Welcome to Deconvolute
The open source security SDK for AI agents and RAG pipelines.
Deconvolute is a security SDK built to protect your AI agents and Large Language Model pipelines from infrastructure- and content-level attacks.
pip install deconvoluteWhether you are building autonomous MCP agents or implementing Retrieval-Augmented Generation (RAG), Deconvolute provides the deterministic enforcement necessary to secure your infrastructure and your data.
Where to start?
Depending on what you are looking to achieve, here are the best places to begin:
- What is Deconvolute?: Read a high-level overview of the SDK and its separation of responsibilities.
- Quickstart Guide: Install the SDK, generate a policy, and secure your first MCP session in minutes.
- The MCP Firewall: Learn how to cryptographically seal your tool definitions to prevent shadowing and rug pull attacks.
- Content Scanners: Discover how to validate untrusted RAG documents and monitor LLM outputs for jailbreaks.
Community & Open Source
Deconvolute is an open source project. We actively welcome contributions, feedback, and security research.
- GitHub Repository: deconvolute-labs/deconvolute
- Benchmarks: deconvolute-labs/benchmarks
- Custom Signatures: Yara-Gen
Automated API Reference
If you are looking for specific class definitions, methods, or type signatures, you can jump straight into the generated API documentation: