PdfParse

io.github.PabloG6v0.1.0Updated Oct 7, 2026

Extract PDFs into structured tables, query records, and configure document inbox workflows.

VerifiedStreamable HTTPWeb executableFiles & Storage Data & Analytics

Overview

AI-generated overview

PdfParse is a hosted MCP service that extracts PDFs into structured tables, lets an assistant query the resulting records, and configures document inbox…

What it does: PdfParse turns PDF documents into structured tables and exposes them through a hosted MCP endpoint. According to the description, an assistant can extract PDFs, query the extracted records, and configure document inbox workflows. The README also mentions companion Claude Code and OpenAI plugin packages and a TypeScript SDK, but the MCP tool list itself is not published in the supplied material.
When to use it: Consider it when you regularly need an assistant to pull tabular data out of PDFs and query that data, or to set up an inbox-style pipeline for incoming documents. It is a hosted service, so it suits users who prefer a managed endpoint over running extraction locally.
Requirements: A remote Streamable HTTP endpoint at reached over the network. A PdfParse account is required, and the connection uses OAuth; project API keys are not accepted by this OAuth endpoint. Processing consumes the PdfParse plan quota. No environment variables or headers are declared in the manifest.

Before you install

The service is account-based and processing consumes your plan quota, so usage may have cost implications. Authorization is OAuth-based; the README states that project API keys are not accepted by the MCP endpoint. Documents you submit are sent to a third-party hosted service, so consider whether the PDFs contain sensitive content. The README notes that ChatGPT directory review is pending and that a registry listing does not imply directory approval.

Installation

In SourceWeft

Open PdfParse in the dashboard and add it to a workspace.
Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "pdfparse": {
      "type": "http",
      "url": "https://mcp.pdfparse.net/mcp"
    }
  }
}

README

PdfParse integrations

PdfParse turns PDFs into structured tables. This repository distributes the official MCP connection, Claude Code and OpenAI plugin packages, and TypeScript SDK. The hosted service requires a PdfParse account; processing consumes your PdfParse plan quota.

MCP

Connect to https://mcp.pdfparse.net/mcp using Streamable HTTP and OAuth. Project API keys are not accepted by this OAuth endpoint. Setup and tool reference.

Claude Code

Sign in to PdfParse when Claude requests MCP authorization. The plugin includes get-started, document extraction, and workflow setup skills. Alternatively: claude mcp add --transport http pdfparse https://mcp.pdfparse.net/mcp.

For Claude web or Desktop, add that URL in Settings → Connectors → Add custom connector and authorize your PdfParse account. A Claude Code marketplace installation does not install a Claude web connector.

ChatGPT and Codex

The release ZIP contains the OpenAI plugin package. ChatGPT directory review is pending; the ZIP is a submission package, not an approved listing. In ChatGPT developer mode, add https://mcp.pdfparse.net/mcp as an OAuth MCP connection. See submission requirements.

TypeScript SDK

Download pdfparse-sdk-0.1.1.tgz from releases and install it with npm install ./pdfparse-sdk-0.1.1.tgz. The public npm listing is pending publisher authentication.

import { Client } from "pdfparse-sdk";const client = new Client(process.env.PDFPARSE_API_KEY!);const tables = await client.tables.list();

import { Client } from "pdfparse-sdk";const client = new Client(process.env.PDFPARSE_API_KEY!);const tables = await client.tables.list();

The SDK ships JavaScript for ESM and CommonJS with TypeScript declarations. Its project API key is separate from MCP OAuth. SDK reference.

The plugin and SDK are licensed under Apache 2.0; each distribution includes its LICENSE file.

Directory metadata

server.json describes the hosted endpoint for the official MCP Registry. A registry listing does not imply ChatGPT or Anthropic directory approval. Website · Support · Privacy · Terms.

Source: README.md at commit a0552c2

Tools

Tool metadata has not been indexed yet.

Version history

v0.1.0LatestOct 7, 2026