ActionDock
Documentation | English | 简体中文
接入你的智能体
發布者自己的指令。不是 PlugMCP 端點。
用我們的金鑰與日誌託管一份副本。
npx -y @actiondock/[email protected]Claude Code 沒有本機指令的安裝連結。複製下方指令並在終端機中執行。
claude mcp add actiondock-mcp -- npx -y @actiondock/mcp@2.22.0在終端機中執行。
{ "mcpServers": { "actiondock-mcp": { "command": "npx", "args": [ "-y", "@actiondock/[email protected]" ] } }}claude_desktop_config.json
遠端 URL 還是本機指令? 兩種安裝方式的差異
來自發布者
Documentation | English | 简体中文
ActionDock MCP Adapter for exposing Actions as Model Context Protocol (MCP) Tools
Write once. Deliver as CLI, MCP, HTTP, and Agent Skill.
An engineering toolchain for developing, testing, building, and distributing AI agent actions and skills. Built-in in-memory testing sandboxes, persistent state, config fallback, run tracking, and reproducible packaging.
# Install official skill for AI agents (Cursor, Claude Code, Antigravity)
npx skills add team4u/actiondock -g -y
# Or install CLI toolchain globally and initialize
npm install -g @actiondock/cli
ad init hello && cd hello
# Scaffold Action and generate typed contracts
ad action create greet --input name:string --output message:string
# Run sub-second in-memory sandbox tests
ad test
# Verify execution locally
ad run greet -- name=World
# Multi-target instant delivery
ad mcp # MCP server (for Cursor, Windsurf, Claude Desktop)
ad serve # RESTful HTTP microservice (:8080)
ad export skill # Agent skill bundle (for Claude Code, Codex, Antigravity)
node:sqlite embedded storage, and node:http microservice engine with zero transpilation overhead.Install official Agent Skill (recommended for AI agents): Equip your AI programming agents (such as Cursor, Claude Code, Codex, Antigravity) to write, test, and ship Actions autonomously:
npx skills add team4u/actiondock -g -y
Install CLI and initialize project:
npm install -g @actiondock/cli
ad init hello
cd hello
Scaffold an Action: CLI parses field declarations, creates typed contracts, and updates manifest:
ad action create greet --input name:string --output message:string
Implement business logic:
Consume generated types directly in actions/greet.ts:
import { defineAction } from "@actiondock/sdk";
import type { ActionInput, ActionOutput } from "../.actiondock/generated/actions.d.ts";
export type Input = ActionInput<"greet">;
export type Output = ActionOutput<"greet">;
export default defineAction<Input, Output>(async (input, ctx) => {
ctx.log.info("Greeting user", input);
return {
message: `Hello, ${input.name}!`,
};
});
Test and run locally:
# Execute in-memory sandbox tests
ad test
# Run locally via CLI
ad run greet -- name=World
Multi-target instant delivery:
# Standard MCP protocol server
ad mcp
# Production RESTful HTTP microservice
ad serve
# Self-contained Agent Skill bundle
ad export skill
ActionDock is structured as a modular monorepo:
For full guides and references, visit the Documentation Center or explore local guides:
Apache-2.0 License.
ActionDock MCP Adapter for exposing Actions as Model Context Protocol (MCP) Tools
{ "mcpServers": { "actiondock-mcp": { "command": "npx", "args": [ "-y", "@actiondock/[email protected]" ] } }}工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
@actiondock/mcp 透過 stdio 在本機執行。
上次檢查時間:2026-11-13。 ActionDock MCP Adapter for exposing Actions as Model Context Protocol (MCP) Tools