PicoBerry

Generate3D models, images, and animationsfor game and 3D workflows from any MCP client — Claude Code, Cursor, Claude Desktop, Cline — with no HTTP glue. A thin wrapper over thePicoBerry/v1API, so you get PicoBerry's multi-engine pipeline directly inside your agent. Several 3D and image engines sit behind one API; calllist_modelsfor the live set and each engine's cost. Generated assets are drafts — useful for prototyping and iteration, and can be reviewed or refined for your project.

📖Full reference:API + MCP docs·PicoBerry API

There's no separate subscription for the MCP or the API.Generation spends the same prepaid PicoBerry credits as the web app, per engine, at rates you can read withlist_modelsbefore you spend anything. (Using the API does require a completed purchase — seeGet an API key.)

// Claude Code: .mcp.json · Claude Desktop: claude_desktop_config.json { "mcpServers": { "picoberry": { "command": "npx", "args": ["-y", "@picoberry/mcp-server"], "env": { "PICOBERRY_API_KEY": "pb_live_xxxxxxxxxxxxxxxx" } } } }

Cursor uses the same shape in~/.cursor/mcp.json.

API access needs a completed purchase: a subscriptionor a one-off credit pack. A purchase entitles you permanently — you don't need acurrentsubscription. (An active paid subscription works too, of course.)
- generate_3d_from_text({ prompt })→ returns an asset{ id }.
- wait_for_asset({ asset_id: id })→ polls untiltaskStatus === 2(succeeded).
- Read the result URL fromfiles.model(GLB) orfiles.image(PNG).

taskStatus:0pending ·1processing ·2succeeded ·3failed. Result URLs are signed and short-lived — download promptly. Errors come back with an actionable message (e.g. an unknown engine returns the list of valid names).

"Make a low-poly treasure chest, retopo it to 3k tris, and give me a Unity FBX."

list_models(category="3d") → pick an engine generate_3d_from_text(prompt="low-poly treasure chest") → { id: A } wait_for_asset(asset_id=A) → taskStatus 2 remesh(asset_id=A, polycount=3000) → { id: B } wait_for_asset(asset_id=B) download_asset(asset_id=B, format="fbx", texture_preset="unity") → signed URL

Run this next toblender-mcpand the agent can generate with PicoBerry, then import into Blender in one flow:

{ "mcpServers": { "picoberry": { "command": "npx", "args": ["-y", "@picoberry/mcp-server"], "env": { "PICOBERRY_API_KEY": "pb_live_..." } }, "blender": { "command": "uvx", "args": ["blender-mcp"] } } }

npm install npm run build # tsc → dist/ PICOBERRY_API_KEY=pb_live_... npm start

RunActions → Publish → Run workflow(or push avtag). It publishes to npm and then to the official MCP registry, in that order — the registry validates by fetching the package's npm metadata and matching itsmcpNameagainstserver.json'sname, so npm has to land first. A guard step checks every invariant (name/version agreement, namespace casing, version not already on npm)beforeanything is published, because npm versions are immutable and a failed half-publish burns the number.

The MCP registry needs no setup at all:mcp-publisherexchanges the Actions OIDC token, and the registry grantsio.github.<repository_owner>/from the token'srepository_ownerclaim. That coversio.github.UModeler/picoberry-mcpand avoids the interactive browser login (which additionally requires org Owner).

Trusted Publishing needsnpm >= 11.5.1, so the workflow runs onNode 24(npm 11.x). Node 22 still bundles npm 10.9 and would fail — thenode-versionpin is load-bearing. A guard step fails the run early if the runner ever ships an older npm.

The namespace is comparedbyte-exactly—io.github.UModeler/..., matching the GitHub org's login. A lowercasedio.github.umodeler/...is rejected 403.

After publishing, claim theGlama listing— unclaimed servers get limited discoverability, andawesome-mcp-serversgates its PRs on a Glama badge in CI.

MCP bridge between an AI agent and a live Aseprite session, same machine or agent-in-a-VM