Show HN: Promptbuild.ai – Version control for LLM prompts
After months of managing prompts in spreadsheets and losing track of which variations performed best, I decided to build a proper solution. PromptBuild.ai is essentially GitHub meets prompt engineering - version control, testing, and performance analytics all in one place.
The problem I was solving:
- Testing 10+ variations of a prompt and forgetting which performed best
- No systematic way to track prompt performance over time
- Collaborating with team members was chaos (email threads, Slack messages, conflicting versions)
- Different prompts for dev/staging/prod environments living in random places
Key features built specifically for prompt engineering:
- Visual version timeline - See every iteration of your prompts with who changed what and why
- Interactive testing playground - Test prompts with variable substitution and capture responses
- Performance scoring - Rate each test run (1-5 stars) and build a performance history
- Variable templates - Create reusable prompts with {{customer_name}}, {{context}}, etc.
- Global search - Find any prompt across all projects instantly
What's different from just using Git:
- Built specifically for prompts, not code
- Interactive testing interface built-in
- Performance metrics and analytics
- No command line needed
- Designed for non-technical team members too
Current status:
- Core platform is live and FREE (unlimited projects/prompts/versions)
- Working on production API endpoints (so your apps can fetch prompts dynamically)
- Team collaboration features coming next month
I've been using it for my own projects for the past month and it's completely changed how I approach prompt development. Instead of guessing, I now have data on which prompts perform best.
Would love to get feedback from this community - what features would make your prompt engineering workflow better?
Check it out: https://promptbuild.ai
P.S. - If you have a specific workflow or use case, I'd love to hear about it. Building this for the community, not just myself!
No comments yet