UK Curriculum MCP Server

An MCP server that gives AI assistants structured access to UK exam specifications — paper structures, assessment objectives, topic breakdowns, and grade boundaries.

12 GCSE specs and growing. New specs added daily via automated extraction.

Available data

| Board | Subject | Code | |-------|---------|------| | AQA | Biology | 8461 | | AQA | Chemistry | 8462 | | AQA | Combined Science Trilogy | 8464 | | AQA | English Language | 8700 | | AQA | English Literature | 8702 | | AQA | Geography | 8035 | | AQA | History | 8145 | | AQA | Mathematics | 8300 | | AQA | Physics | 8463 | | AQA | Religious Studies A | 8062 | | Edexcel | English Language | 1EN0 | | Edexcel | Mathematics | 1MA1 |

Quick start

With Claude Desktop / OpenClaw / any MCP client

{
  "mcpServers": {
    "uk-curriculum": {
      "command": "npx",
      "args": ["-y", "uk-curriculum-mcp"]
    }
  }
}

That's it. The server runs over stdio and exposes 5 tools.

Install globally

npm install -g uk-curriculum-mcp
uk-curriculum-mcp  # starts stdio server

Tools

| Tool | Description | |------|-------------| | list_subjects | List all available subjects, optionally filter by board/level | | get_specification | Get topic hierarchy for a subject (topic areas → subtopics) | | get_paper_structure | Get exam papers — duration, marks, sections, calculator status | | get_assessment_objectives | Get AOs with weighting ranges | | get_grade_boundaries | Get historical grade boundaries by year/tier |

Example

> list_subjects({ board: "AQA" })

{
  "count": 5,
  "subjects": [
    { "board": "AQA", "subject": "Mathematics", "level": "GCSE", "code": "8300" },
    { "board": "AQA", "subject": "English Language", "level": "GCSE", "code": "8700" },
    ...
  ]
}

Data quality

Every data point includes provenance metadata:

{
  "_meta": {
    "source": "aqa-8300-spec.pdf",
    "pageRef": "p.12",
    "confidence": "high",
    "verifiedDate": "2026-03-28",
    "verifiedBy": "agent",
    "tier": "authoritative"
  }
}

• authoritative = directly from exam board spec PDF (always includes pageRef)
• derived = synthesised from multiple sources (may lack pageRef)
• Grade boundaries are historical only — never use them for prediction

Contributing specs

The fastest way to add a new spec:

1. Download the specification PDF from the exam board
2. Run the extraction script: npx tsx scripts/extract-from-pdf.ts <pdf-path> <board> <subject> <level>
3. Validate: npm run validate && npm test
4. Open a PR

See scripts/extract-from-pdf.ts for the extraction pipeline.

Development

git clone https://github.com/AnasInno/uk-curriculum-mcp
cd uk-curriculum-mcp
npm install
npm run build
npm test        # 15 tests
npm start       # run server locally

License

MIT