Laravel MCP Server
A project and task manager MCP server built with Laravel, enabling AI to manage projects and tasks through tools, resources, and prompts.
README
Project Manager — Laravel MCP Server
A small project & task manager exposed as a Model Context Protocol
(MCP) server, built with laravel/mcp. It is a learning-oriented
reference that exercises every core MCP primitive — Tools, Resources, and Prompts —
over a deliberately simple domain (projects that contain tasks).
An MCP client (Claude Desktop, Cursor, Claude Code, or the built-in inspector) connects to the server over stdio and lets an AI create and track tasks, pull read-only context, and run reusable prompt templates.
Domain
- Project —
name,description, and many tasks. - Task — belongs to a project; has a
title,description,due_date, astatus(pending,in_progress,done) and apriority(low,medium,high,critical). A task is overdue when its due date has passed and it is not yet done.
Priorities carry a numeric weight (low = 1 … critical = 4) used to sort reports and triage
overdue work.
Project structure
app/
├── Enums/ TaskStatus, TaskPriority (with label() / weight() helpers)
├── Models/ Project, Task
├── Mcp/
│ ├── Servers/ ProjectManagerServer — registers all primitives
│ ├── Tools/ The 5 callable actions
│ ├── Resources/ The 5 read-only resources
│ └── Prompts/ The 4 prompt templates
└── Support/ ReportBuilder — shared report aggregation
MCP primitives
🔧 Tools — actions the AI can invoke
| Tool | Description |
|---|---|
create-task-tool |
Create a task under a project (title, description, priority, due date). |
list-tasks-tool |
List tasks with optional filters: project, status, priority, overdue. |
update-task-status-tool |
Move a task to a new status and report the before → after change. |
delete-task-tool |
Delete a task by id. |
generate-report-tool |
Aggregate tasks into a portfolio report as Markdown or JSON. |
📚 Resources — read-only context addressed by URI
| Resource | URI | Description |
|---|---|---|
| System documentation | documentation |
Onboarding overview of the server and its entities. |
| Priority rules | priority-rules |
The priority levels, their weights, and when to use each. |
| Task details | task://tasks/{task_id} |
Full detail for a single task (templated). |
| Project summary | project://projects/{project_id} |
Task counts by status, overdue count, next due (templated). |
| Report | report://reports/{format} |
Whole-portfolio report in json or markdown (templated). |
The report tool and the report resource share a single ReportBuilder
so their numbers can never drift apart.
💬 Prompts — reusable, argument-driven templates
| Prompt | Arguments | Description |
|---|---|---|
plan-project-prompt |
goal (required), deadline |
Break a goal into a project plus milestone tasks. |
decompose-task-prompt |
task_id (required) |
Split one real task into concrete subtasks. |
review-overdue-tasks-prompt |
project_id |
Triage and re-prioritize overdue tasks. |
daily-summary-prompt |
project_id |
Produce a stand-up style daily summary. |
Getting started
composer install
php artisan migrate
# Seed at least one project so tasks have somewhere to live:
php artisan tinker --execute "App\Models\Project::factory()->create(['name' => 'Website Redesign']);"
The server is registered in routes/ai.php and exposed over stdio:
Mcp::local('project-manager', ProjectManagerServer::class);
Running & testing the server
# Interactive debugging — drive tools, resources and prompts by hand:
php artisan mcp:inspector project-manager
# Run the server over stdio (this is the command MCP clients launch;
# on its own it will appear to hang, waiting for JSON-RPC on stdin):
php artisan mcp:start project-manager
Connecting from an AI client
Claude Code
claude mcp add project-manager -- php "<full-path>/artisan" mcp:start project-manager
Claude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"project-manager": {
"command": "php",
"args": ["<full-path>/artisan", "mcp:start", "project-manager"]
}
}
}
Passing the full path to artisan lets Laravel resolve the project root regardless of the
client's working directory.
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。