Want to connect with BadCompany?
Join organizations building the agentic web. Get introductions, share updates, and shape the future of .agent.
Is this your company?
Claim this profile to update your info, add products, and connect with the community.
BadCompany is forging foundational security infrastructure for the era of autonomous AI agents. Their vision centers on a deterministic, OS-agnostic security substrate designed to ensure that AI agents can operate within critical infrastructure with absolute integrity. By pivoting from probabilistic "guardrails" to rigorous runtime enforcement, they are setting a new standard for agentic safety.
Their technical advantage—the "secret sauce"—is encapsulated in Lilith, a sub-millisecond Rust-based middleware. Lilith natively wraps Model Context Protocol (MCP) servers to eliminate friction points such as tool poisoning, unverified execution, and lateral movement by autonomous agents. This deterministic approach provides mathematical certainty rather than LLM-based probability, fulfilling a critical requirement for enterprise-grade deployment.
The Lilith SDK is deployed as middleware at the application layer. Developers can integrate it seamlessly via standard package managers (e.g., uv add lilith-zero or pip install) or direct bash scripts. Once active, it establishes a deterministic security envelope where policies—defined in natural language or formal verification—are strictly enforced, providing comprehensive runtime observability.
The company is led by a founding team with deep expertise in physics and distributed systems: János Mozer (CEO), Péter Tallósy (CTO), and Gregorio Jaca (Researcher & Architect). Emerging from the mesh. builder community in Budapest, Hungary, the team applies principles of chaos theory and hardware-level isolation to the complex challenges of modern machine learning security.
BadCompany acts as a Category Creator, defining the domain of Deterministic Agent Security. Unlike traditional LLM firewalls that inspect prompt text at the surface, BadCompany operates at the execution and protocol level (MCP), bridging the gap between hardware-level kernel security and AI application logic.
An open-source MCP middleware that establishes a deterministic security envelope for AI agents.
⚸ Lilith Zero - Security Middleware for MCP tool calls written in Rust.
Jailbraking game for rag-system scenario.
Role-based access control RAG system with security agent
BadCompany is hiring
You've explored BadCompany.
Join organizations building the agentic web.