Skip to main content

MCP server for RenderDoc graphics debugger - enables AI-assisted frame capture analysis

Project description

renderdoc-mcp

MCP server for RenderDoc — let AI assistants analyze GPU frame captures (.rdc files) for graphics debugging and performance analysis.

Platform: Windows x64 only.

Python: Requires Python 3.13 exactly. The bundled renderdoc.pyd is compiled for Python 3.13, and the compile-time and runtime Python versions must match exactly.

Prerequisites

# verify
python --version   # Python 3.13.12
uv --version       # uv 0.10.9

Configuration

No manual installation needed. Add to your MCP client configuration and uvx will handle the rest:

Kiro (~/.kiro/settings/mcp.json):

{
  "mcpServers": {
    "renderdoc": {
      "command": "uvx",
      "args": ["--python", "3.13", "renderdoc-mcp"]
    }
  }
}

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "renderdoc": {
      "command": "uvx",
      "args": ["--python", "3.13", "renderdoc-mcp"]
    }
  }
}

--python 3.13 is required — the bundled renderdoc.pyd is compiled for Python 3.13 and the compile-time and runtime versions must match exactly.

Usage

Just talk to your AI assistant:

"简单介绍下 renderdoc-mcp 所有的工具及作用"

"打开 frame.rdc 看看这一帧在做什么"

"找到渲染角色模型的 draw call,检查它的管线状态"

"为什么我的阴影贴图是黑的?检查深度 pass"

"分析性能,有没有冗余的 draw call?"

Tools (42)

Session — open/close captures, frame overview, GPU quirk detection

Event Navigation — list/search/navigate draw calls and actions

Pipeline Inspection — full pipeline state, shader bindings, vertex inputs, one-call draw analysis

Resource Analysis — enumerate textures, buffers, resources and their usage

Data Extraction — export textures/meshes/render targets, read pixels, buffer data

Shader Analysis — disassembly, reflection, constant buffer contents

Advanced — pixel history, draw call diff, render pass analysis, pixel region scan, per-pixel shader debug

Performance — pass timing, overdraw, bandwidth, state change analysis

Diagnostics — negative/NaN/Inf detection, precision issues, reflection mismatch, mobile GPU risks

Prompts

Prompt Description
debug_draw_call Deep-dive a single draw call
find_rendering_issue Systematic diagnosis from a problem description
analyze_performance Frame-wide performance analysis

How It Works

AI Assistant <--MCP--> renderdoc-mcp <--Python API--> renderdoc.pyd <-> GPU replay

All required RenderDoc DLLs and renderdoc.pyd are bundled — no separate RenderDoc installation needed.

Troubleshooting

ImportError: DLL load failed — verify Python 3.13 is being used; check antivirus is not blocking bundled DLLs.

Server won't start — check MCP client logs; confirm python --version is 3.13.

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

renderdoc_mcp-0.2.3.tar.gz (28.1 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

renderdoc_mcp-0.2.3-py3-none-any.whl (28.2 MB view details)

Uploaded Python 3

File details

Details for the file renderdoc_mcp-0.2.3.tar.gz.

File metadata

  • Download URL: renderdoc_mcp-0.2.3.tar.gz
  • Upload date:
  • Size: 28.1 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for renderdoc_mcp-0.2.3.tar.gz
Algorithm Hash digest
SHA256 9facd3dc891f10d33f68347af1149c1a82086a4711ab30e81fa696adc6292acb
MD5 6580fbb2e33e64e8618952fc11dd2fe5
BLAKE2b-256 5b17dece6cd3cb347155b7e48ec4f94714dca40234a48f5a42a39f0acdf4a3ce

See more details on using hashes here.

File details

Details for the file renderdoc_mcp-0.2.3-py3-none-any.whl.

File metadata

  • Download URL: renderdoc_mcp-0.2.3-py3-none-any.whl
  • Upload date:
  • Size: 28.2 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for renderdoc_mcp-0.2.3-py3-none-any.whl
Algorithm Hash digest
SHA256 3f6bd5aab6e64da04600f8fe3f3a741983828a8dbf42a9f22c9805c053936e8c
MD5 332b90bcfe2a48d81438e9bf7b7a4e14
BLAKE2b-256 0bdb5e42e4cb6c84f018ebb1bbe7e507ed861492eb0524555f7e86fe74c01926

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page