PDF quotes from an AI agent
Asking a model to write a PDF gives you a different layout every time. With the MCP server, the agent only supplies data. Your published template decides how the quote looks.
- 1
Pick the Quote starter
Set your company details and terms, then publish. The fields (client, summary, items, total) become the tool's input schema automatically.
- 2
Connect your agent
In Cursor, paste this into ~/.cursor/mcp.json. In Claude Code, run: claude mcp add --transport http docspout https://docspout.com/mcp --header "Authorization: Bearer sk_test_...". Use a live or test API key; test keys are free and watermark the PDF.
{ "mcpServers": { "docspout": { "url": "https://docspout.com/mcp", "headers": { "Authorization": "Bearer sk_test_..." } } } } - 3
Ask for a quote
"Draft a quote for Lumen Labs: n8n workflow build $1,400 and a training session $300, 50% upfront." The agent calls list_templates, get_template to read the fields, then render_pdf with your Quote template.
- 4
Get a link back
The tool returns a PDF link and the page count. Missing fields come back as errors naming the exact path, so the agent can ask you instead of guessing.