Moose

The product is a simulation and experimentation platform for measuring how AI coding agents discover, install and use software products, tools and interfaces, and for comparing results across agents, models and interaction methods. It sells to developers, product teams and data teams at B2B software companies whose products are increasingly used by coding agents. It is delivered as a cloud service and local command-line interface with managed infrastructure, clean isolated sandboxes per run, and warehoused trace data.

Key features

• Unbiased simulations in isolated sandboxes
• Define experiments and test parameters
• Select AI agents and LLM models
• Compare across agents and models
• Run locally or on managed cloud
• Parallelize runs at scale
• Capture agent transcripts and timelines
• Record OTEL logs and errors
• Track pass/fail and performance metrics
• Measure latency, cost and wall-clock time
• Versioned and warehoused trace data