game-dev-resource-mcp
License-aware game development resource discovery MCP server. Helps AI coding agents find reusable game assets with commercial-use filtering, license classification, and attribution generation.
README
GameDev Resource MCP
License-aware game development resource discovery, stack recommendation and safe asset installation for AI coding agents via Model Context Protocol (MCP).
Why this exists
AI coding agents are good at finding things, but game-development resources are unusually easy to misuse: “free,” “open source,” “downloadable,” “commercial use,” and “redistributable” are not equivalent.
GameDev Resource MCP gives agents a conservative resource layer that keeps provenance, licensing, search metadata, stack planning, attribution and installation safety together.
It is designed for Codex, Claude Code, Trae and other MCP-capable coding agents.
What V1 can do
- Turn a game description into a practical resource stack with
recommend_stack. - Search multiple game-resource providers with one query.
- Filter by commercial-use posture and license obligations.
- Filter by engine, 2D/3D/audio/font/code, style, format, asset type, genre and animation state.
- Search GitHub for reusable game-development projects and inspect detected repository licenses.
- Generate project-level
THIRD_PARTY_ASSETS.mdandCREDITS.mdcontent. - Resolve official provider-hosted files where supported.
- Safely install an explicitly selected file into a local project with path, host, size and hash checks.
5-minute setup
Requires Node.js 20+.
git clone https://github.com/dsaewkfiush03-debug/game-dev-resource-mcp.git
cd game-dev-resource-mcp
npm ci
npm run build
Run the MCP server over STDIO:
node dist/index.js
Codex CLI
codex mcp add game-dev-resource-mcp -- node /absolute/path/to/game-dev-resource-mcp/dist/index.js
Then verify:
codex mcp list
For Codex config.toml, Claude Code, Trae and generic MCP client examples, see docs/client-setup.md.
GITHUB_TOKEN is optional and only improves GitHub API rate limits. Never commit tokens or API keys.
Recommended workflow
For a whole game or feature idea, start with recommend_stack:
user game description
↓
recommend_stack
↓
resource slots + primary candidates + alternatives + gaps
↓
license summary
↓
select resources
↓
plan_asset_install (when supported)
↓
install_asset_file (explicit file only)
↓
generate_project_attribution
↓
engine-specific integration by the coding agent
For a single known need, call find_game_assets directly.
See docs/end-to-end-example.md for a complete example.
recommend_stack
recommend_stack converts a broad game description into deterministic resource slots such as:
- starter/framework examples;
- environment;
- vehicles;
- characters;
- weapons/combat art;
- UI;
- icons;
- sound effects;
- music/jingles;
- fonts;
- shader/GPU examples where relevant.
The planner recognizes common English and Chinese game-development terms, including engine, 2D/3D, pixel, sci-fi, low-poly, survival, RPG, road/vehicle, combat, inventory and related concepts.
It does not use model judgment to invent license rights. Each slot is resolved through the same provider metadata and conservative license filters used by find_game_assets.
The result includes:
- inferred engine/dimension/style/genre/theme signals;
- required versus optional slots;
- a primary recommendation and alternatives per resolved slot;
- provider failures without discarding other successful slots;
- unresolved required gaps;
- a license/risk summary for primary selections;
- asset attribution and provider service-credit requirements where known.
If a requested engine has no verified starter provider, the slot remains unresolved instead of silently substituting a different engine.
