Agent Evaluator
Visualize evaluator-optimizer iterative improvement loops where an evaluator scores agent output and an optimizer refines it until quality thresholds are met.
Features
- Iteration Timeline: Step through each evaluate-optimize cycle with diffs between versions
- Quality Scoring: Radar or bar chart showing scores across multiple evaluation dimensions
- Evaluation Criteria: Configurable rubric cards displaying pass/fail status for each criterion
- Convergence Tracking: Line chart plotting quality scores over iterations to visualize improvement
- Threshold Controls: Adjustable quality thresholds that determine when the loop should stop
Examples
Content Quality Loop
Evaluation results
Evaluate the renewal briefing for evidence and clarity.
Latest score
91 / 100
Four accounts need executive outreach this week.
Specific, evidence-led, and actionable.
Evidence94/100
Clarity88/100
Iteration history
Generator output
Four accounts need executive outreach this week.
Evaluation criteria
Evidence94/100
Clarity88/100
Evaluator feedback
Specific, evidence-led, and actionable.
3 iterations · Threshold 85 · 3 of 5
Installation
npx shadcn@latest add "https://agents-ui.github.io/agents-kit/c/agent-evaluator.json"