Rust trust core
Deterministic TAPL policy, authorisation and bounded execution controls.
Security-first execution infrastructure for AI agents: Rust trust core, Go MCP gateway, deterministic policy and signed audit evidence.
The repository documents core engineering and verification work. Customer-environment sign-off remains open and is required per deployment. This page does not claim independent certification or production acceptance.
Model-generated requests should not automatically become privileged commands. Agent and tool integrations need an explicit point where identity, policy and approval determine what may execute.
The documented runtime places a Go MCP gateway and a Rust-authoritative policy core between agent intent and verified tool dispatch, with result checks and signed audit evidence.
Deterministic TAPL policy, authorisation and bounded execution controls.
Request validation, action firewall and verified MCP tool dispatch.
Brokered secret operations and explicit boundaries around privileged workloads.
Signed, tamper-evident dispatch, outcome and failure records as documented in the source.
The repository README documents an Ubuntu-targeted buyer engineering demo and exact-commit evidence. Its demo is not an independent penetration test or a production deployment.
Authoritative project: MosinShaikh505/TTRL-Agent-Runtime. Source reviewed: its README on 7 November 2026. Readiness must be re-evaluated for the selected commit and target environment.