P-01
100× faster than Gymnasium
Generation runs at over a million rows per second — orders of magnitude beyond OpenAI's Gymnasium. Build a multi-million-row world while the kettle boils.
MODELLING ENGINE · V0.4 Spin up a believable
- 01 Over 1M rows / second
- 02 Demo your software
- 03 Train your AI
Spin up a believable
world in seconds.
Mock Machines generates perfectly linked datasets and live APIs — at over a million rows per second — so you can demo your software and train your agents against worlds that actually behave.
FIG.1 / DATA MODELOSI · DUCKDB
1,000,000+ rows / sec
PLATFORM
The engine, in four parts.
P-02
Perfectly linked tables
Every foreign key resolves, every join is clean. Referential integrity is guaranteed by construction — ideal for exercising a semantic layer or an agentic AI end-to-end.
P-03
Dynamic REST API generation
Each scenario ships a generated REST API so agents and apps interact with a live, mutating data model — not a static dump.
P-04
Instant OSI semantic layer
An OSI-compatible semantic layer definition is emitted with every scenario, so metrics and dimensions are defined the moment your data exists.
OUTCOMES
What you walk away with.
O-01
Generate credible scenarios instantly
Pick a model, set the scale, and get a believable world in seconds — coherent entities, plausible distributions, real relationships.
O-02
Download as DuckDB + semantic layer
Export the whole world as a single DuckDB file with its OSI-compatible semantic layer definition attached. Query it anywhere.
O-03
Interact live over MCP & REST
Connect an agent over MCP, or any client over REST, and operate against a living data model that responds, mutates and ticks forward.
CATALOG
All scenarios→ Start from a scenario.
▶ REPLAY SC-002 Transport Network
London Underground
A live tube graph — eleven lines, hundreds of trains, platforms held as a mutex.
OPEN →
SC-004 Insurance Claims
Claims FNOL
First notice of loss through settlement, with capacity-capped regional adjusters.
OPEN →
SC-007 Commerce Pipeline
Online Shopping
Order fulfilment against live warehouse stock — confirmations, shipments and returns.
OPEN →
▶ REPLAY SC-013 RL / POMDP
Tiger POMDP
The canonical POMDP — a hidden tiger, noisy growls, a finite-memory policy.
OPEN →
SC-014 Reinforcement Learning
Cliff Walking
Sutton & Barto's cliff-walking gridworld — three policies, one dangerous edge.
OPEN →
GET ACCESS
Pick a world. Or build your own.
OPTION B · WAITLIST Request editing access
Author and edit your own scenarios. Rolling out in batches.