Home  ›  Playgrounds  ›  MCP tool explorer

pdfnative-mcp tool explorer

The tools an AI assistant gets when you wire pdfnative-mcp into Claude Desktop, Cursor, Continue, Zed, or any other MCP client (Cline, Windsurf, Goose, Gemini CLI…). The server exposes 28 tools — including the v1.6.0 PAdES ladder add_ltv / timestamp_pdf (with RFC 3161 timestamps on sign_pdf), update_metadata, the read-only inspect_layout pagination preview, all 13 block kinds in generate_basic_pdf, print production and charts v2, the v1.5.0 add_chart / form / encryption tools, the v1.4.0 markup annotate_pdf and network-free draft_governance_issue, and the v1.3.0 page-tree trio merge_pdfs / split_pdf / extract_pages — a pdfA flag on every document tool, multi-script lang on add_international_text, token-frugal read modes (verbosity / fields), and a per-tool _meta.apiVersion stability contract. Each card below shows a real example prompt, the pdfnative library payload this browser demo executes (the MCP tool schemas differ — see the MCP guide for the exact schemas), and a Generate PDF button that produces the very PDF the AI assistant would receive — running the underlying pdfnative library directly in your browser. No API keys, no install, no LLM cost.

For the full integration story (config, security model, signed-document workflow), see the MCP guide.

1. Configure your client

Pick your client. The snippet below is what you paste into its MCP servers config. npx downloads pdfnative-mcp on first use; no global install needed.

Other clients: any tool that speaks the Model Context Protocol over stdio works — Cline, Windsurf, Goose, Aider, LibreChat, Sourcegraph Cody, Gemini CLI, and your own custom client. Use the same command: "npx" / args: ["-y", "pdfnative-mcp"] pair in whatever JSON shape your client expects.

2. Try the tools

Loading pdfnative from esm.sh… this happens once per page load.

The interactive cards below cover the core document-authoring tools, the v1.3.0 page-tree trio (merge_pdfs, split_pdf, extract_pages), the v1.4.0 annotate_pdf, and the v1.6.0 update_metadata — all of which build their own source PDFs in the browser and run the real pdfnative 1.7.0 API. Two cards produce no PDF by design: inspect_layout returns a pagination report (run it to see the JSON), and draft_governance_issue validates a report locally, network-free — use its Copy args button. Eleven of the 28 tools have no card here: nine need a real PDF as input and run best from your own MCP client (validate_pdf, verify_pdf, add_attachment, extract_attachments, extract_text, read_form_fields, fill_form, encrypt_pdf, decrypt_pdf), and two need the operator-configured network providers a browser page cannot supply (add_ltv, timestamp_pdf — TSA / OCSP / CRL endpoints set via environment variables). See the MCP guide for their full schemas.

3. Now ask your AI assistant

Once you've pasted the config snippet above and restarted your client, ask it anything PDF-shaped:

The AI assistant will pick the right tools, fill the JSON arguments, and pdfnative will render the PDF — no code from you.

Resources