OnLeash — AI Agent Governance & Control Platform

OnLeash monitors, controls, and audits autonomous AI agent actions in real-time. Policy enforcement engine evaluates every action in under 8 milliseconds. Emergency kill switch, PII detection, semantic firewalls, and tamper-proof cryptographic audit trails for teams deploying AI agents into production.

Governance Layers

• Discover — Agent inventory, runtime fingerprinting, shadow AI detection
• Understand — Risk scoring, behavioral DNA profiling, intent classification
• Control — Kill switch, circuit breakers, rate limiting, quarantine
• Enforce — Policy engine, approval gates, action evaluation
• Compliance — EU AI Act, NIST AI RMF, SOC 2, ISO 27001

Integrations

Works with Cursor, Claude Code, Windsurf, Cline, GitHub Copilot, LangChain, AutoGPT, CrewAI, and any MCP-compatible AI tool. TypeScript SDK, Python SDK, CLI, REST API, and MCP Server integration.

Novel Capabilities

• Agent Taming — behavioral state machine from WILD to TRUSTED
• Z3 Policy Prover — formal verification of governance policies
• Semantic Firewalls — prompt injection defense
• ABOM — Agent Bill of Materials for supply chain integrity
• WebSocket Shield — real-time message scanning