mcp-gdbmi
Enables debugging of programs through GDB/MI, allowing breakpoints, stepping, and register analysis via natural language.
README
MCP server for GDB/MI
An MCP server for interacting with GDB/MI (GDB Machine Interface).
Output from GDB/MI is "compressed" using json5 to reduce the token count for
data structures. Hex values are also "compressed" by stripping out long
0x00000000000 prefixes ie 0x0000004 becomes 0x4
[!CAUTION] GDB/MI can be very verbose, and this tool will destroy your quota VERY quickly if you are not careful with what you ask for! You have been warned.
Installation
Clone the repo and install the deps:
pip install .
Setup
Claude CLI
# run the mcp server
fastmcp run gdb_mcp.py:mcp --transport http --port 3333
# add it to claude
claude mcp add --transport http gdb http://localhost:3333/mcp
Gemini CLI (FastMCP and uv)
This command simplifies the process and makes the server capabilities instantly available and configured within Gemini CLI.
fastmcp install gemini-cli gdb_mcp.py
Gemini CLI (Manual)
.gemini/settings.json
{
"mcpServers": {
"GDB MCP Server": {
"command": "fastmcp",
"args": ["run", "/absolute/path/to/mcp-gdbmi/gdb_mcp.py:mcp"]
}
}
}
Usage
Start your program with gdbserver:
gdbserver localhost:1234 ./program
Then use the tools through your MCP client:
Ask Claude/Gemini to:
- Set breakpoints
- Run the program
- Step through code
- Analyse registers
- Anything that is possible via the GDB machine interface
Tools
connect(target: str)
Connect to GDB, which connects to the target (e.g., "localhost:1234" for a gdbserver instance).
disconnect()
Disconnect and terminate the GDB session.
command(command: str, wait_for_done: bool = True)
Execute a GDB/MI command (e.g., "-break-insert main", "-exec-run").
Common GDB/MI Commands
-break-insert LOCATION- Set breakpoint-break-delete NUM- Delete breakpoint-exec-run- Start program-exec-continue- Continue execution-exec-step/-exec-next- Step through code-stack-list-frames- Show stack frames-data-evaluate-expression EXPR- Evaluate expression
See GDB/MI docs for more.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。