Proactive security and compliance enforcement for AI

Agnostic Security's Yashigani platform sits between an organization's agents, people and APIs and everything they reach for, including models, MCP tools, other agents and documents.

Every agent runs in its own isolated container with its own cryptographic identity, and every call is inspected in both directions by pattern rules, a machine learning classifier and an LLM judge, and is then checked against policy before it runs.

Version 6.0 runs on KUROSHIO, the company's own inference engine, which can operate fully offline, and every verdict is written to a hash chained, tamper evident audit trail.

The platform is built for organizations that need sensitive data answered by local models while other traffic is routed to cloud providers by policy.

The Community edition is open source under Apache 2.0 on GitHub, and every paid tier ships from that same codebase.