roadmap-skill
Enables AI agents and humans to collaboratively plan and manage tasks with a shared kanban and dependency graph, all stored locally.
README
Roadmap Skill
<p align="center"> <img src="./imgs/roadmap-skill-web.png" alt="Roadmap Skill web workspace with kanban and graph views" width="900"> </p>
<p align="center"> <sub>The local planning workspace: status in Kanban, execution order in Graph View.</sub> </p>
<p align="center"> <strong>Shared roadmap for humans and AI Agents</strong><br> Visual kanban and graph views for you. MCP server for Agents. Local-first. </p>
<p align="center"> <a href="README.zh.md">简体中文</a> | English </p>
<p align="center"> <a href="https://www.npmjs.com/package/roadmap-skill"><img src="https://img.shields.io/npm/v/roadmap-skill" alt="npm version"></a> <a href="https://github.com/shiquda/roadmap-skill/blob/main/LICENSE"><img src="https://img.shields.io/github/license/shiquda/roadmap-skill" alt="License"></a> <a href="https://www.npmjs.com/package/roadmap-skill"><img src="https://img.shields.io/npm/dm/roadmap-skill" alt="npm downloads"></a> </p>
<p align="center"> <a href="https://insiders.vscode.dev/redirect/mcp/install?name=roadmap&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22roadmap-skill%22%5D%7D"><img src="https://img.shields.io/badge/VS_Code-Install%20MCP%20Server-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white" alt="Install in VS Code"></a> <a href="https://insiders.vscode.dev/redirect/mcp/install?name=roadmap&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22roadmap-skill%22%5D%7D&quality=insiders"><img src="https://img.shields.io/badge/VS_Code_Insiders-Install%20MCP%20Server-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white" alt="Install in VS Code Insiders"></a> </p>
AI-readable docs: llms.txt · llms-full.txt · Architecture · Tool Interface Standard
What Makes It Different
Shared context — Chat with your Agent to plan tasks, or edit directly in the Kanban — both stay in sync. Your Agent always sees the latest state.
Agent-native — Built as an MCP server. Your AI can create, update, and query tasks directly in conversation.
Visual workspace — Open localhost:7860 when you feel like it, or ask your Agent to open it with open_web_interface. Kanban for status, Graph View for sequencing. Or just stay in chat. All three work.
Local-first — Your data lives on your machine. No accounts, no cloud sync, no vendor lock-in.
New in v0.3.0: Graph View
Graph View is one of the headline features in v0.3.0.
<p align="center"> <img src="./imgs/roadmap-graph.png" alt="Roadmap Skill graph view for dependency planning" width="900"> </p>
<p align="center"> <sub>Graph View turns roadmap tasks into a dependency map, so you can inspect order, blockers, and ready work at a glance.</sub> </p>
Kanban is great for tracking status. Graph View is for understanding execution order.
It helps you and your Agent answer questions like:
- What has to happen first?
- Which tasks are ready right now?
- What is blocked by unfinished work?
- Which dependency should be added, reversed, or removed?
Instead of treating roadmap tasks as a flat list, Graph View turns them into a visual dependency map with node connections, ready/done states, detailed task cards, and export support. It is designed for planning sessions where sequencing matters just as much as status.
If you are using Roadmap Skill as an MCP workspace, this is the view that makes agent-assisted planning feel concrete: your Agent can build and update dependency graphs in chat, and you can immediately inspect the structure visually in the web UI.
Quick Start (Zero Config)
# Integrate with your AI assistant (MCP config below)
<details> <summary><b>View platform configurations</b></summary>
<details> <summary><b>Claude Code</b></summary>
claude mcp add roadmap "npx -y roadmap-skill"
</details>
<details> <summary><b>Claude Desktop</b></summary>
{
"mcpServers": {
"roadmap": {
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>Copilot</b></summary>
/mcp add
Or edit ~/.copilot/mcp-config.json:
{
"mcpServers": {
"roadmap": {
"type": "local",
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>Cursor</b></summary>
{
"mcpServers": {
"roadmap": {
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>JetBrains (via Cline)</b></summary>
Install Cline plugin, then add to cline_mcp_settings.json:
{
"mcpServers": {
"roadmap": {
"type": "stdio",
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>Roo Code</b></summary>
{
"mcpServers": {
"roadmap": {
"type": "stdio",
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>Opencode</b></summary>
Edit ~/.config/opencode/opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"roadmap": {
"type": "local",
"command": [
"npx",
"-y",
"roadmap-skill"
],
"enabled": true
}
}
}
</details>
<details> <summary><b>Trae</b></summary>
Go to Settings -> MCP -> Add new MCP Server:
{
"mcpServers": {
"roadmap": {
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>VS Code</b></summary>
{
"mcp": {
"servers": {
"roadmap": {
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
}
</details>
<details> <summary><b>Windsurf</b></summary>
Follow Windsurf MCP documentation and use:
{
"mcpServers": {
"roadmap": {
"command": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
</details>
<details> <summary><b>Zed</b></summary>
{
"context_servers": {
"roadmap": {
"command": {
"path": "npx",
"args": ["-y", "roadmap-skill"]
}
}
}
}
</details>
<details> <summary><b>Codex CLI / Codex IDE</b></summary>
One-liner (recommended):
codex mcp add roadmap -- npx -y roadmap-skill
Manual config — edit ~/.codex/config.toml:
[mcp_servers.roadmap]
command = "npx"
args = ["-y", "roadmap-skill"]
Windows: No
cmd /cwrapper needed — Codex handles the subprocess directly.
