JasmyLab has started offering JANCTION Render, a Blender cloud GPU rendering service for AI agents. Blender files and scenes written by an AI can be rendered on GPUs in a Tokyo data center, from an environment that has neither a GPU nor Blender. No sign-up is required, and the first previews need no payment.
Service site: https://render.janction.jp/
What is JANCTION Render
JANCTION Render is a service that takes on Blender rendering on cloud GPUs. What sets it apart from earlier render farms is that, in addition to a screen operated by people, it has had entry points that AI agents can call directly (an MCP server and an HTTP API) from the start. By adding a single connection to ChatGPT, Claude, Codex, Cursor and similar tools, you can leave the whole flow to the AI: it builds the scene, renders it with JANCTION Render, looks at the result and fixes it.
There are three ways to use it.
- In the browser: place a .blend, a 3D file such as GLB/FBX/USD, or a bpy script on the site and it renders.
- From an AI: add the MCP URL (https://render.janction.jp/mcp) to ChatGPT, Claude, Codex, Cursor or a similar tool, and ask it to "render this scene".
- From the terminal: render from coding agents such as Claude Code and Codex through the MCP server distributed on PyPI.
In each case, you check composition and materials with a "preview" that comes back in seconds, make fixes, and then move on to the "final render" (final rendering and animation). The amount is shown before rendering, so there are no unexpected costs.
Main specifications
- Supported Blender versions: 3.6 / 4.2 / 4.5 / 5.0 / 5.2 (Cycles and EEVEE)
- Accepted files: .blend, .glb / .gltf, .fbx, .usd / .usdz, .obj, .stl, bpy scripts
- GPU: NVIDIA RTX PRO 6000 (Blackwell, 96GB), in a Tokyo data center
- Output: still images, animation (MP4), turntables, product shots (4 directions)
- Data retention: 24 hours after last use (normally 7 days if you have topped up)
Pricing
Each new API key comes with 100 JPY of free credit, and no card registration is required. Registering a card adds a further 400 JPY (you are not charged). After that, pricing is pay-as-you-go at 0.1 JPY per GPU-second, and only what you use is deducted from prepaid credit. Top-ups start at 100 JPY and are denominated in yen. Previews are free up to 2 GPU-minutes each day; free keys can run final renders up to 1080p, and topped-up keys up to 4K. See the pricing page for details.
Since the release
JANCTION Render was released as a free beta on September 30, 2026, and switched to credit-based service on October 8. In about one month since release, renders from outside users have exceeded 1,600 (as of November 4, 2026, excluding use by the operator). The MCP server code is published under the MIT license, and it can also be installed from the official MCP Registry, PyPI, Smithery and others.
- Code: https://github.com/JasmyLab-JANCTION/janction-render
- Usage and examples: https://render.janction.jp/examples
Looking ahead
JANCTION Render is the first service to offer JANCTION's distributed GPU infrastructure as "computing resources used by AI". Going forward, we will expand the supported AI tools and output formats while building an environment in which creators in 3DCG and video production do not lose time waiting for renders. For corporate use or integration into production workflows, please get in touch through the contact page.