PlainGov-MCP
Retrieves and explains government program information from official Canadian sources using a strict retrieval-first approach, with deterministic eligibility checks and full source attribution.
README
Retrieval-First MCP Server
Tagline: "Government information, retrieved and explained."
This MCP server provides government program information based solely on freshly retrieved documents from pre-approved sources. It operates on a strict retrieval-first principle, ensuring all responses are grounded in official sources without prior knowledge or assumptions.
Quick Start
npm install
npm run build
npm run inspector # Test the server
See Deployment Guide for complete installation instructions.
What This MCP Does
- Retrieves verbatim text from pre-approved government URLs
- Provides retrieved text directly from official sources
- Performs conservative eligibility assessments using rules engine
- Includes source URLs and verification dates in all outputs
Strict compliance: No summaries, explanations, or answers without retrieval. Returns verbatim retrieved text only.
Scope (v1)
- Country + Province: Canada + Alberta
- Domains: Taxes & Benefits, Small Business Administration
- Programs: GST Credit, Canada Child Benefit, Alberta Family Employment Tax Credit, GST Registration, Payroll Deductions
Tools Exposed
explain_program(program_id)- Retrieved text from official sourceget_eligibility_criteria(program_id)- Retrieved text from official sourceeligibility_check(program_id, user_context)- Conservative eligibility assessment using rules enginegenerate_checklist(program_id)- Retrieved text from official sourcetimeline(program_id)- Retrieved text from official sourcequestions_for_professional(program_id)- Retrieved text from official source
Data Sources
- Canada.ca
- CRA official publications
- Alberta.ca program pages
- Public PDFs, bulletins, guides
No scraping forums. No third-party "tips".
Architecture
- Retrieval-First Principle: All responses based on freshly retrieved documents from pre-approved URLs
- Verbatim Storage: Retrieved text stored exactly as received
- Retrieval Operation: Only verbatim text from pre-approved sources (no external data)
- Rules Engine: Deterministic eligibility checks (no LLM involvement)
- Error Handling: Retrieval failures return errors only, no fallbacks
- Output Requirements: All responses include source URL and last_verified date
Design choice: Retrieval only. Fail loudly on constraint violations.
What We Explicitly DO NOT Build
❌ "How to minimise tax"
❌ "Best way to qualify"
❌ Edge-case exploitation
❌ Advice framed as instruction
We are a translator, not a strategist.
This keeps us out of trouble and attractive to institutions.
Pricing Strategy
- Individual: $8–12/month CAD
- Small business: $19–29/month
- Professional / org: Seat-based or bulk licensing
Cheap insurance against mistakes.
Documentation
Getting Started
- Deployment Guide - Complete installation and configuration instructions
- Usage Examples - Real-world scenarios and tool demonstrations
Technical Documentation
- Architecture - Technical design, data flow, and design decisions
- API Reference - Tool specifications, parameters, and schemas
- Testing Guide - Testing procedures and validation
Support & Contributing
- Troubleshooting - Common issues and solutions
- Contributing Guidelines - How to contribute to the project
Installation
Quick Install
# 1. Install dependencies
npm install
# 2. Build the server
npm run build
# 3. Test with Inspector
npm run inspector
MCP Client Configuration
Add to your MCP client configuration (e.g., Claude Desktop):
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"plain-gov-mcp": {
"command": "node",
"args": [
"/absolute/path/to/PlainGov-MCP/build/index.js"
]
}
}
}
Important: Use absolute paths and restart your MCP client after configuration.
See Deployment Guide for detailed instructions.
Usage
Available Tools
- explain_program - Get program information from official sources
- get_eligibility_criteria - Retrieve eligibility requirements
- eligibility_check - Check eligibility based on user context
- generate_checklist - Get application checklist
- timeline - Get key dates and deadlines
- questions_for_professional - Get questions to ask professionals
Example Usage
Check GST Credit eligibility:
User: "Am I eligible for GST Credit? Income: $45,000, live in Canada"
Server uses eligibility_check tool:
- Status: eligible
- Includes source URL and verification date
- Provides disclaimer
See Usage Examples for comprehensive scenarios.
Features
Retrieval-First Architecture
- All responses based on freshly retrieved documents
- No cached content or prior knowledge
- Verbatim text from official sources
- Source attribution in every response
Conservative Eligibility Assessment
- Deterministic rules engine (no LLM)
- Errs on side of caution
- Clear status indicators (eligible/not_eligible/unclear)
- Identifies missing information
Compliance & Trust
- Strict adherence to retrieval-first constraints
- Fail loudly on errors (no fallbacks)
- Complete source traceability
- Regular verification dates
See Architecture Documentation for technical details.
Compliance with Constraints
- No summaries/explanations without retrieval: All content based on freshly retrieved documents
- Retrieval restrictions: Only verbatim text from pre-approved sources, no prior knowledge
- Retrieval failures: Return error objects only, no fallbacks
- Output standards: Include source URL and last_verified date in all responses
- Rules engine: Eligibility uses deterministic logic, no LLM
- Fail loudly: System errors on any constraint violation
Ethical Guidelines
- No advice or interpretation
- Conservative eligibility assessments
- Direct to official sources only
- Clear disclaimers: "This is not advice"
- No optimization or loophole suggestions
Testing
Manual Testing with Inspector
npm run inspector
Open the provided URL in your browser to test tools interactively.
Automated Testing
See Testing Guide for comprehensive testing procedures.
Troubleshooting
Common issues and solutions are documented in the Troubleshooting Guide.
Quick fixes:
- Server not appearing? Check configuration file syntax and restart client
- Retrieval failures? Verify internet connection
- Tools not working? Test with MCP Inspector
Contributing
We welcome contributions that align with the retrieval-first principles and core mission.
Before contributing:
- Read Contributing Guidelines
- Review Architecture Documentation
- Understand constraint compliance requirements
Ways to contribute:
- Add new government programs
- Improve documentation
- Report bugs
- Suggest enhancements
See CONTRIBUTING.md for detailed guidelines.
Project Status
Current Version: 0.1.0 Status: Production-ready with comprehensive documentation
Completed:
- ✅ Core retrieval-first implementation
- ✅ 5 government programs (Canada & Alberta)
- ✅ 6 tools for program information
- ✅ Deterministic eligibility rules engine
- ✅ Comprehensive documentation suite
- ✅ MCP Inspector integration
- ✅ Error handling and compliance mechanisms
Roadmap:
- LLM integration for formatting (with strict constraints)
- Additional provinces and programs
- Caching layer with time-based invalidation
- Enhanced rules engine
- Automated testing suite
Support
Documentation: Start with Deployment Guide Issues: Check Troubleshooting Guide Questions: Review Usage Examples Technical: See Architecture Documentation
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgments
Built with:
- Model Context Protocol (MCP)
- TypeScript and Node.js
- Official Canadian government sources
Last Updated: 2026-01-10
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。