PromptAncestry AI

Prompt optimization & experimentation workspace

Prompt optimization for individual builders.

Analyze, revise, test and version your AI prompts. AI suggestions are treated as hypotheses — only your own test runs count as proof.

Move beyond scattered prompt notes and manual rewrites — keep analysis, testing and prompt versioning in one workspace.

Start optimizing — freeCompare plans

Free includes 2 optimizations per monthly usage period, with version history, branching and locked preferences. No credit card required.

Pro: $11.99/month · Power: $29.99/month · Per account

The Clinical Recruiter sample

V3 and V4 are example prompts, not proven results. No experiment has been run on this sample.

Actual PromptAncestry workspace comparing sample V3 and V4 prompts, locked preferences and highlighted changes
V3 baseline alongside the untested V4 challenger.
Saved sample V4 challenger, linked to V3, with branch, revert, export and promotion controls
A saved sample revision with its parent version and an explicit untested-hypothesis label.

Test a revision before relying on it

Analyze a prompt and generate three alternatives, then compare them using your own criteria and test cases. Repeated trials store actual outputs; AI-judge opinions are labeled subjective.

Keep the version that already works

Preserve each version and its parent. Branch, revert and inspect differences without overwriting earlier work. Lock facts, prices and tone that revisions are instructed to preserve.

Choose with evidence, not polished wording

Review measured data, manual scores and AI-judge opinions separately. Decide which version meets your needs and promote it yourself — the AI does not choose your winner.

Frequently asked questions

Do I need to write code or configure files?

No. Create an account and a prompt project, then enter your prompt, context, and preferences in the workspace. You can compare and save versions without writing code or configuring YAML files.

What does the Free plan include?

Free includes 2 optimizations each monthly usage period. Each successful optimization produces three candidate prompts. Version history, branching, comparisons, and locked preferences are included. AI-judged test runs require Pro or Power. No credit card is required for Free.

Does an AI revision prove my prompt is better?

No. Revisions are untested hypotheses. Run your own test cases and review the outputs before choosing a version. AI-judge scores are subjective signals, not guarantees or claims of statistical significance.

What counts as one test run?

One AI-judged run evaluates one prompt version against one test case. Two versions tested against two cases use four runs per trial. Repeated trials use additional runs.

What happens to my prompts and test data?

Your projects, prompt versions, and experiment outputs are saved in your account. When you start an AI operation, relevant content is sent to an AI provider for processing. Avoid entering sensitive personal information. See our Privacy Policy for the published data-handling details.

Read the Privacy Policy
Are plans for an individual account or a whole organization?

Plans apply to one account, with monthly usage allowances rather than organization-wide seats. Pro is $11.99 per month; Power is $29.99 per month. Unused allowances do not roll over, and there are no automatic overage charges.

Questions? PromptAncestry@pm.me

Our honesty rules

  • • We never show AI output unless it was actually generated for you.
  • • “Proven” is reserved for results from tests you ran.
  • • AI-judge scores are labeled subjective. We don’t claim statistical significance.
  • • A new winner is only set when you approve it.