Meshy 3D Agent is the world's first conversational AI agent for 3D creation. Describe an idea, upload a photo or sketch, brainstorm directions, generate a batch of concepts, refine by chatting, and export all in one conversation.
Open to all plans, including Free · Runs on Meshy's standard credit system · Export to FBX, OBJ, GLB, USDZ, STL, BLEND, 3MF & DXF
3D has lived behind a wall of software hours of tutorials before you can place a single cube. Even Text to 3D tools solve generation but not three other friction points:
The blank canvas: knowing you want "something for my daughter's birthday" without knowing what, exactly.
One-shot generation: a single prompt gives one result no room to explore before committing.
Tool-switching overhead: ideation in one app, generation in another, questions answered in a help center third.
Meshy 3D Agent folds brainstorming, batch generation, conversational refinement and built-in 3D knowledge into a single chat.
Example: one brainstorm, several concepts pick a direction before going to 3D
Pitches creative directions before generating anything useful when you know the category but not the specific design yet.
Produces multiple concepts in one thread while keeping a consistent style built for asset sets, not one-offs.
Turns a chosen concept into a downloadable 3D model without leaving the conversation, then keeps refining it conversationally ("rounder, bigger helmet").
Ask about wall thickness for printing, polygon budgets for a target engine, or export formats directly in the chat, no help center required.
Use the agent when the idea is fuzzy; use Text to 3D or Image to 3D when the spec is sharp.
| 3D Agent | Classic Text/Image to 3D | |
|---|---|---|
| Interaction | Ongoing conversation, multi-step | One prompt → one output per run |
| Starting point | Photo, sketch, description, story, or a vague direction | A composed prompt or prepared image |
| Ideation | Built in suggests directions before generating | None you arrive knowing what you want |
| Exploration | Batches of concepts in one thread | Serial regeneration, prompt by prompt |
| Refinement | Conversational ("rounder, bigger helmet") | Manual prompt edits and tool switching |
| 3D knowledge | Ask questions in-chat (printing, formats, topology) | Consult docs and help center separately |
| Best for | Open-ended ideas, beginners, gifts, exploring style directions | Production users who know exactly what they need, batch pipelines, API automation |
| Status | Beta (launched Jun 4, 2026), all users | Mature, fully documented |
"Agent" means different things depending on where you meet it. All four run on the same underlying Meshy API and credit system.
The chat interface inside meshy.ai itself what this page is about. Open to every plan, including Free.
An open-source meshy-3d-agent repo of Markdown skills that let coding assistants (Claude Code, Cursor, OpenClaw) call the Meshy API directly.
The structured, Node.js alternative to Skills same capabilities, built-in auto-save, for assistants that speak MCP.
The developer-facing interface underneath everything else, for custom pipelines and automation. See API pricing.
The meshy-3d-agent skills repo lets Claude Code, Cursor, OpenClaw and 20+ other assistants generate, retexture, rig, animate and print-prep 3D models requires a Pro plan or above for API access.
| meshy-3d-generation (core) | Credits | meshy-3d-printing (optional) | Credits |
|---|---|---|---|
| Text to 3D | 20-30 | White Model Print | 20 |
| Image to 3D | 20-30 | Multicolor Print | 40 |
| Retexture | 10 | Slicer Detection | 0 |
| Remesh | 5 | Analyze Printability | 0 (free) |
| Auto-Rigging | 5 | Repair Printability | 10 |
| Animation | 3 | ||
| Text to Image / Image to Image | 3-9 |
A third module, meshy-openclaw, unifies these for the OpenClaw ecosystem with the API key stored locally only.
Prefer structure over simplicity? The MCP server offers the same capabilities with built-in auto-save and typed tools instead of Markdown + Python skills for simplicity, MCP for structure.
Any plan, including Free, at meshy.ai.
It sits alongside Text to 3D and Image to 3D in the workspace.
A description, a photo, a sketch, or just a vague direction.
Let the agent pitch directions before generating anything.
Review several concepts in the same thread, not one at a time.
Wall thickness, polygon budgets, export formats no tab-switching.
Turn the chosen concept into a model, then export.
Remesh, AI Texturing and rigging live outside the chat for now.
Official product tour
3D Agent walkthrough
Available to every plan, including Free, since the beta launch. No separate fee usage runs on Meshy's standard credit system.
Require a Pro plan or above, since they authenticate with a Meshy API key. See API pricing.
Whether you generate through the chat, the app, or a coding assistant, a Text to 3D generation costs the same 20-30 credits either way.
An AI agent for 3D creation built into meshy.ai, launched June 4, 2026. It helps users develop ideas, generate series of visual concepts, ask 3D-related questions, and turn selected concepts into downloadable 3D models all through a single chat conversation.
Traditional Text to 3D generates one result from one prompt. The 3D Agent supports a multi-step, chat-based process: you can explore directions, refine ideas conversationally, and generate multiple concepts before committing any of them to 3D.
The beta launched as available to all users on meshy.ai including free accounts. Usage runs on Meshy's standard credit system rather than a separate fee.
Almost anything: a text description, a creative direction, a story, a photo (a pet, a product), or a sketch even a child's drawing.
Yes built-in 3D creation knowledge is one of the four launch capabilities. You can ask about printing and modeling topics, like wall thickness, directly in the chat. See the 3D printing page for more.
They're ordinary Meshy assets: downloadable for creative, printing, or production workflows, and editable with the platform's standard tools (Remesh, AI Texturing, rigging).
No. The 3D Agent is Meshy's own chat interface inside meshy.ai. The MCP server is the reverse direction: it lets external AI assistants use Meshy as a tool.
For the front of the pipeline, yes direction-finding, concept spreads, and style exploration. For volume production against a locked art bible, the classic tools remain sharper.
An open-source set of pure Markdown agent skills that let coding assistants drive the Meshy API directly text/image to 3D, retexture, remesh, rigging, animation, 2D image generation, and a full 3D-printing workflow.
Yes the skills authenticate with a Meshy API key, and API keys require a Pro plan or above.
Both expose the same Meshy capabilities. Skills are simpler (copy Markdown files, Python + requests) and work across 20+ assistants; the MCP server offers structured tools and built-in auto-save.
Very nearly. The optional printing skill runs printability analysis (free), repairs mesh issues, prepares white-model OBJ or multicolor 3MF output, and auto-detects seven slicers see the full 3D printing workflow.
Meshy 3D Agent is open to every plan, including Free no separate fee, just your standard credits.
Try the Agent →