MCP tools
| Tool | Purpose |
|---|---|
recommend_stack |
Turn a game description into a multi-category resource stack with license summary and gaps. |
find_game_assets |
Preferred targeted cross-provider search with ranking and license filters. |
search_game_assets |
Search the curated source registry. |
search_live_assets |
Search one provider directly. |
get_asset_files |
Resolve official provider-hosted file metadata where available. |
plan_asset_install |
Plan a safe installation without writing to disk. |
install_asset_file |
Download one explicitly selected verified file into a local project. |
list_asset_providers |
List provider modes. |
search_open_source_projects |
Search GitHub for reusable game-development repositories. |
inspect_repository |
Inspect repository metadata and detected license. |
check_license |
Classify common licenses conservatively. |
generate_attribution |
Generate attribution for one resource. |
generate_project_attribution |
Generate project-level third-party asset and credits manifests. |
Providers
| Provider | Mode | Typical content | License posture | Auto-install |
|---|---|---|---|---|
| Poly Haven | Live API | HDRIs, textures, 3D models | CC0 assets | Yes, for allowlisted official file URLs |
| Kenney | Verified catalog | 2D, 3D, UI, vehicles, audio | CC0 | No |
| Quaternius | Verified catalog | 3D characters, vehicles, environments | CC0 | No |
| Game Icons | Verified catalog | Game/UI icons | Conservatively treated as CC BY 3.0 | No |
| Google Fonts | Verified catalog | Game-suitable fonts | SIL OFL 1.1 | No |
| Godot demos | Verified catalog | Code, shaders, starter demos | MIT at repository level | No |
| Phaser starters | Verified catalog | Web-game starter templates | MIT code templates | No |
OpenGameArt and arbitrary GitHub repositories remain per-item/per-repository review sources, not blanket-whitelisted providers.
Search model
find_game_assets supports structured filters including:
providerscategoriesenginesdimensionsstylesformatsassetTypesgameGenresanimatedcommercialOnlyallowAttributionallowShareAlike
Default license filtering is conservative:
commercialOnly = true
allowAttribution = true
allowShareAlike = false
Search scores and stack recommendations are retrieval heuristics, not legal clearance.
License model
The project tracks these separately rather than collapsing them into one “safe” flag:
- commercial use
- modification
- redistribution
- attribution
- share-alike / copyleft-like obligations
- canonical source URL
- license source URL
- API/service-level obligations
Examples of recognized licenses include CC0, MIT, BSD, Apache-2.0, CC BY, CC BY-SA, GPL/LGPL families, OpenGameArt attribution licenses and SIL OFL 1.1.
Unknown, custom, missing or non-commercial licenses fail closed to manual review/rejection.
Repository-level licenses must not be assumed to cover independently licensed bundled assets.
Safe installation
V1 automatic installation is intentionally narrow. Only Poly Haven currently exposes a sufficiently verifiable automatic acquisition path.
install_asset_file:
- requires an absolute project root;
- prevents path traversal outside the project;
- only accepts allowlisted HTTPS download hosts;
- defaults to a 128 MiB per-file limit;
- enforces a 1 GiB hard maximum;
- validates provider size metadata;
- validates MD5 when available;
- does not overwrite unless explicitly requested;
- never executes downloaded content;
- never extracts archives;
- never runs package managers or shell commands;
- never automatically clones third-party repositories.
See docs/installation.md.
Attribution manifests
After adopting resources, use generate_project_attribution to create content for:
THIRD_PARTY_ASSETS.md
CREDITS.md
This does not replace legal review, but it makes provenance and release obligations much harder to lose during AI-driven development.
Development
npm ci
npm run validate
CI validates Node.js 20 and 22, then packs the npm tarball, installs it into a clean temporary project and smoke-starts the installed MCP binary.
Design principles
- Correctness over coverage.
- Registry first; asset hosting last.
- Primary-source licensing evidence over AI inference.
- Commercial use and redistribution are separate questions.
- Unknown rights fail closed.
- Never store secrets.
- Never execute downloaded third-party content as part of asset installation.
- Recommendation must reuse the same conservative license filters as direct search.
See AGENTS.md and CONTRIBUTING.md.
V1 scope and limitations
V1 is a practical foundation, not a universal asset index. Verified catalog coverage is intentionally incomplete. Some providers expose only canonical source links because a stable, safe public acquisition API is not available.
The project provides technical metadata and conservative automated classification, not legal advice. Before shipping a commercial game, verify the original license text and provenance of every incorporated resource.
Release notes
See RELEASE_NOTES.md.
License
Project source code is MIT licensed. Third-party resources discovered through this project retain their original licenses.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.