Ataraxis
Bridging AI Coding Assistants and Scientific Hardware
Ataraxis is an open-source framework that enables AI coding assistants to interact with laboratory
hardware. It provides optimized hardware interface libraries, Model Context Protocol (MCP) servers
for structured device discovery, and domain-specific skills that encode expert workflows. AI agents
use these components to generate efficient data acquisition pipelines, configure systems, and
troubleshoot hardware issues.
Core Insight: AI assistance operates at configuration time while runtime data acquisition
remains deterministic and AI-independent. This separation ensures that network latency, API rate
limits, or model errors never disrupt a running experiment.
Authored by Ivan Kondratyev.
Copyright: 2026, NeuroAI Lab, Cornell University.
Features
Hardware Discovery & Validation
- MCP-based device enumeration: AI agents can query connected cameras, microcontrollers, and
motor controllers through structured tool interfaces.
- Pre-session diagnostics: Validate hardware connectivity and configuration through natural
language queries.
- Real-time status checking: Query device responsiveness, serial port status, and camera
capabilities without manual debugging loops.
Optimized Hardware Interfaces
- High-speed camera acquisition: Support for OpenCV and GeniCam cameras with real-time FFMPEG
encoding (CPU/GPU).
- Microcontroller communication: Bidirectional serial communication with Arduino and Teensy
boards at microsecond speeds.
- Precision timing: Microsecond-accurate timers using C++ chrono library bindings.
- Inter-process data sharing: Thread-safe shared memory arrays and scalable data logging.
AI-Assisted Development
- Code generation: AI agents generate hardware interface code following established patterns.
- Configuration management: Interactive experiment configuration using task templates.
- Domain-specific skills: Reusable workflows for camera interfaces, microcontroller modules,
and system health checks.
- Cross-repository coordination: Skills encode knowledge spanning multiple interdependent
libraries.
Deterministic Runtime
- Static acquisition pipelines: Experiments run independently of AI systems.
- Validated configurations: Pre-runtime parameter validation ensures reliable data collection.
- Reproducible execution: Configuration files capture complete experimental setups.
Architecture
┌─────────────────────────────────────────────────────────────────────────────┐
│ Ataraxis Architecture │
├─────────────────────────────────┬───────────────────────────────────────────┤
│ Configuration Time │ Runtime (No AI) │
├─────────────────────────────────┼───────────────────────────────────────────┤
│ │ │
│ ┌─────────────────────┐ │ ┌─────────────────────────┐ │
│ │ AI Agent (Claude) │ │ │ Static Acquisition │ │
│ └──────────┬──────────┘ │ │ Pipelines │ │
│ │ │ └────────────┬────────────┘ │
│ ▼ │ │ │
│ ┌─────────────────────┐ │ ▼ │
│ │ Skills & MCP │ │ ┌─────────────────────────┐ │
│ │ Discovery Tools │ ─────┼────▶│ Ataraxis Libraries │ │
│ └──────────┬──────────┘ │ └────────────┬────────────┘ │
│ │ │ │ │
│ ▼ │ ▼ │
│ ┌─────────────────────┐ │ ┌─────────────────────────┐ │
│ │ Config Files & │ │ │ Physical Hardware │ │
│ │ Pipeline Code │ │ └────────────┬────────────┘ │
│ └─────────────────────┘ │ │ │
│ │ ▼ │
│ │ ┌─────────────────────────┐ │
│ │ │ Session Data & Logs │ │
│ │ └─────────────────────────┘ │
└─────────────────────────────────┴───────────────────────────────────────────┘
Libraries
Core Infrastructure