Security middleware for AI Agents. Intercepts shell commands before execution using a multi-layer pipeline: binary allowlist, regex patterns, deterministic intent coherence mapping, and LLM semantic check as last resort.

Sandboxed code execution daemon with enforced memory, CPU, and output limits. JSON API for humans and AI agents.

Safe execution layer for LLM tools: fatigue-aware gates, witness logs and policy checks before agents touch real systems.

Results utilities and wrappers made for JS

Control plane for multi-agent engineering work: intent, plan, execute, review, synthesize, with trajectories and routing policy.

Güvenli ve kontrollü yerel masaüstü ajan sistemi (simülasyon + kullanıcı onayı) v1.4

Safe-by-construction local execution substrate for mini-swe-agent-style bash-action coding agents.

Lightweight AI code execution sandbox — run untrusted code safely with resource limits, timeout protection, and output capture

A v1 command execution sandbox in Zig with timeout control, output capture, and process handling.

A secure MCP server for executing whitelisted development commands with comprehensive security controls.

Terminal-native AI agent orchestrator with safety gates, repo-signal integration and mq ecosystem support