Introducing MalimGraph: Open-Source Knowledge Graph Engine for Claude

███╗   ███╗ █████╗ ██╗     ██╗███╗   ███╗ ██████╗ ██████╗  █████╗ ██████╗ ██╗  ██╗
████╗ ████║██╔══██╗██║     ██║████╗ ████║██╔════╝ ██╔══██╗██╔══██╗██╔══██╗██║  ██║
██╔████╔██║███████║██║     ██║██╔████╔██║██║  ███╗██████╔╝███████║██████╔╝███████║
██║╚██╔╝██║██╔══██║██║     ██║██║╚██╔╝██║██║   ██║██╔══██╗██╔══██║██╔═══╝ ██╔══██║
██║ ╚═╝ ██║██║  ██║███████╗██║██║ ╚═╝ ██║╚██████╔╝██║  ██║██║  ██║██║     ██║  ██║
╚═╝     ╚═╝╚═╝  ╚═╝╚══════╝╚═╝╚═╝     ╚═╝ ╚═════╝ ╚═╝  ╚═╝╚═╝  ╚═╝╚═╝     ╚═╝  ╚═╝

We are excited to announce the public release of MalimGraph v0.2.0, an experimental open-source knowledge graph engine built specifically for Claude Code and Claude Desktop.

MalimGraph transforms dense policy documents, research papers, and regulatory reports into interactive, explorable knowledge graphs. It extracts entities and relationships with verbatim source citations, exports to multiple graph formats, and generates visual discovery maps—all orchestrated by Claude's native intelligence.

The Problem We're Solving

Knowledge workers across research, policy, legal, and intelligence sectors face a common challenge: extracting actionable insights from hundreds of pages of dense documentation. Traditional text search and manual reading cannot capture the complex web of relationships between entities, jurisdictions, regulations, and concepts.

MalimGraph addresses this gap by providing automated knowledge extraction and visualization infrastructure that works seamlessly within the Claude ecosystem.

Core Capabilities

1. Intelligent Entity Extraction

MalimGraph leverages Claude's reasoning capabilities to identify entities (organizations, persons, locations, regulations, concepts) and map their relationships with complete source traceability. Every extracted relationship includes verbatim citations from the source document.

2. Multi-Format Export

Knowledge graphs are exported in three industry-standard formats:

• JSON for application integration
• Cypher for Neo4j graph databases
• Apache AGE SQL for PostgreSQL with graph extensions

3. Visual Discovery Suite

Two complementary visualization engines:

• Premium Visualizer (vis.js): Interactive physics-based explorer with detail inspection
• High-Performance Explorer (Sigma.js): WebGL-powered rendering for large-scale graphs (1000+ nodes)

4. RAG-Ready Infrastructure

Built-in document chunking with token-aware overlap and native pgvector integration for embedding storage, enabling downstream retrieval-augmented generation workflows.

5. MCP Server Architecture

Deployed as a Model Context Protocol (MCP) server, MalimGraph integrates natively with Claude Code and can be accessed as a serverless endpoint at mcpserver.malim.my/mcp.

Installation and Usage

Quick Start (Claude Code)

pip install malimgraph
claude mcp add malimgraph -- malimgraph-plugin

The /kg Command

Type /kg in Claude Code to initiate the full knowledge discovery workflow:

1. Extract: Parse PDF content and metadata
2. Analyze: Claude identifies entities and relationships with evidence
3. Build: Generate knowledge graph in JSON, Cypher, and SQL
4. Visualize: Launch interactive browser explorer

Skill-Based Invocation

MalimGraph provides natural language triggers:

• "knowledge graph" — Full extraction and export
• "visualise graph" — Launch visual explorer
• "chunk for RAG" — Prepare document for vector embeddings
• "load into Neo4j" — Import to local graph database

Real-World Application: Financial Crime Intelligence

The initial development of MalimGraph was driven by work in financial crime intelligence and anti-money laundering (AML) research. Documents like the Asia/Pacific Group on Money Laundering's Typologies Report contain critical intelligence across multiple dimensions: jurisdictions, regulatory frameworks, enforcement bodies, criminal typologies, and cross-border relationships.

Extracting this structure manually is time-intensive and error-prone. MalimGraph automated this process, generating a graph with 103 entities and 109 relationships from a single report, enabling rapid pattern recognition and cross-jurisdictional analysis.

Technical Architecture

Core Technologies:

• Python 3.10+
• PyPDF2 for document parsing
• MCP protocol for Claude integration
• vis.js and Sigma.js for visualization
• Support for Neo4j and PostgreSQL (Apache AGE)

Deployment Options:

• Local installation via pip
• Claude Code plugin
• Serverless MCP endpoint (Railway/Vercel)

Current Status and Roadmap

MalimGraph v0.2.0 is in experimental stage. Core extraction and visualization workflows are functional, but expect rough edges in edge cases, large-scale deployments, and certain document formats.

Planned Enhancements:

• Multi-document graph merging
• Temporal relationship tracking
• Enhanced entity disambiguation
• Custom extraction templates for domain-specific documents
• Performance optimization for documents >500 pages

Open Source and Community Contribution

MalimGraph is released under the MIT License. The codebase is publicly available on GitHub, and we actively welcome contributions from developers, researchers, and domain specialists.

How to Contribute:

• Star the repository: github.com/malim-ai-labs/malim-graph-plugin
• Report issues or request features via GitHub Issues
• Submit pull requests for bug fixes or enhancements
• Share use cases and experimental results

Why Open Source Matters

At Malim AI Labs, we believe knowledge infrastructure should be accessible, transparent, and community-driven. Proprietary black-box solutions create vendor lock-in and limit innovation. Open-source development enables:

• Transparency: Full visibility into extraction logic and data handling
• Customization: Adapt the engine to domain-specific requirements
• Trust: Auditable code for sensitive regulatory and intelligence work
• Collaboration: Collective improvement through distributed expertise

Conclusion

MalimGraph represents our commitment to building AI-powered tools that democratize access to structured knowledge. Whether you work in policy research, regulatory compliance, competitive intelligence, or academic research, we invite you to experiment with MalimGraph and contribute to its development.

This is early-stage infrastructure. It will evolve through real-world use and community feedback. We look forward to seeing what you build with it.

Install MalimGraph:

pip install malimgraph

GitHub Repository:
github.com/malim-ai-labs/malim-graph-plugin

MCP Endpoint:
https://mcpserver.malim.my/mcp

License:
MIT License © 2026 Malim AI Labs

Malim AI Labs is a Malaysian social enterprise building AI-powered research and knowledge access tools. We develop open-source infrastructure for extracting, structuring, and navigating complex information.