Is prompt version control available on the free tier?

Basic prompt templates are available on the free tier. A/B experiments and advanced versioning features are available on the a paid platform tier ($29.99/mo) and above.

How is this different from storing prompts in git?

Git tracks prompt text as part of your application code. Tack Room tracks prompts as operational artifacts with A/B testing, live rollback, and performance metrics attached to each version. You can change the active prompt without redeploying your application.

Can I use Tack Room with any model?

Yes. Prompt templates work with any of Stockyard's 40+ supported providers. You can test the same prompt across different models and compare results.

Stockyard Prompt Version Control — Track, Test, and Roll Back Prompt Changes

Prompts are code. Treat them that way.

Your LLM prompts change constantly. A tweak to the system prompt, a new few-shot example, a different output format instruction. Each change affects output quality, cost, and latency. But most teams track prompt changes in Slack messages or git commits buried in application code.

Prompt version control gives you a dedicated system for managing prompt iterations: named versions, side-by-side diffs, A/B testing, and instant rollback.

How Tack Room works

Tack Room is Stockyard's prompt management app. It stores prompt templates with named versions, lets you run A/B experiments between versions, and tracks which version is active in production.

# Create a prompt template
curl -X POST http://localhost:4200/api/studio/templates \
  -d '{"name": "classify-ticket", "version": "v1",
       "system": "Classify this support ticket as: bug, feature, question",
       "model": "gpt-4o-mini"}'

# Create a new version
curl -X POST http://localhost:4200/api/studio/templates \
  -d '{"name": "classify-ticket", "version": "v2",
       "system": "Classify as: bug, feature, question, billing. Reply with ONLY the category.",
       "model": "gpt-4o-mini"}'

# A/B test: 50% traffic to each version
curl -X POST http://localhost:4200/api/studio/experiments \
  -d '{"template": "classify-ticket", "variants": ["v1", "v2"], "split": [50, 50]}'

What you get

Version history. Every prompt change is stored with a version tag, timestamp, and the author. See the full history of how a prompt evolved.

Side-by-side diffs. Compare any two versions of a prompt to see exactly what changed. Useful for debugging when output quality shifts.

A/B experiments. Split traffic between prompt versions and compare output quality, cost, and latency. Make decisions with data instead of gut feel.

Instant rollback. If a new prompt version causes problems, roll back to the previous version with one API call. No redeployment, no code change.

Model + prompt testing. Combine prompt versioning with request replay to test a new prompt against a different model. Find the cheapest model that works with your new prompt.