{"slug": "meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow", "title": "Meshy plugs its 3D generator into ChatGPT and lets Astra run the workflow", "summary": "Meshy released a ChatGPT plugin on September 20th that lets OpenAI's Astra model call Meshy's 3D generation and Auto Split segmentation services and return manipulable 3D assets inside a conversation. The plugin, listed in ChatGPT's plugin directory, has Astra orchestrate the broader project while Meshy handles geometry work, following Meshy's September 9th blog post arguing its purpose-built models offer faster generation and more predictable per-asset costs than Astra's code-and-tools approach. Meshy founder Ethan Yuanming Hu's company is betting specialist models can retain pricing power inside general-purpose agents, with OpenAI having opened its ChatGPT app directory to outside submissions in December 2025 via its Apps SDK and Model Context Protocol connections.", "body_md": "# Meshy plugs its 3D generator into ChatGPT and lets Astra run the workflow\n\n**The plugin generates, segments and displays 3D assets in chat, making OpenAI's generalist model the orchestrator for Meshy's specialist APIs.**\n\n        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)\n        · Published \n\nPrimary source: [MeshyAI on X](https://x.com/MeshyAI/status/2101683299550925127)\n\n## Why it matters\n\nMeshy is betting that specialist AI models can retain pricing power inside general-purpose agents. ChatGPT owns the interface, while Meshy supplies the geometry Astra cannot efficiently produce alone.\n\n[Meshy](https://x.com/MeshyAI/status/2101683299550925127), the 3D generation company founded by [Ethan Yuanming Hu (@YuanmingH)](https://x.com/YuanmingH), released a ChatGPT plugin on September 20th that lets OpenAI's Astra model call Meshy's services, return a generated model in the conversation and split it into separately manipulable parts.\n\n[https://x.com/MeshyAI/status/2101683299550925127](https://x.com/MeshyAI/status/2101683299550925127)\n\nMeshy's launch demo starts with a description of a piece of furniture. Astra sends the request to Meshy, receives the 3D model and then invokes Meshy's segmentation service to divide the object into components including the frame, seat, backrest and rails. Astra uses those assets to assemble a furniture catalog site where each component can be dragged and rotated.\n\nThat sequence matters because it assigns the two models distinctly different jobs. Astra plans and executes the broader project while Meshy handles the specialized geometry work. A user stays in ChatGPT instead of moving between a conversational model, a 3D generator, downloaded asset files and a web development environment.\n\nThe [ChatGPT listing](https://chatgpt.com/plugins/plugins_6a744fc2d0c08191ba89ff29dbd2f8c9) is live, although OpenAI says access to individual plugins can vary by account, plan, workspace, region and product surface. Users can invoke supported plugins through an @ mention or select them from ChatGPT's tools menu, according to [OpenAI's plugin documentation](https://help.openai.com/en/articles/20001256/).\n\n### Meshy turns a competitor into its front end\n\nThe release follows an unusually direct public comparison between the two products. In a September 9th [company blog post](https://www.meshy.ai/blog/meshy-vs-openai-astra), Meshy described Astra as a general-purpose agent capable of creating 3D geometry through code and tools such as Blender. Meshy argued that its own purpose-built models offered faster generation and more predictable per-asset costs.\n\nMeshy also identified the combined workflow that the plugin now formalizes: use Meshy to produce the base asset, then give Astra responsibility for refinement, scene assembly or integration. Eleven days later, Meshy put that division of labor inside ChatGPT.\n\nThe strategy gives Meshy distribution through the conversational interface that could otherwise reduce the visibility of standalone creative software. OpenAI opened its ChatGPT app directory to outside submissions in December 2025, allowing approved tools to be triggered inside conversations. OpenAI said apps could bring external context and actions into a chat through its Apps SDK and Model Context Protocol connections in its [developer announcement](https://openai.com/index/developers-can-now-submit-apps-to-chatgpt/).\n\nFor Meshy, that structure preserves the value of its models even when users begin a project with Astra. Meshy becomes infrastructure called by the agent rather than another tab competing for the user's attention.\n\n### Segmentation makes the output usable\n\nThe technically important step in Meshy's demo comes after generation. Meshy's [Auto Split API](https://docs.meshy.ai/en/api/auto-split) can divide a generated model automatically, follow a prompt naming specific structural pieces or separate an object by color regions. Formats including GLB, OBJ, FBX, USDZ, Blender and 3MF can retain each resulting part as a separate object.\n\nThat lets Astra treat a chair as a collection of components rather than one opaque 3D file. In the catalog demo, the distinction makes the frame, seat and backrest independently available to the site Astra builds.\n\nMeshy introduced the split endpoint on September 9th and updated it on September 16th to accept textured input models, according to Meshy's [API changelog](https://docs.meshy.ai/en/api/changelog). The split result does not preserve the original texture map. Meshy rebuilds the cut surfaces and assigns flat vertex colors, a limitation that could matter for production catalogs requiring finished materials.\n\nThe September 20th video demonstrates a working chain of generation, segmentation and web presentation. It does not independently measure geometry quality, latency or the number of prompts required outside the edited demonstration.\n\n### Hu's graphics research becomes an agent tool\n\nHu built Meshy after studying computer graphics and high-performance computing at MIT CSAIL. His doctoral work centered on Taichi, a programming language for graphics and compute-intensive visual workloads. Hu's [personal site](https://yuanming.taichi.graphics/) says he completed his Ph.D. defense in 2021 after earning an undergraduate degree from Tsinghua University's Yao Class.\n\nThat background is reflected in Meshy's product position. Meshy is selling structured geometry generation and processing as a callable service, rather than asking a general model to approximate every object through generated scripts.\n\nMeshy said on July 21st that it had raised nearly $400 million in a Series B at a $1.5 billion valuation. The figures, along with claims of 12 million registered users, 100 million generated models and roughly 12-fold year-over-year recurring revenue growth, came from Meshy's [funding announcement](https://www.prnewswire.com/news-releases/meshy-raises-nearly-400-million-at-a-1-5-billion-valuation-the-largest-round-to-date-in-ai-3d-302828384.html). Crunchbase lists IDG Capital, Matrix Partners China and Monolith Capital as lead investors, with Granite Asia among Meshy's seven recorded backers.\n\nThe ChatGPT plugin gives that heavily funded model provider a route into projects that begin as ordinary conversations. Astra keeps the user relationship and coordinates the work. Meshy gets the API call when the project requires an actual 3D asset.", "url": "https://wpnews.pro/news/meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow", "canonical_source": "https://runtimewire.com/article/meshy-chatgpt-plugin-astra-3d-generation", "published_at": "2026-09-21 15:00:00+00:00", "updated_at": "2026-09-21 15:25:19.563806+00:00", "lang": "en", "topics": ["ai-agents", "ai-products", "generative-ai", "agent-protocols", "ai-tools"], "entities": ["Meshy", "ChatGPT", "OpenAI", "Astra", "Ethan Yuanming Hu", "Auto Split API", "Blender", "Model Context Protocol"], "alternates": {"html": "https://wpnews.pro/news/meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow", "markdown": "https://wpnews.pro/news/meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow.md", "text": "https://wpnews.pro/news/meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow.txt", "jsonld": "https://wpnews.pro/news/meshy-plugs-its-3d-generator-into-chatgpt-and-lets-astra-run-the-workflow.jsonld"}}