The Qualixar Suite
Ten shipped research instruments for making AI agents reliable, safe, and production-ready. Three more in the forge. Every instrument is backed by an arXiv preprint.
Runtime Behavioral Contracts for AI Agents
Define YAML rules and check supplied structured agent state in Python. Place checks at an action boundary and raise on hard violations.
Supply Chain Security for AI Agent Skills
Static analysis, behavioral sandboxing, and cryptographic attestation for the AI agent skill ecosystem. Detects malicious skills before they execute.
Local-first Memory for AI Agents
Store and recall project context across AI agent sessions through MCP or CLI, with workspace isolation, explicit operating modes and auditable retrieval.
Peer-to-peer communication layer for AI coding agents
P2P agent communication with 8 MCP tools. Agents discover each other, send messages, share state, and lock files. Works with any MCP-compatible agent. SQLite + UDS for <100ms delivery. 480 tests, 100% coverage.
Control your Mac with your hand through the built-in camera
Point to move the cursor, pinch to click, thumb and ring finger to scroll, wave to switch apps. Fully local, written in Rust, with a 19-minute video course on building it.
The World's First MCP Gateway That Learns
One hub process, every MCP server, every AI client. 430+ tools exposed through 3 meta-tools — 79% fewer processes, 150K tokens saved per session. Federated tool discovery with shared cache, cost telemetry, and retrieval learning.
AI Reliability Engineering — The Universal Runtime for AI Agents
Orchestration runtime of the Qualixar AI Reliability Engineering platform. 13 execution topologies (sequential, parallel, debate, mesh, hybrid, and more), Forge AI auto-design, adversarial judge pipeline, cost-quality-latency routing. 2,936 tests passing. Paper on arXiv.
Regression Testing for Non-Deterministic AI Agents
Token-efficient stochastic behavioral testing framework purpose-built for non-deterministic AI agent workflows. Part of the Qualixar AI Reliability Engineering platform.
The Harness Decides DONE — Never the Agent's Own Word
A worker, an independent gate the worker cannot write to, and a declared budget. Termination is proved in closed form under a global repair budget, no node reaches DONE without a gate verdict in an append-only hash-chained ledger, and the spend ceiling is enforced inside an attempt rather than between attempts. Nine enforced bounds plus a kill switch. The instrument found 47 vacuous gates in shipped, reviewed code.
Jev Decision Layer
Typed Decisions for Coding Agents — Execution Stays With the Agent
Route bounded task, tool, skill, and review choices through TypeSafe Jev and get back a typed answer with probabilities and a local, inspectable receipt. A local policy broker checks workspace scope, screens recognizable secrets, and enforces daily limits before anything leaves the machine. 20 workflow contracts, 32 recipe specifications, five operating modes. An answer is advice, never permission to act.
Project Echo
Multi-agent communication degradation benchmarking.
Project Sentinel
Reliability analysis for AI-generated code.
Project Rewind
Time-travel debugging for autonomous agents.
Project Aurora
Chaos engineering principles applied to AI agent systems.
Project Bridge
Migration engineering across agent frameworks.
Project Nexus
Composition testing for agent pipelines.
Building Philosophy
Every tool I build follows the same principle: research first, then ship. Each product is backed by a formal paper with mathematical proofs and real experiments.
I build for production. Years of enterprise architecture taught me that tools need to work at scale, handle edge cases, and integrate with existing workflows. If it can't be pip installed and used in 5 minutes, it's not ready.