Guides
Rig and animate a 3D model from Claude or ChatGPT with the Vi3W MCP server
Rig and animate 3D models from Claude or ChatGPT with the Vi3W MCP server: connect it, add bones, pick or describe a move, and export FBX or GLB.
6 min read

Quick answer
Can I rig and animate a 3D model from Claude or ChatGPT?
Yes, through an MCP connector to Vi3W. Once Vi3W's server is added to Claude or ChatGPT, the assistant can generate a character, add a humanoid or animal skeleton, and apply a move such as walk, run, jump, wave or dance, or one you describe. You then download the animated model as FBX or GLB.
What the Vi3W MCP adds to a chat
Claude and ChatGPT can't hold a mesh or a skeleton. They read and write text. The Model Context Protocol lets them call tools that do hold those things. Vi3W's MCP server exposes the same model tools you'd use on Vi3W's 3D screen, so an assistant can run them for you in order:
- Generate a model from text or an image.
- Add bones: rig it. A humanoid skeleton by default, or a creature rig when you tell it the model is an animal.
- Make it move: apply a preset move (Idle, Walk, Run, Jump, Wave, Dance) or a move you describe, matched from a motion-clip library.
- New paint job, Clean shape and the downloads, for retexturing, retopology and export.
The chat becomes the place where you decide what you want. Vi3W does the 3D work.

Connect the server
Both assistants add remote MCP servers as custom connectors. You need Vi3W's MCP server URL, which is listed in the Vi3W app.
In Claude
On Pro and Max plans: Customize ‣ Connectors, then + Add ‣ Add custom connector. Name it "Vi3W", paste the server URL exactly as given, review the authentication step, sign in to your Vi3W account and click Add. On Team and Enterprise, an Owner adds the connector for the organisation and members click Connect. Anthropic's custom connector guide has the current menus. Custom connectors work on every Claude plan, including Free.
In ChatGPT
Go to Plugins, select the plus button, then Add custom MCP server. Enter a name, paste the HTTPS endpoint, set up authentication, accept the risk warning and select Create as a plugin. In a chat, type @ and pick Vi3W. OpenAI lists full MCP support (tools that change things, which rigging does) for Business and Enterprise/Edu workspaces, on the web. Pro accounts can connect custom servers with read/fetch permissions only. Details: OpenAI's help article.
More on both setups: how to connect an MCP server to ChatGPT and Blender MCP with Claude, which uses the same connector screens.
A chat that rigs and animates
A realistic session, start to finish:
- "Generate a stylised knight character, standing in a T-pose, arms out, no cape, no shield."
- Check the preview the assistant returns. If the arms are fused to the body, ask for a regenerate with "arms well away from the torso". The rigger needs separated limbs.
- "Add bones to it. It's a humanoid."
- "Make it walk." Walk and Run come with the rig.
- "Now give me a version that waves." Each new move re-rigs the model, so expect a fresh job, not an edit of the last one.
- "Download the waving version as FBX."
For a dog, a horse or a dragon, say so at step 3: "Add bones. It's an animal." That switches to the creature rig, which is built for four-legged and other non-human bodies.
Describing a move
You can ask for moves beyond the six presets in plain words: "a kick", "punching", "a happy jump", "sitting down". Vi3W matches the description to a library of motion clips. It does not invent new motion frame by frame. If nothing in the library fits, it says so and asks for other words. "Celebrate" might land on a cheer, and "hop" lands on a jump. Short verbs work better than long descriptions.
Good prompts:
- "Make it run."
- "Make it dance."
- "Find a punching move."
Prompts that won't work as you hope:
- "Make it limp across the room, then trip over a rock." That's a sequence, and the library has single clips.
- "Make it do the dance from [a specific film]." That's choreography the library doesn't contain, and probably someone else's.
Prepare the model so rigging works
Auto-rigging fails for predictable reasons, and an assistant can't see most of them in a thumbnail. Check these before you ask for bones:
- One body, one mesh. Floating swords, separate hats and capes are better left out of the prompt, or added after rigging.
- A neutral pose. T-pose or A-pose, legs slightly apart, fingers not clenched into the palm.
- Separated limbs. Arms touching the hips get skinned to both, and the hips drag the hands when the character walks.
- A reasonable face count. Very dense AI meshes rig slower and deform worse. Run Clean shape first to rebuild the mesh at a lower face count with the texture baked on. See retopology through the Vi3W MCP.
The AI character topology guide covers why joints need extra edge loops.
Take it into Blender or an engine
Download FBX for Unity, Unreal and Blender, or GLB for web viewers, three.js and Godot. In Blender, import the FBX and the armature arrives with its action, which you can play on the timeline. If you want more moves on the same skeleton, retarget them in Blender: retargeting with an AI agent.
Fixes for the common import problems are in rig an AI-generated character and Blender rig not moving the mesh.
What it can't do
- No custom choreography. Moves come from a clip library. For a unique performance, keyframe it in Blender or use motion capture.
- No facial rig. The rig moves the body. Blinking, lip sync and expressions need shape keys or a face rig built separately.
- No editing an existing rig. Each move rebuilds the skeleton. If you've hand-tuned weights in Blender, retarget clips onto that rig instead.
- No guarantees on odd bodies. A character with six arms, or a creature that's half snake, may rig badly. Check the preview before you download.
- The assistant doesn't see the playback. It reports what the tools returned. Watch the animation yourself before you ship it.
Costs and limits
Rigging, moves, retexturing and Clean shape are paid tools and cost XP. Vi3W shows the cost before each run. Your chat plan's usage limits apply too: a long session with many previews uses more of your Claude or ChatGPT allowance. For how credits compare to tokens, see AI credits explained.
Frequently asked questions
ai rigging and animation
AI rigging places a skeleton inside a mesh and computes how each part of the surface follows each bone. AI animation then applies motion to that skeleton. In Vi3W the two are separate steps, Add bones and Make it move, and both can be called from a chat through the Vi3W MCP server.
can claude animate 3d models
Not on its own: Claude writes text and code. With tools connected it can. Over Blender MCP it can keyframe objects and bones in Blender, and over the Vi3W MCP server it can rig a generated character and apply a motion clip to it.
ai 3d model and rigging
Generate the model first, then rig it. Rigging works best on a single connected character mesh in a neutral pose with arms away from the body. Vi3W can do both in one session, in the app or from a chat.
Last reviewed

