← Back to directory
G

GladeKit MCP

Community
Connect Unity and Godot editors to any MCP-compatible AI client.
Category
Dev Tools #74 of 438
Stars
★ 209 Popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Python
Credentials
Optional API key
License
MIT
Last commit
Tools
8
63FMRS · C

This server is feature-rich and one of the most comprehensive MCP implementations for game development, but requires careful configuration and permission management.

Strongest · Documentation 16/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
12/20
Maintenance
13/20
Documentation
16/20
Setup experience
14/20
Why each score
Reliability 8/20
The README and manifest describe a coherent stdio/HTTP MCP server (gladekit-mcp v0.7.22 on PyPI) with an auto-detecting Unity (:8765) / Godot (:8766) bridge; the happy path is plausible and troubleshooting/error paths are documented. No CI workflows, committed tests, or source code were provided in the review material, so the init/tool-list handshake and declared 260+ (also stated as 235+ elsewhere) tool counts cannot be verified. Minor inconsistencies between the repo description, manifest, and README example log (bridge v0.7.6 vs server 0.7.22) cost further points. Per static-calibration rules, reliability is capped at 12; scored 8.
Security and permissions 12/20
No red-line behavior is evident from the supplied material: no embedded tokens, no covert exfiltration, and the README openly discloses data flows (embeddings to OpenAI, optional cloud layer at api.gladekit.com, GLADE.md project-context injection). The design shows least-privilege and confirmation awareness: loopback-only HTTP binding with DNS-rebinding protection, opt-in LAN exposure, host allowlist for asset downloads, explicit licenseAcknowledged gates, and a disable switch for the asset pipeline. However, this is self-reported README evidence only; the actual guard implementations, dependency supply chain, and cloud API handling cannot be audited statically, so the score is held to 12.
Maintenance 13/20
Signals point to an actively maintained project: MIT license, semver v0.7.22, 0 open issues, not archived, documented release artifacts (Unity UPM path, Godot bridge zips) and a detailed contributing guide. However, the supplied material contains no commit history or dates, no release cadence, no dependency-update evidence, and no security-response channel (SECURITY.md); since cadence and responsiveness cannot be verified and uncertainty lowers the score, it is held to 13.
Documentation 16/20
Documentation is layered and unusually complete for this category: per-client configs (Claude Code, Cursor, Desktop, Cline, Windsurf, Unity AI Gateway, VS Code), an env-var table, troubleshooting, architecture diagram, security/asset-license discipline section, and explicit limitations (Godot Mono unsupported, Claude Code 128-tool cap, bridge version drift). Gaps: no per-tool parameter reference for the claimed 260+ tools (only examples), no instructions for obtaining GLADEKIT_API_KEY, and no changelog/release notes in the supplied material. Scored 16.
Setup experience 14/20
Setup is clear and short on the server side: install uv once, then paste a stdio config (uvx gladekit-mcp) into any of seven documented clients; Mac/Linux/Windows variants are given, and an alternative pip/python path plus a PATH troubleshooting fix are documented. The remaining fragility is the editor side: a bridge must be installed and running in Unity or Godot on a fixed port, with documented failure modes (port conflicts, UPM cache drift). Per static-calibration rules setup cannot exceed 15 without CI/test execution evidence; scored 14.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesRuns commands or codeUses the network

Best for

  • Indie game developers and small teams looking to integrate AI into their Unity or Godot workflows.
  • Developers who want fast prototyping and asset import.
  • Tech-savvy developers who are comfortable configuring MCP servers.

Not for

  • Large studios or teams that require strict asset curation may not want the AI-driven asset pipeline (can be disabled).
  • Environments requiring full offline operation or strict security may not suit the optional cloud features.
  • Godot Mono/C# projects or web export targets are not supported yet.

Required permissions

  • Accesses local editor bridges via HTTP (ports 8765 for Unity, 8766 for Godot).
  • Reads and writes project files, including importing assets, creating GLADE.md, and reading scripts.
  • Optionally calls cloud services with OPENAI_API_KEY or GLADEKIT_API_KEY for semantic search, RAG, and cross-session memory.
  • Executes editor operations such as creating scenes, modifying objects, and running editor commands.

Risks and side effects

  • AI could perform unintended editor actions; use with version control and backups.
  • Asset pipeline allows only CC0-licensed assets, but users must verify license acceptance.
  • Cloud features (if enabled) send data to OpenAI or GladeKit servers, posing privacy risks.
  • The bridge is editor-only and does not affect exported games, but opens local network ports.

