Trainer
Train and evaluate agents that perform work tasks through screen recording
About Trainer
Trainer records the screen, mouse, keyboard, and microphone narration to turn tasks into structured trajectories such as natural language, JSON, and action DSL, and fine-tunes agent policies from them. Once the SDK is connected, production runs are evaluated on step accuracy, coverage, and sequence completeness and then feed the next training loop — a good fit for handing repetitive industry work to agents.
Best for
Best use cases
Record repetitive workflows
Train task-executing agents
Evaluate production run performance
Deploy industry work automation
Features
Screen task recording
Captures screen, mouse, keyboard, and microphone narration in sync to form time-aligned task records.
Structured trajectories and training
Organizes recorded steps into natural language, JSON, action DSL, and Natural DSL, and uses them for policy fine-tuning.
Production run evaluation
Evaluates production runs on step accuracy, coverage, and sequence completeness, displayed side by side with human baselines.
SDK integration and deployment
Connect to the Trainer SDK via an API key so production runs flow back as training runs into the next round of training.
Pros
- No prompt engineering or manual data labeling required
- Covers screen, mouse, keyboard, and microphone narration
- Evaluated on step accuracy, coverage, and sequence completeness
Pricing Plans
Free
- Full product access during the launch period
- First 10 hours of recording free
- No credit card required
Pro
- Keep training agents
- Deploy agents to production
* Pricing information is for reference only. Please refer to the latest announcement on Trainer's official website.