Chat Testing

Test your chat prompts in the most flexible prompt playground. Simultaneously stream and compare two chat interactions to fine-tune different models, system messages, or chat templates.

2 prompt chat templates side-by-side
"Batches is hugely helpful in seeing the impact of changes I make to my prompts. Not quite the right term, but batch testing gives me more confidence in the statistical significance of the changes I make when I can see the output generated many times. If you're not batch testing your prompts, you're probably missing out on some low-hanging fruit."
Headshot of Jade Samadi
Jade Samadi
Founder, Smart Recover

How it works

Three stacked rectangles
Create a chat prompt
2 blue cog wheels spinning together
Test variations side-by-side. Use different models, parameters, and data
Cursor with a check mark above it
Collaborate, approve, and merge

Dual Streams

Simultaneously stream and compare two conversations to evaluate variations across different models, system messages, or chat templates. Get better outputs through better testing.

Chevron pointing right

What you can build

Graph icon
Scalable content creation

Embed content creation prompts in Notion

Life flotation device icon
Client support form

Connect documentation to a form so that clients can get quick answers

Bell Icon
Lead magnet

Create mini-apps that drive value for website visitors

blocks icon
Connect custom data
(Gdocs, csv, Excel)
lock icon
Your prompts are secure
Code icon
Embed anywhere

Intuitive versioning and collaboration

Compare your chat prompts to previous versions, review new changes through merge requests before final approval, and ensure every update is refined and effective.

Chevron pointing right
A list of previous prompt versions to compare against

Retry user messages

Wouldn't it be helpful to restart the conversation from a specific message, rather than recreate a whole new conversation? Now you can!

Chevron pointing right

Few-shot prompting made easy

Quickly train any model by providing just a few examples within your chat template. Get better outputs that match your desired structure, tone, and style, without needing extensive datasets.

Chevron pointing right
4 messages stacked on each other with purple background

Join the waitlist

Organize your prompts, test them thoroughly, and get better outputs

Got questions? Schedule a demo with the founder