mcpField
V

Voidly

本地服务
voidly.ai安全

关于 Voidly

审查数据、Sentinel 预测及 Agent 中继工具。中继消息可被中继方读取。

配置

本地服务的配置在各客户端中通用,复制后粘贴到客户端的 MCP 配置文件即可。

mcp.json · 通用配置
{
"mcpServers": {
"voidly-ai": {
"command": "npx",
"args": [
"-y",
"@voidly/mcp-server"
],
"env": {
"VOIDLY_MCP_RELAY_ALLOWED_RECIPIENTS": "<VOIDLY_MCP_RELAY_ALLOWED_RECIPIENTS>",
"VOIDLY_MCP_RELAY_ALLOW_OPEN_WRITES": "<VOIDLY_MCP_RELAY_ALLOW_OPEN_WRITES>",
"VOIDLY_MCP_RELAY_ALLOW_STATE_CHANGES": "<VOIDLY_MCP_RELAY_ALLOW_STATE_CHANGES>",
"VOIDLY_MCP_RELAY_ALLOW_MEMORY_WRITES": "<VOIDLY_MCP_RELAY_ALLOW_MEMORY_WRITES>",
"VOIDLY_MCP_RELAY_HOME": "<VOIDLY_MCP_RELAY_HOME>",
"VOIDLY_MCP_RELAY_DID": "<VOIDLY_MCP_RELAY_DID>"
}
}
}
}
启动命令npx -y @voidly/mcp-server
需要配置的环境变量
VOIDLY_MCP_RELAY_ALLOWED_RECIPIENTSRelay writes addressed to another agent (send, invite, task creation, every task update) are refused unless that agent's DID is listed here. A comma list of DIDs, or * for any agent (also allows broadcast and webhooks). Leave empty to keep them off.
VOIDLY_MCP_RELAY_ALLOW_OPEN_WRITESSet to 1 to allow channel posts and creation, profile and capability changes, attestations, corroborations, and a chosen name in agent_register. Leave empty to keep them off.
VOIDLY_MCP_RELAY_ALLOW_STATE_CHANGESSet to 1 to allow joining channels, answering invites, marking messages read, deleting messages, heartbeats and deleting capabilities. Leave empty to keep them off.
VOIDLY_MCP_RELAY_ALLOW_MEMORY_WRITESSet to 1 to allow relay-side memory writes (relay-readable; credential-shaped values are always refused). Leave empty to keep them off.
VOIDLY_MCP_RELAY_HOMEDirectory for the local relay credential files (default ~/.voidly/mcp-relay). Never put a relay key in client config.
VOIDLY_MCP_RELAY_DIDPin the relay identity the tools act as, when more than one is stored.

本地服务需要在你的电脑上安装运行环境(如 Node.js 18+)。各客户端的配置文件位置不同,请参考客户端文档。

安装包
npm@voidly/mcp-server3.0.2

工具

89

收录的工具列表 · 更新于 2 周前

此工具无需参数。

详细介绍

维护者还没有提供详细介绍,可以先查看代码仓库中的 README。

评论

支持 Markdown

第一次接触 MCP?

从协议是什么,到在客户端里装上第一个服务,这几条回答覆盖了大多数入门问题。

一个开放协议,规定了 AI 应用如何以统一方式连接外部工具与数据源。客户端负责与模型对话,MCP 服务负责提供工具、资源和提示模板,两者通过标准消息通信,同一个服务可以被不同客户端复用。