mcp-graph-workflow

概述

mcp-graph-workflow 是一个本地优先的 CLI 工具，通过将 PRD 文本文件转换为 SQLite 任务图，提供持久化任务管理。它提供 54 个 MCP 工具，通过从分析到监听的 9 个阶段管理任务生命周期。该解决方案解决了 AI 编码常见问题，提供持久化内存、结构化工作流和可追溯性，无需云依赖或 API 密钥。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• import_prd

  Reads PRD text file and creates tasks with acceptance criteria in the graph

  试试:Import requirements from my-prd.md into the workflow graph试试:Convert my product requirements document into tasks
• start_task

  Starts working on a specific task from the graph with compressed context

  试试:Start working on task 42试试:Begin implementing feature from my graph
• finish_task

  Finishes a task after validating done criteria and updates status

  试试:Mark task 15 as done with validation试试:Complete the current implementation task
• next

  Gets the next available task for the agent to work on

  试试:What's the next task I should work on?试试:Show me the next task in my workflow
• status

  Shows the current state of the task graph

  试试:Show me the current status of all tasks试试:What tasks are in progress?

可对比工具

安装

npm install -g @mcp-graph-workflow/mcp-graph

{
  "mcpServers": {
    "mcp-graph": {
      "command": "npx",
      "args": ["-y", "@mcp-graph-workflow/mcp-graph"]
    }
  }
}

mcp-graph init

FAQ

这需要云服务或 API 密钥吗？

不需要，它设计为 100% 本地运行，零云依赖。LLM 集成是可选的。

这可以与我现有的 IDE 集成吗？

可以，它提供 130 多个 REST 端点用于 IDE 集成，并包含用于可视化的 React 仪表板。

mcp-graph-workflow 对比