Run the server from source

Try an unreleased change, or point a client at a local build. You need Node.js 22 or later, pnpm and a clone of the repository, on a platform the PDF engine supports.

Build and run

From the repository root:

pnpm install
pnpm turbo run build --filter=@sezzlee/pdf-mcp
node packages/servers/pdf-mcp/dist/cli.js /absolute/path/to/documents

Build through Turbo, not pnpm --filter @sezzlee/pdf-mcp build: the server uses @sezzlee/file-core through its built dist/, and only Turbo builds it first.

In a client configuration, use "command": "node" and "args": ["/absolute/path/to/sezzlee/packages/servers/pdf-mcp/dist/cli.js", "/absolute/path/to/documents"]. Add "--ocr" and a binding path after the folder to enable OCR.

Call it by hand

npx @modelcontextprotocol/inspector node packages/servers/pdf-mcp/dist/cli.js /absolute/path/to/documents

This opens the Inspector's web interface with the server connected, where you can call every tool.

Test it

pnpm turbo run test --filter=@sezzlee/pdf-mcp
pnpm turbo run check-types lint --filter=@sezzlee/pdf-mcp

The OCR tests run against fake ports. A suite against a real Ollama is skipped unless SEZZLEE_PDF_OCR_URL is set; it reads SEZZLEE_PDF_OCR_MODEL too, and defaults to deepseek-ocr:3b.