Paperlib

Pap erlib MCP Server provides an end-to-end platform for academic literature management and retrieval. It enables PDF import with text extraction, powerful hybrid search, a knowledge graph built from entities and relations, and automated generation of literature reviews from evidence packs. This makes it easier to organize papers, discover connections, and produce structured reviews.

How to use

You interact with the Paperlib MCP Server through an MCP client. Start the server using one of the available deployment methods, then issue operations such as importing PDFs, performing hybrid searches, extracting and visualizing knowledge graphs, and generating literature reviews. Typical workflows include importing a set of PDFs, running a hybrid search for your topic of interest, extracting a knowledge graph from a document, clustering topics, and drafting a review from evidence packs.

Example usage patterns include: importing a PDF to index its text and embeddings, performing a comprehensive search combining full-text and semantic results, building topic communities from a document set, and generating a draft literature review from a focused evidence pack.

How to install

Prerequisites you need before starting are PostgreSQL 16+ with pgvector, an S3-compatible object storage (MinIO or similar), and an OpenRouter API key.

Choose one of the deployment methods below and follow the exact commands shown.

{
  "mcpServers": {
    "paperlib-docker": {
      "command": "docker",
      "args": [
        "exec",
        "-i",
        "paperlib-mcp",
        "python",
        "-m",
        "paperlib_mcp.server"
      ]
    }
  }
}

Configure in your MCP client to point at the local docker-based server. The exact invocation shown here starts the MCP server inside the running container and exposes its API to the client.

{
  "mcpServers": {
    "paperlib": {
      "command": "uvx",
      "args": ["paperlib-mcp"],
      "env": {
        "POSTGRES_HOST": "localhost",
        "POSTGRES_USER": "paper",
        "POSTGRES_PASSWORD": "paper",
        "POSTGRES_DB": "paperlib",
        "S3_ENDPOINT": "http://localhost:9000",
        "MINIO_ROOT_USER": "minio",
        "MINIO_ROOT_PASSWORD": "minio123",
        "OPENROUTER_API_KEY": "your-api-key"
      }
    }
  }
}

For local development or quick trials, you can install the MCP client as a Python package and run the server directly, using environment variables to configure connections to PostgreSQL and MinIO.

{
  "mcpServers": {
    "paperlib": {
      "command": "paperlib-mcp",
      "args": [],
      "env": {
        "POSTGRES_HOST": "localhost",
        "POSTGRES_USER": "paper",
        "POSTGRES_PASSWORD": "paper",
        "POSTGRES_DB": "paperlib",
        "S3_ENDPOINT": "http://localhost:9000",
        "MINIO_ROOT_USER": "minio",
        "MINIO_ROOT_PASSWORD": "minio123",
        "OPENROUTER_API_KEY": "your-api-key"
      }
    }
  }
}

If you prefer local development with a quick start, you can clone the project, sync the environment, and run the server directly with Python.

uv sync
cp .env.example .env
# Edit .env

uv run python -m paperlib_mcp.server

Additional configuration and notes

The server expects several environment variables to control access to the database and storage services. The primary required key is OPENROUTER_API_KEY. Other variables configure the PostgreSQL connection and MinIO storage.

{
  "env": {
    "OPENROUTER_API_KEY": "<your-key>",
    "POSTGRES_HOST": "localhost",
    "POSTGRES_USER": "paper",
    "POSTGRES_PASSWORD": "paper",
    "POSTGRES_DB": "paperlib",
    "S3_ENDPOINT": "http://localhost:9000",
    "MINIO_ROOT_USER": "minio",
    "MINIO_ROOT_PASSWORD": "minio123"
  }
}

Usage examples

# Import a PDF into the system
> import_pdf file_path="/papers/study.pdf" title="Study Title"

# Perform a hybrid search across documents
> search_hybrid query="monetary policy" k=10

# Build a knowledge graph from a document
> extract_graph_v1 doc_id="abc123"
> build_communities_v1 level="macro"

# Create a literature review draft from an evidence pack
> build_evidence_pack query="CBDC" k=40
> draft_lit_review_v1 pack_id=1

Available tools

health_check

System health check

import_pdf

Import PDF documents

download_pdf

Download PDF by title to local directory

search_hybrid

Hybrid search (full-text + semantic)

get_document

Get document metadata

list_documents

List all documents

extract_graph_v1

Extract knowledge graph from a document

build_communities_v1

Build topic communities within the knowledge graph

summarize_community_v1

Generate summaries for topic communities

build_evidence_pack

Build evidence packs for reviews

draft_lit_review_v1

Generate a literature review draft