Inclusio MCP

What inclusio-mcp MCP server does

The inclusio-mcp MCP server gives MCP clients an interface to Inclusio, a Python-based publishing engine for LaTeX documents. Its primary workflow is document discovery, rendering, and PDF accessibility validation. Agents can inspect the registered content tree, select a document, produce an output artifact, and check a PDF against the configured audit process.

The server provides four tools:

• list_docs returns the documents registered in the content tree.
• doc_count returns the number of available documents.
• render builds a tagged, conformant PDF or another supported output format for a selected document.
• audit_pdf runs an accessibility audit on a PDF using veraPDF.

Three read-only resources are also available: inclusio://meta for the project manifest, inclusio://audit/latest for the most recent audit report, and inclusio://version for the engine version information.

How it works

The server operates against an Inclusio content tree containing registered documents. A client can first call list_docs or doc_count, then pass a document to render. The rendering pipeline uses the Inclusio engine and its LaTeX toolchain to create tagged output. The documented PDF target supports PDF/UA-2, WTPDF, and PDF/A-4f conformance.

For validation, audit_pdf uses veraPDF to inspect a generated PDF. The README describes this as the strict accessibility audit gate; a failed check causes the corresponding command-line validation flow to report failure. Multi-format output, including HTML5, JATS XML, and EPUB3, depends on the broader Inclusio installation and its optional Pandoc dependency.

The inclusio-mcp MCP server can run through standard input and output, or through Streamable HTTP. The command examples use inclusio-mcp for stdio and inclusio-mcp --http --port 8765 for HTTP.

Setup and configuration

Install the MCP extra with pip install 'inclusio[mcp]'. The project requires Python 3.11 or newer and a LuaLaTeX toolchain on PATH. Linux, macOS, and WSL are supported; native Windows supports the Python interface, while the LaTeX step requires WSL or a TeX Live installation.

Set INCLUSIO_CONTENT_DIR to the absolute path of the content tree when configuring a client. The README shows a Claude Code configuration using the inclusio-mcp executable and this environment variable. Install veraPDF separately if PDF auditing is needed. Install Pandoc when using the HTML5, JATS XML, or EPUB3 emission paths.

Tools and capabilities

The inclusio-mcp MCP server is suited to agent workflows that need to discover available publication inputs, create accessibility-oriented outputs, and inspect the latest audit state. It does not expose the engine’s full command-line surface as MCP tools: the listed MCP interface covers document listing, counting, rendering, and PDF auditing.

The underlying project also includes provenance and signing features, but the supplied MCP tool list does not identify separate MCP operations for C2PA, PAdES, or SLSA. Likewise, the README describes judging, resume import, and CV tailoring as engine features rather than tools exposed by this server.