Bring your dataset
Drop in labelled examples — the input/output pairs every candidate prompt is scored against.
Everything you need
Downloads.
The source, the CLI, sample datasets and configs, and the prompt templates that drive every round — all in one place.
How it works
Three plain steps from your dataset to a better prompt — no commands to memorize.
Start the backend
Point PromptPotter at the pipeline you want to improve. It reads the tunable params and never edits your code.
Open your browser
Watch the loop generate, score and critique — round after round — until your numbers move.
Install
Free and open source — no license keys, no seats. Pick how you want to run it.
CLI
Runs where your data already is. Your examples and your API keys stay on your own machine — nothing is uploaded, and there is no account to create before you can try it. Start a campaign, close the laptop, pick it up where it left off.
# install — needs Python 3.13 or newer
git clone https://github.com/PromptPotter/prompt-potter-optimizer
cd prompt-potter-optimizer
pip install -e ".[all]"
# point it at your provider
echo "OPENROUTER_API_KEY=your_key" > .env
# start a campaign
python -m promptpotter new my-first-campaign
# pick it back up, any time
python -m promptpotter resume
Datasets & configs
Drop-in examples so you can watch the loop move on round one — no setup.
Prompt templates
The meta-prompts that drive each stage of the loop — read the source that checks in, generates, critiques, reframes the task, and replans.