Vibeleaderboard
← Back to Tools
Index / tool

PageIndex MCP

https://github.com/vectifyai/pageindex-mcp
Visit pageindex.ai
Category
AI Agents
Rank
No. 646Tools index
Pricing
Open Source
Type
TOOL
Builder
vectifyai
GitHub
385 stars
Latest release
v1.8.2
Date

About

MCP server for PageIndex, a vectorless reasoning-based RAG system that retrieves like a human expert via tree search.

What it does

PageIndex MCP lets compatible assistants work with long PDFs through PageIndex. Developers can connect to a hosted HTTP endpoint, while the local package bridges standard input and output clients to the remote service and adds local PDF upload support.

Why it's ranked here

The tool offers unusually broad MCP integration choices: hosted access, OAuth for app users, API-key authentication for developers, and a local package. Its value is strongest for PDF workflows, but the local server still depends on PageIndex’s remote backend.

What's good

Setup fits several environments without much custom plumbing. The server exposes both tools and resources, discovers remote capabilities dynamically, validates tool inputs, retries failed remote calls, supports two HTTP transport styles, and prevents remote tools from replacing its local document-processing tool.

Tradeoffs

The package is a remote-service bridge, not a self-contained document engine. OAuth requires opening a browser, running a temporary callback server, and storing credentials locally. Its JSON Schema conversion handles common types only, falling back to unrestricted values for unsupported shapes. The README and package metadata also disagree on the minimum Node.js version.

How to use it well

Use it when an MCP-capable agent must repeatedly question long local or online PDFs without adding vector-database infrastructure. Choose hosted API-key access for agent frameworks, OAuth for PageIndex App accounts, or the local package for uploads. Do not choose it for offline or fully self-hosted retrieval.

Technical notes+

src/server.ts implements a stdio MCP server with tool, resource, and resource-template handlers, connecting lazily through PageIndexMcpClient. src/client/mcp-client.ts prefers StreamableHTTPClientTransport, falls back to SSE during transport construction, and retries tool and resource operations twice with p-retry. src/tools/remote-proxy.ts fetches remote tool definitions, excludes get_signed_upload_url and submit_document, and converts a limited JSON Schema subset to Zod; unknown forms become z.any(). src/client/oauth-provider.ts implements browser OAuth, a localhost callback, and a home-directory JSON credential store written with mode 0o600. package.json defines an ES module, an npm executable, MCPB packaging, and a Node.js engine requirement of at least 20.8.1.

Observed

License
MIT
Primary language
TypeScript source packaged as an ES module
Install surface
Published npm package usable through npx, plus an MCPB bundle for one-click Claude Desktop installation
Interfaces
Hosted HTTP MCP endpoints and a local stdio MCP server
Authentication
Bearer API keys for developers and OAuth for PageIndex App users
Runtime
Package metadata requires Node.js 20.8.1 or newer
MCP capabilities
Exposes tools, resources, and resource templates

Read from README.md, package.json, src/index.ts, src/server.ts, src/config.ts, src/result.ts, src/globals.d.ts, src/tools/index.ts, src/client/auth.ts, src/tools/types.ts, src/types/global.d.ts, src/resources/index.ts, src/client/mcp-client.ts, src/tools/remote-proxy.ts, src/client/oauth-provider.ts.

What it can do

  • Search and retrieve information using vectorless reasoning

    Search query or question → Relevant information from indexed documents

  • Perform tree-based search through document content

    Search parameters and document corpus → Structured search results with reasoning paths

  • Generate expert-level responses using RAG system

    User questions and document context → Contextually relevant answers with source attribution

  • Index and organize documents for retrieval

    Document files or text content → Searchable document index

  • Analyze document relationships through reasoning

    Multiple documents or text sources → Conceptual connections and relationships between content

Tags

mcpragpageindexretrievalsearch

Tech Stack

Node.jsTypeScript

Media

PageIndex MCP

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.