MUTX provides a governed runtime for AI agents that records every action and evaluates tool calls against policy before execution. The platform links intent, policy decisions, approvals, and outcomes into verifiable receipts, giving operators a complete audit trail. It offers a macOS app, API, and self-hostable control plane for production environments.
Funding
Funding not disclosed
Founders
Product
Problem
AI agents increasingly execute actions autonomously, but their behavior often lacks governance and auditability. Without a runtime that enforces policy before tool calls execute, organizations face unapproved actions, compliance gaps, and an inability to prove what happened after the fact.
Solution
MUTX provides a governed runtime for AI agents that records every submitted action and evaluates registered tool calls against policy before they execute. The platform captures intent, binds it to policy decisions, requires human approval at critical boundaries, and produces cryptographic-style receipts linking each action to its observed outcome. Operators monitor fleet health, active runs, interventions, and spend through a unified hierarchy, enabling them to act before incidents escalate. The system supports macOS, API, and self-hostable control plane deployments, making it adaptable to existing infrastructure.
Target Audience
Primary customers are engineering and operations teams deploying AI agents in production environments who need policy enforcement, audit trails, and human oversight for autonomous actions.
Features
- Governed runtime that evaluates tool calls against policy before execution, with DENY and DEFER verdicts keeping handlers uncalled
- Immutable run records linking intent, policy decisions, approvals, and outcomes into verifiable receipts
- Human-in-the-loop approval workflow for actions crossing policy boundaries, with scope preservation and expiry controls
- Unified operator dashboard showing fleet health, active runs, interventions, spend, and evidence in one hierarchy
- macOS app, API, and self-hostable control plane for flexible deployment
- Policy gates that match governed moves to defined boundaries, preserving action context for review