Measure agent performance locally
LLM evaluations that make it effortless to run and compare your agents
Get started for free. No signup required.
Trusted by engineers at
Analyze multi-step agent behavior across runs
FAST ITERATION CYCLES
Run evaluations in seconds and iterate as fast as you build. No waiting on pipelines or external services
LOCAL-FIRST EVALUATION
Keep runs, data, and prompts on your machine. Nothing leaves your machine, no added risk.
FULLY CUSTOMIZABLE
Use custom evaluators or tailor metrics to your workflow whether it’s tool usage, task success, or multi-step reasoning.
MEASURED PROGRESS
Track performance across every run with clear metrics and chart. Regressions are visible the moment they happen.
AgentHub Evaluation Workflow
RUN
Evaluate observed runs or datasets
EVALUATE
Drill into evaluator details and metrics
COMPARE
Validate changes vs previous evaluations
Before You Run Your First Evaluation
Do I need Railtracks?
Evaluations is built on Railtracks. We’re exploring support for other frameworks and import formats.
Is my data private?
100%. Local runs never leave your machine.
What about human evaluations?
Human evaluators are supported in Conductr cloud (coming soon!). Local stays automated.
