.md](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/installation.md).\n\nThe API-based tools need a key — native image / video / document reading doesn't:\n\n`DASHSCOPE_API_KEY`\n\n—`vision_chat`\n\n/`ocr`\n\n/`grounding`\n\n/`transcribe_audio`\n\n/ generation / video-memory build`SERPER_API_KEY`\n\n—`web_search`\n\n/`web_extractor`\n\n/`image_search`\n\nExport them in your shell, or persist them to `~/.qwen-mm-plugins/config`\n\n(read whenever a var isn't already in the environment — so GUI-launched harnesses pick them up too). The guided installer's Configure step writes that file for you:\n\n```\nbash install.sh configure\n```\n\nFor non-interactive/automation setup and the full environment-variable catalog, see [ docs/en/installation.md](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/installation.md).\n\nOnce a capability is installed, reference a file in your harness and just ask — the model picks the right tool automatically. Reading is **dynamic-resolution**: every image, video frame, and document page is auto-scaled to the VL model's patch grid, so a 4K screenshot's fine print and a tiny thumbnail both come in at the detail they need — no manual resizing.\n\n```\n# core — read images / video / docs / 3D models, plus OCR · grounding · segmentation · ASR · web search\n@dashboard-4k.png      Read every number in this dashboard.\n@report.pdf            Summarize page 3.\n@receipt.jpg           OCR this and total the line items.\n@street.jpg            Draw a box around every car in the scene.   # grounding\n\n# video-memory — QA over long videos; the first query auto-builds memory\n@lecture-2h.mp4        What are the main points, with timestamps?\n\n# video-edit — image / video / audio generation + editing workflows\n                       Generate a 1024×1024 image of a red panda coding at night.\n@/path/to/media        Help me edit this video down to about 3 minutes.\n\n# blender — drive a running Blender to model / texture / light / render (thin client, 22 tools)\n                       Model a low-poly wooden stool, add a warm key light, and render it.\n\n# freecad — parametric CAD in a running FreeCAD (thin client, 14 tools; STEP/STL, FEM)\n                       Model an M6 hex bolt 30 mm long and export it as STEP.\n\n# edu-agent — turn a math/science problem into a step-by-step Chinese explainer video (skill-only)\n@geometry-problem.png  Explain how to solve this as a narrated video.\n```\n\nSee each capability's 🍳 [cookbook](/QwenLM/Qwen-MM-Plugins/blob/main/cookbooks) for every tool, setup, and a worked case.\n\nDevelopment setup, contribution guidelines, and verification commands are in\n[ CONTRIBUTING.md](/QwenLM/Qwen-MM-Plugins/blob/main/CONTRIBUTING.md). Detailed guides:\n\n[local development](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/local_development.md)·\n\n[adding a capability](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/how_to_add_new_capability.md)·\n\n[testing](/QwenLM/Qwen-MM-Plugins/blob/main/docs/en/testing.md).\n\nApache-2.0 — see [ LICENSE](/QwenLM/Qwen-MM-Plugins/blob/main/LICENSE). The Blender and FreeCAD capabilities vendor third-party MIT-licensed code; see\n\n[and](/QwenLM/Qwen-MM-Plugins/blob/main/src/capabilities/blender/NOTICE.md)\n\n`src/capabilities/blender/NOTICE.md`\n\n[for attribution.](/QwenLM/Qwen-MM-Plugins/blob/main/src/capabilities/freecad/NOTICE.md)\n\n`src/capabilities/freecad/NOTICE.md`", "url": "https://wpnews.pro/news/qwen-mm-plugins", "canonical_source": "https://github.com/QwenLM/Qwen-MM-Plugins", "published_at": "2026-08-10 05:17:24+00:00", "updated_at": "2026-08-10 05:35:48.486843+00:00", "lang": "en", "topics": ["artificial-intelligence", "developer-tools", "ai-agents"], "entities": ["Alibaba", "Qwen", "Qwen-MM-Plugins", "Claude Code", "Codex", "Qoder", "OpenClaw", "Qwen Code"], "alternates": {"html": "https://wpnews.pro/news/qwen-mm-plugins", "markdown": "https://wpnews.pro/news/qwen-mm-plugins.md", "text": "https://wpnews.pro/news/qwen-mm-plugins.txt", "jsonld": "https://wpnews.pro/news/qwen-mm-plugins.jsonld"}}