GitHub
Your AI agent shows you the plan before it does the work, and a second agent checks what it did. - hfadhlullah/kaizen
What it does
You know the feeling. You ask your AI for a feature, or an email, or a migration plan. It goes quiet for a bit, then proudly hands you something. A week later you find the part where it confidently made something up.
kaizen adds two checkpoints so that stops happening:
Before: it shows you the plan. Nothing gets built until you say yes.
After: a *second* agent reviews the work. It never saw the first agent's reasoning, so it has no ego to protect. It just tells you what's wrong, numbered and graded by severity.
You pick what gets fixed. That's the whole thing.
It also works for things that aren't code: emails, docs, runbooks, research. In one demo the reviewer caught "Your price will not change" in a pricing email. That line wasn't true for annual plans, and it's the kind of mistake that turns into a support ticket.
**New in 1.
Product details
Profile
- Maker
- @futonefu
- Category
- AI Tools
- Platforms
- Web, Desktop
- Website
- github.com
Submitted by
- Launch date
- Launch tags
- Pricing
- Not specified
FAQ
Product questions
You know the feeling. You ask your AI for a feature, or an email, or a migration plan. It goes quiet for a bit, then proudly hands you something. A week later you find the part where it confidently made something up. kaizen adds two checkpoints so that stops happening: Before: it shows you the plan. Nothing gets built until you say yes. After: a *second* agent reviews the work. It never saw the first agent's reasoning, so it has no ego to protect. It just tells you what's wrong, numbered and graded by severity. You pick what gets fixed. That's the whole thing. It also works for things that aren't code: emails, docs, runbooks, research. In one demo the reviewer caught "Your price will not change" in a pricing email. That line wasn't true for annual plans, and it's the kind of mistake that turns into a support ticket. **New in 1.
builder, team, product mindset people
anyone who use ai for building product but get cheap result or not as expected
Community
Discuss and review GitHub
Read maker updates, ask questions, or leave product feedback for other builders evaluating this launch.
Discovery paths
