Official JANCTION Render: https://render.janction.jp
JANCTION Render is a cloud GPU render farm for Blender that AI agents call as an MCP server or HTTP API. This post is about one specific gap we closed this week: how to get a .blend file that sits on your computer into a render job when the AI you are talking to runs in a browser tab.
The problem
Claude.ai and ChatGPT connect to JANCTION Render as a remote MCP server. That works well for scenes the AI writes itself (a bpy script) or for files that already have an https URL. It does not work for a file on your desktop: a remote connector cannot read your disk, and most people will not install a CLI or a Blender add-on just to try something once.
Until now the honest answer was "install the add-on, or use the CLI". That is where most first attempts stopped.
The fix: an upload link
There is now a page at https://render.janction.jp/upload. Drop a file there and you get a link that works for 12 hours. No account, no login, no card. Then you paste the link to your AI.
Three steps, nothing to install:
- Open https://render.janction.jp/upload and drop your .blend (or a glTF / GLB, FBX, USD, OBJ, STL, PLY, Alembic file, or a bpy script).
- Copy the link the page shows.
- Tell the AI: "Render this file: ".
The agent passes the link as scene_url, JANCTION reads the file directly (it does not go back out over the internet), and the first thing you get is a 4-frame preview. Approve it, then ask for the final frames or an MP4.
What happens behind the link, in plain terms
- The file is stored for the lifetime of the link (12 hours). Once a render job has used it, the scene it created follows the normal rule: inputs and results are deleted 24 hours after last use, and nothing is used for training.
- Result links are signed and valid for about 24 hours. Download what you need.
- A link carries one file. If your glTF or OBJ references textures or a .bin, pack it as a .glb or into the .blend first. The scene reader (
scene_info) tells you if something is missing before any GPU time is spent.
Free beta limits, as of October 2026
- 10 GPU-minutes per API key per day (the key is created for you on first connection).
- Final renders up to 240 frames at 1080p per job.
- One GPU, shared with another workload that has priority, so a job can wait. The status page and the agent's
etasay so instead of guessing. - Nothing is charged. Pricing is planned, not billed today; it will be announced on the pricing page before it starts.
When this is the wrong tool
- You want to model or edit inside your own Blender: use a local Blender MCP for that, and hand only the render to JANCTION.
- You render all day, every day: a render farm you pay per GPU-hour, or your own GPU, is cheaper at that volume.
- You need a guaranteed uptime number: there is none during the beta.
Also, JANCTION Render is not related to Render.com. Same word, different product.
Try it with a scene that cannot fail
If you want to see the round trip before touching your own file, paste this instead of a link:
"Render a preview of https://render.janction.jp/samples/cube_scene.py"
It is a small scene with a camera and a light already in place, so the only thing being tested is the connection.
Official JANCTION Render: https://render.janction.jp (upload page: https://render.janction.jp/upload, facts: https://render.janction.jp/facts). Code and client: https://github.com/JasmyLab-JANCTION/janction-render
Top comments (0)