Server name: Must match
^[a-zA-Z0-9_-]+$. Names with spaces or parentheses are rejected.
Note: Both Codex CLI and the Codex VSCode extension share the same
~/.codex/config.toml. A syntax error breaks both.
</details>
</details>
<br>
After setup, simply tell your AI:
"Create a website redesign project with tasks for homepage, about page, and contact form"
The AI will immediately create the project and tasks, saving them to local storage. You can open http://localhost:7860 anytime to view the visual Kanban board.
Agent Skills
This repository includes installable skills under skills/, distributed through the git repository rather than the published npm package.
Available skills:
skills/roadmap/- lightweight background and routing for roadmap-skillskills/roadmap-task-flow/- capture, enrich, tag, and prioritize roadmap tasksskills/roadmap-planning-views/- build focused planning graphs from task subsetsskills/roadmap-web-visualization/- open and use the local roadmap web workspace
Recommended installation pattern for skill-aware tools:
# install from the git repository when the installer supports repo sources
npx skills add shiquda/roadmap-skill
These skills are still in beta. If you run into issues or have suggestions, feedback is very welcome.
Who Is This For?
Solo developers, AI power users, and vibe coders who want their Agent to be a real collaborator — not just a code generator.
Typical Use Cases
1. Capture Ideas Without Breaking Flow
Mid-session, something pops into your head:
"Note this down: switch the auth module to JWT + refresh token rotation"
Use /quick-capture or just say "help me note this down" — your Agent turns it into a task, saves it locally, and you keep going. No context switch, no forgotten ideas.
2. Let Agent Help You Prioritize
"This month's OKR is finishing the user system refactor — can you list the related tasks and figure out which ones should come first?"
Use /auto-prioritize to let your Agent analyze the current project state and reorder task priorities accordingly.
3. Let Agent Recommend What to Do Next
"What's left on the current project? Any high-priority tasks I haven't touched yet?"
Or use /suggest-tasks — your Agent recommends the next actions based on current progress and priorities.
No tab switching. No copy-pasting. Just ask
Core Features
| Feature | Description |
|---|---|
| Kanban View | Four columns: Todo, In Progress, Review, Done |
| Graph View | Visual dependency map for sequencing work, spotting blockers, and reviewing execution order |
| Ready / Blocked Signals | Quickly see which tasks are ready now and which ones still depend on unfinished work |
| Dependency Editing | Add, reverse, and remove task dependencies directly from the planning graph |
| Graph Export | Export dependency graphs as images for sharing, review, or release planning |
| Drag and Drop | Drag tasks to update status, WYSIWYG |
| Quick Create | Click "+" button on any column to instantly add tasks |
| Dual View | Compact mode for overview, detailed mode for full info |
| Smart Search | Quick task search across all projects, keyword filtering |
Supported Platforms
- ✅ Claude Code, Codex, OpenCode — Command-line AI assistants
- ✅ Cursor, VS Code — AI-powered IDEs
- ✅ Any MCP Client — Standard MCP protocol support
Data Storage
Roadmap Skill uses pure local JSON file storage. All data is saved in your user directory:
| Platform | Storage Path |
|---|---|
| Windows | %USERPROFILE%\.roadmap-skill |
| macOS | ~/.roadmap-skill/ |
| Linux | ~/.roadmap-skill/ |
You can export or import backup files from the web interface.
License
MIT © shiquda
<p align="center"> <strong>If this project helps you, please consider giving it a star!</strong> </p>
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。