Meshy CLI

Meshy from the command line

One command turns a prompt or an image into a finished, textured model.

terminal — meshy
$ meshy auth login
✓ signed in via your browser — no API key to paste
$ meshy make "girl with a pearl earring, painted bust" -o pearl.glb
● plan: text-to-3d preview (20) → refine (10) · 30 credits
task_019… queued → running
✓ preview mesh ready
✓ refine complete — PBR textures baked
✓ pearl.glb · glTF binary · 224 s · 30 credits
$ meshy balance

Browser sign-in, one make command, a priced plan before anything is charged, and a GLB at the end.

Install meshy-cli

~/your-project

bash

$ npm install -g meshy-cli

Browser login

Sign in once in the browser instead of pasting an API key into every environment you work in.

meshy auth login

Built for CI

JSON on stdout by default and a distinct exit code per failure class, so a pipeline can branch on what actually went wrong.

--json

Scriptable without an agent

One command — meshy make — takes a prompt or an image to a finished, textured model from a shell script, with no agent in the loop.

meshy make

Prefer an agent? Run Meshy over MCP

The MCP server puts the same 24 tools inside Claude Code, Codex, Cursor and any MCP client — same API, same credits.

1

Add the server

Already in a session with Claude Code, Cursor or Codex? Paste this and it will handle the setup.

text
Install the Meshy MCP server for me. Docs: https://github.com/meshy-dev/meshy-mcp-server
Use this env var: MESHY_API_KEY=msy_YOUR_API_KEY
2

Activate and verify

Restart OpenClaw; it reads MCP servers from its config on launch.

Verify with

List the meshy tools available

Then just ask for the model you want.

Paste this into your prompt and the agent reads the full tool reference itself.

What your agent can do now

One connection puts 24 Meshy tools in the conversation — generation is only the start.

Brainstorm before you build

Ask for directions and 2D concepts first, and spend credits only when the idea is right.

Batch out whole sets

One prompt, a full prop set — crates, signs, baskets and chests, ready for your level.

Turn anything into 3D

Text, photos and sketches become textured meshes — or clean sculpts for printing.

It knows 3D

Formats, printability, broken meshes — the tools carry the know-how, so the agent answers and fixes instead of guessing.

24Meshy tools

One connection, and every one of them is callable by name.

Frequently asked questions

Yes. Install it with npm install -g meshy-cli, then run meshy auth login — sign-in happens in your browser, with no API key to paste.

Everything the API exposes. meshy make takes a prompt or an image to a finished model in one command, every endpoint has a full command behind it, and meshy api is the escape hatch. The MCP server covers the same ground inside a terminal agent.

No. Sign-in opens your browser once — an OAuth flow, with no key created or pasted into a shell profile. An API key is still accepted as an override for CI. The MCP server available today does use an API key.

Yes. Alongside the browser flow there is a device-code path for headless machines — the CLI prints a link and a short code to approve from any other device — plus a manual paste-back fallback.

The MCP server is for an AI agent to call. The CLI is for you and for CI to call. Same API underneath, different caller.

meshy-cli is developed in the open in the Meshy GitHub organisation, alongside the MCP server.