Skip to main content
Glama
README.md
# 3DAssets.dev for agents

Free, web-optimised CC0 GLB models and game asset packs for three.js, React Three Fiber, Blender, Godot, Unity and any glTF-capable tool. This repository holds the machine-readable pieces of https://3dassets.dev in one place for directories and agent tooling that index GitHub:

- `SKILL.md`: the agent skill, a copy of https://3dassets.dev/skill.md (kept in sync daily by a workflow)
- `bin/3dassets-mcp.js`: the `3dassets-mcp` npm package, a stdio bridge to the hosted server
- `.mcp.json`: MCP client configuration for the hosted server
- `server.json`: the official MCP Registry entry
- `llms-install.md`: setup notes for agents that install from a README

The site, the API and the MCP server itself are hosted at 3dassets.dev, so the only code here is the thin stdio bridge; the catalogue is at https://3dassets.dev/assets.

## MCP server

Two ways in. Hosted (Streamable HTTP) needs nothing installed; the npm package runs the same server over stdio for clients that only speak stdio or want a pinned dependency. Both expose the same tools.

```sh
# stdio, via npm (bridges to the hosted server; no local state)
npx -y 3dassets-mcp
```

```json
{
  "mcpServers": {
    "3dassets": { "command": "npx", "args": ["-y", "3dassets-mcp"] }
  }
}
```

Set `THREEDASSETS_API_KEY` in the environment to submit models on the user's behalf; leave it unset for read-only use.

### Hosted endpoint

Streamable HTTP endpoint: `https://3dassets.dev/mcp`. Search and download need no auth. Submitting models on a user's behalf takes their API key as a Bearer token; https://3dassets.dev/auth.md explains how an agent obtains one without the user ever handling a password.

```sh
# Claude Code
claude mcp add --transport http 3dassets https://3dassets.dev/mcp

# Codex CLI
codex mcp add 3dassets --url https://3dassets.dev/mcp

# VS Code
code --add-mcp '{"name":"3dassets","type":"http","url":"https://3dassets.dev/mcp"}'
```

Cursor, Windsurf or any `mcp.json`:

```json
{
  "mcpServers": {
    "3dassets": { "url": "https://3dassets.dev/mcp" }
  }
}
```

Tools: `search_assets`, `search_packs`, `get_pack`, `list_demos`, `get_demo`, `get_asset`, `get_asset_usage`, `list_categories`, `list_tags`, `list_licenses`, `list_ai_models`, `create_account`, `verify_account`, `submit_asset_from_url`, `get_upload_url`, `finalize_upload`, `my_assets`, `update_asset`, `submit_pack`, `my_packs`. Full reference: https://3dassets.dev/docs/mcp.

## Skill

```sh
npx skills add 3dassets-dev/3dassets
```

Or point any agent at https://3dassets.dev/skill.md directly. The skill uses plain HTTP, so it works without MCP.

## REST API

Base `https://3dassets.dev/api/v1`, OpenAPI at https://3dassets.dev/api/v1/openapi.json, guide at https://3dassets.dev/docs/api. Every asset also has a Markdown view at `https://3dassets.dev/assets/{slug}.md`.

## Discovery documents

- https://3dassets.dev/.well-known/mcp/server-card.json
- https://3dassets.dev/.well-known/api-catalog
- https://3dassets.dev/.well-known/agent-skills/index.json
- https://3dassets.dev/.well-known/ai-catalog.json
- https://3dassets.dev/llms.txt

## Licence

Every asset on the site is CC0 1.0. The files in this repository are CC0 1.0 too.

Questions: hello@3dassets.dev

TDQS

A3.7/5.0

Scored across 20 tools

Disambiguation5/5

Each tool has a clearly distinct purpose: account management, asset upload steps, asset and pack queries, metadata lists, and usage examples. The stepwise upload flow (get_upload_url, submit_asset_from_url, finalize_upload) is well-separated, and user-specific vs. general queries are unambiguous.

Naming Consistency3/5

Most tools follow a verb-first pattern (get_, list_, search_, submit_, create_, verify_, finalize_), but 'my_assets' and 'my_packs' break this convention by using a noun phrase. This inconsistency is noticeable but not chaotic, as the verbs are otherwise regular.

Tool Count4/5

With 20 tools, the server is slightly above the typical 3-15 range, but the breadth is justified by covering accounts, assets, packs, demos, metadata, and usage. It feels comprehensive rather than bloated, and each tool serves a distinct function.

Completeness5/5

The surface covers the full asset lifecycle: account creation/verification, both upload methods, asset retrieval/search/update, pack creation/search, demo listing, and supporting metadata (categories, tags, licenses, AI models). No obvious missing operations like delete are required given the immutable design (assets cannot be changed after upload).

Maintenance

ActivityMaintained
ResponsivenessNo issues