What is DCC-MCP-Core?

DCC-MCP-Core is a foundational Rust library with Python bindings for the DCC (Digital Content Creation) Model Context Protocol (MCP) ecosystem. It enables AI assistants to interact with DCC software (Maya, Blender, Houdini, 3ds Max, etc.) through a unified, high-performance interface.

Built with a Rust core exposed to Python via PyO3 and compiled by maturin, it combines the performance and thread safety of Rust with the accessibility of Python.

Core Workflow

mermaid

flowchart LR
    AI([AI Assistant]):::aiNode
    MCP{{MCP Server}}:::serverNode
    Core{{DCC-MCP-Core}}:::coreNode
    DCC[/DCC Software/]:::dccNode

    AI -->|1. Request| MCP
    MCP -->|2. Discover Actions| Core
    Core -->|3. Execute in DCC| DCC
    DCC -->|4. Result| Core
    Core -->|5. Structured Result| MCP
    MCP -->|6. Response| AI

    classDef aiNode fill:#f9d,stroke:#f06,stroke-width:2px,color:#333
    classDef serverNode fill:#bbf,stroke:#66f,stroke-width:2px,color:#333
    classDef coreNode fill:#fbb,stroke:#f66,stroke-width:2px,color:#333
    classDef dccNode fill:#bfb,stroke:#6b6,stroke-width:2px,color:#333

Key Features

ToolRegistry — Thread-safe action registration, search, and versioning
SkillCatalog — Progressive skill discovery and loading; scripts auto-registered as MCP tools via SKILL.md (Skills-First architecture since v0.12.10)
EventBus — Publish/subscribe system for DCC lifecycle events
MCP HTTP Server — Streamable HTTP server (2025-03-26 spec) for serving MCP tools to AI clients
MCP Protocol Types — Type-safe definitions for Tools, Resources, Prompts, and Annotations
Transport Layer — Connection pooling, service discovery, and session management for DCC communication (TCP, Named Pipes, Unix Sockets)
Shared Memory — Zero-copy scene data transfer between DCC and agent processes
Process Management — DCC process launch, monitoring, and crash recovery
Telemetry — Tracing and metrics infrastructure via OpenTelemetry
Sandbox — Security policy, input validation, and audit logging for AI actions
Capture — DCC viewport screenshot capture
USD Bridge — Scene exchange via OpenUSD stage representation
Type Wrappers — RPyC-safe wrappers (BooleanWrapper, IntWrapper, FloatWrapper, StringWrapper)
Zero Python Dependencies — Pure Rust core compiled to a native Python extension

Architecture

DCC-MCP-Core is a Rust workspace with 14 sub-crates, compiled into a single Python extension module dcc_mcp_core._core via maturin:

dcc-mcp-core/
├── src/lib.rs                  # PyO3 module entry point (_core)
├── crates/
│   ├── dcc-mcp-models/         # ToolResult, SkillMetadata, ToolDeclaration
│   ├── dcc-mcp-actions/        # ToolRegistry, EventBus, Pipeline, Dispatcher, Validator
│   ├── dcc-mcp-skills/         # SkillScanner, SkillCatalog, SkillWatcher, Resolver
│   ├── dcc-mcp-protocols/      # MCP types: ToolDefinition, ResourceDefinition, Prompt, DccAdapter
│   ├── dcc-mcp-transport/      # IPC (ipckit), DccLinkFrame, IpcChannelAdapter, SocketServerAdapter
│   ├── dcc-mcp-process/        # PyDccLauncher, ProcessMonitor, CrashRecovery
│   ├── dcc-mcp-telemetry/      # ToolRecorder, ToolMetrics, TelemetryConfig
│   ├── dcc-mcp-sandbox/        # SandboxPolicy, SandboxContext, AuditLog, InputValidator
│   ├── dcc-mcp-shm/            # PySharedBuffer, PyBufferPool, PySharedSceneBuffer
│   ├── dcc-mcp-capture/        # Capturer, CaptureFrame
│   ├── dcc-mcp-usd/            # UsdStage, UsdPrim, VtValue, SdfPath
│   ├── dcc-mcp-http/           # McpHttpServer, McpHttpConfig, McpServerHandle, Gateway
│   ├── dcc-mcp-server/         # Binary entry point, gateway runner
│   └── dcc-mcp-utils/          # Filesystem, constants, type wrappers, JSON helpers
└── python/
    └── dcc_mcp_core/
        ├── __init__.py          # Re-exports ~140 public symbols from _core
        ├── skill.py             # Pure-Python skill script helpers
        └── _core.pyi            # Type stubs for all public APIs

Python API Surface

All public APIs are available from the top-level dcc_mcp_core package. The library exports ~140 public symbols across 14 domains:

python

from dcc_mcp_core import (
    # Actions
    ToolRegistry, ToolDispatcher, ToolPipeline, ToolValidator,
    ToolRecorder, ToolMetrics, EventBus,
    ToolResult, success_result, error_result,

    # Skills — Skills-First architecture
    SkillCatalog, SkillSummary, SkillMetadata, ToolDeclaration,
    SkillScanner, SkillWatcher, scan_and_load,

    # MCP HTTP Server
    McpHttpServer, McpHttpConfig,

    # Transport
    IpcChannelAdapter, GracefulIpcChannelAdapter, SocketServerAdapter, DccLinkFrame,

    # Protocols
    ToolDefinition, ToolAnnotations, ResourceDefinition, PromptDefinition,

    # Shared Memory
    PySharedSceneBuffer, PySharedBuffer, PyBufferPool,

    # Process
    PyDccLauncher, PyProcessWatcher, PyCrashRecoveryPolicy,

    # Telemetry
    TelemetryConfig, is_telemetry_initialized,

    # Sandbox
    SandboxPolicy, SandboxContext, InputValidator, AuditLog,

    # Capture
    Capturer, CaptureFrame,

    # USD
    UsdStage, UsdPrim, VtValue, SdfPath,
)

See the API Reference for complete documentation of every symbol.

Version & Python Support

Current version: 0.14.18
Python: 3.7–3.13 (abi3-py38 wheel, tested in CI across all versions)
Rust: Edition 2024, MSRV 1.85
Build: maturin + PyO3; zero runtime Python dependencies

dcc-mcp-maya — Maya MCP server implementation
dcc-mcp-unreal — Unreal Engine adapter (in development)
dcc-mcp-photoshop — Photoshop UXP WebSocket bridge (in development)
dcc-mcp-zbrush — ZBrush HTTP REST bridge (in development)
dcc-mcp-rpyc — RPyC bridge for remote DCC operations

What is DCC-MCP-Core? ​

Core Workflow ​

Key Features ​

Architecture ​

Python API Surface ​

Version & Python Support ​

Related Projects ​