RenderDoc MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@RenderDoc MCP Server Analyze C:\captures\frame.rdc and find all textures containing 'normal'"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
RenderDoc MCP Server
一个用于 RenderDoc 的 MCP (Model Context Protocol) 服务器,允许 AI 助手分析和检查 RDC 捕获文件。
功能
打开和分析 RDC 捕获文件
获取捕获的基本信息(API 类型、驱动、帧数等)
列出和搜索纹理资源
列出和搜索缓冲区资源
获取 Draw Call 列表和详情
查看着色器信息
获取像素数据
安装
前置要求
RenderDoc: 从 renderdoc.org 下载并安装
Python 3.10+: 确保已安装 Python 3.10 或更高版本
安装 MCP 服务器
# 克隆仓库
git clone https://github.com/yourusername/renderdoc-mcp.git
cd renderdoc-mcp
# 安装依赖
pip install -e .配置 RenderDoc Python 模块
要使用原生 Python API(推荐),需要设置 PYTHONPATH 包含 RenderDoc 的 Python 模块路径:
Windows:
set PYTHONPATH=%PYTHONPATH%;C:\Program Files\RenderDoc\plugins\python或者在环境变量中永久设置。
如果不设置,服务器会自动回退到使用 renderdoccmd 命令行工具。
配置 Claude Desktop
在 Claude Desktop 配置文件中添加:
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"renderdoc": {
"command": "python",
"args": ["-m", "renderdoc_mcp.server"],
"env": {
"PYTHONPATH": "C:\\Program Files\\RenderDoc\\plugins\\python"
}
}
}
}macOS/Linux: ~/.config/claude/claude_desktop_config.json
{
"mcpServers": {
"renderdoc": {
"command": "python",
"args": ["-m", "renderdoc_mcp.server"],
"env": {
"PYTHONPATH": "/usr/share/renderdoc/plugins/python"
}
}
}
}可用工具
工具名称 | 描述 |
| 检查 RenderDoc 是否可用 |
| 打开 RDC 捕获文件 |
| 关闭当前捕获文件 |
| 获取捕获的详细信息 |
| 获取所有纹理列表 |
| 获取所有缓冲区列表 |
| 获取 Draw Call 列表 |
| 获取着色器详细信息 |
| 获取指定位置的像素数据 |
| 获取纹理缩略图 |
| 分析特定 Draw Call |
| 搜索资源 |
使用示例
在 Claude Desktop 中,你可以这样使用:
请帮我分析 C:\captures\scene.rdc 这个捕获文件中的纹理资源列出这个捕获中的所有 Draw Call搜索名称包含 "normal" 的纹理开发
# 安装开发依赖
pip install -e ".[dev]"
# 运行测试
pytest许可证
MIT License
Available Tools
12 toolsrenderdoc_analyze_draw_callC
Perform detailed analysis of a specific draw call including input/output resources, shaders, and state.
Inputs: event_id (required).
renderdoc_check_availableA
Check if RenderDoc is available on the system. Returns availability status and installation info.
Inputs: no named parameters.
renderdoc_close_captureA
Close the currently open RDC capture file and release resources.
Inputs: no named parameters.
renderdoc_get_buffersB
Get a list of all buffers in the capture with their properties (size, type, etc.).
Inputs: filter.
renderdoc_get_capture_infoB
Get detailed information about the currently open capture including API, driver, and resource counts.
Inputs: no named parameters.
renderdoc_get_draw_callsB
Get a list of all draw calls/events in the capture with their properties.
Inputs: start_index, count.
renderdoc_get_pixel_dataC
Get pixel color data at a specific location in a texture.
Inputs: texture_id (required), x (required), y (required), mip.
renderdoc_get_shader_infoC
Get detailed information about a specific shader including its reflection data.
Inputs: resource_id (required).
renderdoc_get_texturesA
Get a list of all textures in the capture with their properties (dimensions, format, mip levels, etc.).
Inputs: filter.
renderdoc_get_texture_thumbnailA
Get a base64 encoded thumbnail image of a texture. Useful for quickly visualizing texture contents.
Inputs: texture_id (required), mip, max_size.
renderdoc_open_captureA
Open an RDC capture file for analysis. Returns basic information about the capture including API type, frame count, and resource counts.
Inputs: file_path (required).
renderdoc_search_resourcesC
Search for resources (textures, buffers, shaders) by name pattern.
Inputs: pattern (required), resource_type.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Each tool has a clearly distinct purpose targeting specific aspects of RenderDoc capture analysis, such as checking availability, opening/closing captures, retrieving lists of resources, analyzing draw calls, or fetching specific data like pixel colors or thumbnails. There is no significant overlap in functionality, making tool selection straightforward for an agent.
All tool names follow a consistent 'renderdoc_verb_noun' pattern with snake_case, such as 'renderdoc_open_capture' and 'renderdoc_get_textures'. This uniformity enhances readability and predictability, allowing agents to easily infer tool purposes from their names.
With 12 tools, the server is well-scoped for analyzing graphics captures, covering core operations like capture management, resource listing, detailed analysis, and data retrieval. Each tool serves a specific and necessary function without redundancy, making the count appropriate for the domain.
The toolset provides comprehensive coverage for capture analysis, including opening/closing captures, retrieving resource lists, analyzing draw calls and shaders, and accessing pixel data. A minor gap exists in operations like modifying captures or exporting data, but core analysis workflows are fully supported, allowing agents to perform most essential tasks.
Related MCP Connectors
The Polar Signals MCP server enables AI assistants to connect directly with performance profiling data, allowing users to analyze application performance through natural language queries. Key capabilities include querying CPU performance and memory usage, exploring profiling metadata like profile types and labels, and providing AI-driven code optimization suggestions directly within development environments like Claude Code or Cursor.
DXF and PDF/X-4 for AI agents: structured facts, PNG renders, an interactive in-chat viewer.
Read-only AI coding tools for change verification, release readiness, capacity, and guidance.
Savecraft serves real save game data and expert game knowledge to AI assistants.