# Is the new prompt actually better?

JustEvals puts your eval runs side by side, so you can see whether a model or prompt change made the output better or just got lucky. Your pipeline uploads the results, you pick the better output per case, and your coding agent reads your notes before the next change.

Request access: contact@justevals.com · Documentation: https://justevals.com/docs

## How does it work?

1. **Upload.** Your pipeline sends outputs, cost and latency with one command. The models keep running where they run today.
2. **Review.** Go case by case with the keyboard. The input sits next to every output, zoom follows across all of them, and you can flip between outputs to catch small differences.
3. **Improve.** Your agent reads which output won and why through the API, fixes the prompt and runs the suite again.

## Every eval decision in one place

- **Hide the model names.** Names, costs and scores stay hidden until you pick. One key shows them.
- **Light or detailed scoring.** Just pick the better one, or add good or bad, 1–5 scores or your own criteria.
- **Cost and failures in view.** Wins, losses, cost and median time per run, updated as you vote.
- **Versioned test cases.** Change a case and get a new version. Old runs stay comparable.

Hosted in the EU. Private by default. Your data never trains a model.

## Made for coding agents

Everything works through the API. The agent skill tells Claude Code, Codex and other agents how to upload a run and read your notes before they touch a prompt.

```bash
python justevals.py push manifest.json
python justevals.py feedback --suite exterior
```

Agent skill: https://justevals.com/skill.md · API reference: https://justevals.com/docs/api.md · OpenAPI: https://justevals.com/openapi.json

## Questions

### Does JustEvals run my models?

No. Your pipeline calls the models with your own code and keys. JustEvals stores the results and the human judgments. There is no model gateway and nothing to proxy.

### Which models and tools does it work with?

Any. JustEvals only needs the outputs: images, videos, text or JSON, plus optional cost, latency and scores. If your pipeline can send an HTTP request or run a Python file, it can upload.

### Is it only for image models?

Image work is where side-by-side review matters most, so the review screen is built for it: the input next to each output, zoom that follows across all of them, and flipping between outputs. Video outputs play in place, and text outputs show as text.

### Why not build our own viewer?

Most teams start that way: a quick HTML page or notebook per eval. It works until you need last month's runs, a second reviewer, or a note your agent can read. JustEvals keeps every run in one place and takes one command from the pipeline you already have.

### How much does it cost?

JustEvals is in a private beta, and there is no public price yet. Write to contact@justevals.com to get access.

Updated 2026-10-14. Imprint: https://justevals.com/imprint