openclaw-output-vetter-mcp
MCP server for verifying AI agent claims vs reality — single-transcript inline grounding-check that flags when an agent's response states facts not in the input context, when its code silently swallows exceptions and substitutes mock data, or when its multi-turn transcript contains contradictions or unverified completion claims. Sub-second, local, free, no API calls.
代码分析
0 个平台
0 项功能
安装
uvx openclaw-output-vetter-mcp
把上面的配置写入 MCP 客户端的配置文件（Claude Desktop 为 claude_desktop_config.json），重启客户端后生效。
{
"mcpServers": {
"openclaw-output-vetter-mcp": {
"command": "uvx",
"args": [
"openclaw-output-vetter-mcp"
]
}
}
}把上面的配置写入 MCP 客户端的配置文件（Claude Desktop 为 claude_desktop_config.json），重启客户端后生效。
数据来源：公开的 MCP Server 目录。本站为独立第三方目录，与各 MCP Server 维护方均无隶属或背书关系；名称与简介如实标注，介绍文案由本站再加工。
