ROADMAP
Where Mock Machines is headed.
Mock Machines today is a fast, faithful data-generation engine. The intent is to grow it into a testing tool for agentic coding, analytics, and reinforcement learning — a place where you can author a world by talking to an agent, train against it from Python, and generate datasets at a scale that breaks lesser tools. The work below is how we get there.
PLANNED CAPABILITIES
What we're building next.
| Initiative | What it unlocks | Direction |
|---|---|---|
| Scenario creation & editing via UI and chat/MCP | Author and refine a whole scenario from the studio or from a chat agent over MCP — define machines, fields and transitions, validate, and deploy without hand-writing YAML. | Agentic coding |
| Modelling documentation review | Treat the modelling docs as a first-class artefact: review, tighten, and keep them in step with the engine so authors and agents share one accurate reference. | Agentic coding |
| CGo or gRPC integration with Python for reinforcement learning | Drive a running scenario from Python in-process (CGo) or over gRPC, so a reinforcement-learning loop can step the world, read observations, and apply actions at engine speed. | Reinforcement learning |
| More calculations and LLM support for field updates and event sampling | Richer expressions for field updates and probabilistic event selection, plus optional LLM-backed sampling — so an entity can decide its next state from context, not just a fixed matrix. | Analytics · RL |
| Billion-row dataset generation | Scale streaming generation past a billion rows — flushing retired entities to Parquet and reclaiming memory — so larger-than-memory worlds export as a single queryable dataset. | Analytics |
GET ACCESS