Setup

Before you start

Runtime:Python

OPENAI_API_KEY optionalsecret Get at platform.openai.com/api-keys; enables ranked semantic script search.
GLADEKIT_API_KEY optionalsecret GladeKit cloud key; unlocks RAG knowledge base and cross-session memory.
Other optional settings (4)
GLADEKIT_MCP_DISABLE_ASSET_PIPELINE optional Set to 1 to suppress the asset pipeline tools for curated-asset workflows.
GLADEKIT_MCP_SUPPRESS_BRIDGE_WARNING optional Set to 1 to silence the stderr bridge-version warning.
UNITY_BRIDGE_URL optional Overrides the Unity bridge URL, default http://localhost:8765.
GLADEKIT_MCP_FORCE_ENGINE optional Set to unity or godot to pin engine detection when both editors run.
  1. Install the editor bridge:
  • Unity: Add git URL in Package Manager: https://github.com/Glade-tool/glade-mcp.git?path=/unity-bridge
  • Godot (4.3+): Download com.gladekit.mcp-bridge.zip, extract to addons/ directory, and enable the plugin in Project Settings.
  1. Install uv (one-time):
  • Mac/Linux: curl -LsSf https://astral.sh/uv/install.sh | sh
  • Windows: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
  1. Add the MCP config to your AI client, e.g.:
   {
     "mcpServers": {
       "gladekit-mcp": {
         "command": "uvx",
         "args": ["gladekit-mcp"]
       }
     }
   }
   

For Claude Code, run: claude mcp add --transport stdio gladekit-mcp --scope user -- uvx gladekit-mcp.

claude_desktop_config.json
{
  "mcpServers": {
    "gladekit-mcp": {
      "command": "uvx",
      "args": [
        "gladekit-mcp"
      ]
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "gladekit-mcp": {
      "command": "uvx",
      "args": [
        "gladekit-mcp"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add gladekit-mcp -- uvx gladekit-mcp

Run it in a terminal; replace any <…> placeholders with your own values first.

Check that it works

Run uvx gladekit-mcp and you should see the version banner on stderr; in your client's tool list confirm get_relevant_tools and find_asset appear, then ask about the scene structure to prove the Unity/Godot bridge is connected.

Troubleshooting

  1. Verify the bridge is listening on expected ports (Unity: 8765, Godot: 8766).
  2. Ensure uv is installed, or use `pip install gladekit-mcp` and adjust command.
  3. If tools don't appear in Claude Code, use `get_relevant_tools` to discover extended tools (default shows ~80 core tools).
  4. GLADE.md must be in the project root with exact filename (case-sensitive on Mac/Linux).
  5. If asset pipeline doesn't work, check if `GLADEKIT_MCP_DISABLE_ASSET_PIPELINE` is set.

Things to try

Once connected, you can ask your AI assistant things like:

  • Find me free CC0 placeholder art: a platformer character and tile sprites.
  • Import Platformer Pack Redux into Assets/Sprites/Platformer/ - I accept the CC0 license.
  • Audit my imported assets for any license that requires attribution.
  • Search the project scripts to see how enemy spawning works.

Tools 8

get_relevant_tools read-only
Discovers relevant tools based on task description, injecting RAG context.
remember_for_session writes
Stores facts in the current session for later recall.
recall_session_memories read-only
Retrieves facts stored in the current session.
batch_execute read-only
Executes multiple tool steps in one call.
search_project_scripts read-only
Semantic search over project scripts using OpenAI embeddings.
find_asset read-only
Searches for free CC0 assets by description, type, style, and license.
import_asset writes
Downloads, extracts, and imports CC0 assets with a license sidecar file.
list_imported_assets read-only
Lists imported assets and their license info for auditing.

Use cases

Create and modify Unity scenes, GameObjects, scripts, and materials directly from AI conversation.
Quickly locate scripts in large projects using semantic search, even if filenames don't match.
Import free CC0 assets (like Kenney packs) directly from the AI client.
Automatically generate and maintain project context (GLADE.md) during development.
Adapt AI to your skill level and workflow via skill calibration and session memory.

Supported clients

Claude Desktop
Claude Code
Cursor
Windsurf
VS Code
Unity AI Gateway
GitHub Copilot

Listed from the project's documentation, not tested by this site.

Overview

GladeKit MCP is an open-source server that allows AI assistants like Cursor, Claude Code, and Windsurf to directly control the Unity and Godot editors. It provides 260+ granular tools for Unity and 105 native tools for Godot, covering scene manipulation, scripting, physics, animation, UI, and more. The server auto-detects which editor is running and exposes the matching toolset. It also offers an editor-aware system prompt, project context file (GLADE.md), semantic script search, skill calibration, and an optional cloud intelligence layer with RAG and cross-session memory.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →

Source revision a1e34ce3aad5 Data synced 2026-10-12 Read the FMRS scoring method