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.
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.