Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.benchgen.com/llms.txt

Use this file to discover all available pages before exploring further.

The infrastructure for self-improving agents

BenchGen is a Synthetic Data Factory — it creates a digital twin of your business from your enterprise data, then uses that twin to simulate realistic environments, train RL agents, and generate unlimited synthetic data and trajectories. The result: agents that improve continuously on their own, grounded in your real business context. BenchGen — Digital Twin of Your Business

How it works

BenchGen takes your existing enterprise data — CRM, ERP, data warehouse, support tickets — and builds a living simulation of your business. Agents are trained inside that simulation using reinforcement learning, then challenged against real-world benchmarks before deployment.

Agents

Build and deploy AI agents with topics, actions, and integrations using the Agentspace visual builder.

Eval

Benchmark any model against curated datasets and track performance metrics over time.

Train

Fine-tune models with LoRA on your own datasets and export adapters ready for inference.

The self-improvement loop

Understand how Simulate → Train → Generate forms a self-reinforcing cycle.

Why BenchGen

More accurate

Agents trained in your digital twin make better decisions in the complex scenarios your business actually faces.

Lower risk

Test and break agents in simulation before they touch production. Deploy with fewer surprises.

Faster iteration

The simulation loop runs continuously. Test more, ship faster, iterate without waiting for real-world data.

Enterprise impact

Drive measurable results across the organization — from support to ops to sales — with agents that get better over time.

API reference

API reference

Integrate BenchGen into your own pipelines via the REST API.