tei-mcp: TEI P5 for AI Agents

If you’ve ever used an AI coding assistant to write TEI XML, you’ve probably noticed that it gets things wrong. Elements appear where they shouldn’t. Attributes are invented. Nesting rules are ignored. The model has a rough sense of what TEI looks like, but no reliable knowledge of the specification.

tei-mcp solves this by giving AI agents direct, tool-based access to the TEI P5 Guidelines.

What is MCP?

The Model Context Protocol (MCP) is an open standard that allows AI applications to connect to external data sources and tools. Think of it as a USB port for AI: a single protocol that lets any compatible client — Claude, Cursor, Windsurf, and others — plug into specialised services.

An MCP server exposes tools that the AI can call during a conversation. Instead of relying on memorised training data, the model can query a live, authoritative source.

What tei-mcp does

tei-mcp parses the TEI P5 ODD specification and exposes 16 tools that cover the most common questions an editor or encoder would ask:

• What is this element? Look up any element, class, macro, or module by name, with case-insensitive matching and typo suggestions.
• What attributes does it take? Resolve attributes across the full class hierarchy — local attributes first, then inherited ones in order.
• What can go inside it? Expand content models into structured trees, or get a flat list of valid children.
• Can this element go here? Check parent-child nesting, or trace reachability through the full element hierarchy.
• Is my document valid? Validate a TEI XML file against the spec: content models, attribute values, closed value lists, reference integrity, and deprecation warnings.
• What about my project schema? Load an ODD customisation file to constrain all of the above to your project’s specific subset of TEI.

Why it matters

TEI encoding requires constant reference to the Guidelines. Experienced encoders internalise the most common patterns, but even they need to check the spec for less familiar elements or complex content models. For AI assistants, which have no such internalised knowledge, the problem is worse: they hallucinate plausible-looking but incorrect markup.

With tei-mcp, the AI doesn’t have to guess. It can look up the answer in the specification before writing a single angle bracket. The result is markup that conforms to TEI P5 — or to your project’s ODD customisation.

Getting started

Install from PyPI:

pip install tei-mcp

Then add it to your MCP client’s configuration:

{
  "mcpServers": {
    "tei": {
      "command": "uvx",
      "args": ["tei-mcp"]
    }
  }
}

The server downloads the TEI specification on first run and works with any MCP-compatible client.

Source code and full documentation: github.com/Pantagrueliste/tei-mcp