Use cases
Native MCP

PDF tools for AI agents

Connect supported AI clients to one hosted Streamable HTTP endpoint. The client discovers the tools and opens OAuth sign-in. You do not need to install a package or run a local process.

Create an account Read the setup guide
HTTPS Remote endpoint
OAuth Browser sign-in
0 Server packages

Supported AI platforms

Each connection path is verified against the client's documentation.

Verified

Claude / Desktop

Add the hosted endpoint as a custom connector from Claude's Connectors settings.

  • Remote Streamable HTTP
  • Browser-based OAuth sign-in
  • Tools and resources
  • No desktop JSON for this remote server
Official client documentation
Verified

Claude Code

Register the remote URL with the Claude CLI's HTTP transport and authenticate from /mcp.

  • Official HTTP transport flag
  • Automatic OAuth discovery
  • User or project scope
  • No local server process
Official client documentation
Verified

VS Code

Add an HTTP server through the Command Palette or the servers object in mcp.json.

  • servers top-level key
  • type http + remote URL
  • Trust and authentication UI
  • Workspace or user profile config
Official client documentation
Verified

Cursor

Add the URL under mcpServers in a project or global Cursor MCP configuration.

  • mcpServers top-level key
  • Streamable HTTP support
  • OAuth for remote servers
  • Project or global configuration
Official client documentation
Plan dependent

ChatGPT

Create a custom app from the remote endpoint in ChatGPT developer mode where the workspace allows it.

  • Remote MCP endpoint
  • OAuth during tool scan
  • Workspace admin controls
  • Plan-dependent write-tool access
Official client documentation

What is MCP?

Model Context Protocol is an open protocol for connecting AI models securely to external tools and data sources. It gives AI agents a standard way to use APIs and services.

  • Secure connection OAuth discovery, PKCE, scoped access, and per-client tool controls
  • Standard interface One Streamable HTTP endpoint; each client uses its documented configuration
  • Live discovery The client retrieves current tool names and JSON schemas from the server
mcp-protocol.log

> TOOL_CALL html_to_pdf

> INPUT { html: "<h1>Report</h1>..." }

> EXECUTING Streamable HTTP request...

> SUCCESS PDF stored privately

> OUTPUT { document_id, view_url }

> COMPLETE Tool result returned to LLM

What agents can build

Documents generated from conversations, retrieved context, and automated analysis.

Documents from conversations

Users request a document in natural language. The agent interprets the request, gathers the data, and generates a PDF on demand.

Input User: "Create an invoice for project X"
Output Agent generates formatted PDF invoice with all details
Documents from RAG pipelines

Research assistants can turn retrieved context into summary documents and downloadable reports.

Input Agent retrieves relevant documents from vector DB
Output Compile findings into formatted PDF report
Multi-agent workflows

Use pdf-mcp as a shared tool across multiple AI agents. One agent researches, another analyzes, and a third generates the final document.

Input Research agent completes analysis
Output Documentation agent creates the final PDF
Customer support documents

During a support conversation, an AI agent can generate personalized documentation, troubleshooting guides, or case summaries as PDFs.

Input Support ticket requires documentation
Output Generate personalized PDF guide for customer
Automated reporting

Agents that monitor systems or analyze data can generate periodic PDF reports with findings and recommendations.

Input Scheduled analysis or threshold alert
Output Generate PDF report with findings and charts
Contracts and proposals

Sales agents can create customized proposals and contracts from conversation context and customer requirements.

Input Customer requirements captured in conversation
Output Generate tailored proposal PDF with pricing

How it works

Connect the hosted URL, sign in with OAuth, and enable the tools you need.

1

Create an account

Sign up for pdf-mcp. Interactive MCP clients authenticate your account through OAuth.

2

Add the remote URL

Choose HTTP, Remote MCP, or Custom connector in the client and enter https://api.pdf-mcp.io/mcp.

3

Sign in

The client discovers pdf-mcp's OAuth metadata, opens the browser flow, and retrieves the live tool schemas.

4

Enable the tools

Select the pdf-mcp tools in the client and ask it to create, inspect, transform, or manage a PDF.

Client configuration

These examples configure remote HTTP connections. They do not start a local server.

Claude Code

terminal 
claude mcp add --transport http --scope user pdf-mcp https://api.pdf-mcp.io/mcp

# Then open Claude Code and run:
/mcp

Visual Studio Code

.vscode/mcp.json 
{
  "servers": {
    "pdf-mcp": {
      "type": "http",
      "url": "https://api.pdf-mcp.io/mcp"
    }
  }
}

Cursor

.cursor/mcp.json 
{
  "mcpServers": {
    "pdf-mcp": {
      "type": "http",
      "url": "https://api.pdf-mcp.io/mcp"
    }
  }
}

Claude and Claude Desktop use Customize → Connectors → Add custom connector; ChatGPT uses a custom app in developer mode where the plan and workspace allow it. See the client-by-client setup guide and primary sources.

PDF output for RAG pipelines

Retrieval-Augmented Generation workflows often need to produce final documents from retrieved context. pdf-mcp renders those documents as PDFs.

1. Retrieve Query vector database for relevant documents
2. Augment LLM synthesizes findings with context
3. Generate pdf-mcp renders final document as PDF
rag-pipeline.log

[14:32:01] RETRIEVE Querying vector DB...

[14:32:02] FOUND 12 relevant documents

[14:32:03] AUGMENT Sending to LLM with context

[14:32:08] SYNTHESIZE Generating report content...

[14:32:09] GENERATE Calling pdf-mcp API

[14:32:10] COMPLETE research-report.pdf ready

Why use pdf-mcp with AI?

  • Agent workflows The MCP tools are designed for LLM clients
  • No local runtime The hosted service receives MCP requests directly over HTTPS
  • Live tool schemas Clients discover the current PDF, document, template, asset, and theme tools
  • Three storage modes Managed storage, BYOB, or fully stateless mode
agent-session.log

USER Create a summary report of our Q4 sales data

AGENT I'll analyze the Q4 data and generate a report.

> TOOL query_database(period="Q4")

> OK Retrieved 1,247 transactions

> TOOL html_to_pdf(html="...")

> OK PDF generated successfully

AGENT Here's your Q4 Sales Report: [Download PDF]

Connect your AI client

Use the remote Streamable HTTP endpoint. OAuth handles sign-in, with no package or local server to install.

Create a free account Explore use cases