PromptLine
In betaWriting
Say what you want, roughly.
You already know what you want — you just don’t want to write four paragraphs to get it. Type the scrappy version. Five models each fix a different thing, in order, and one finished prompt comes back with every change attributable to the stage that made it.
What it does
Five models, five different jobs
One model asked to “improve this prompt” gives you a longer prompt. Five models each given a narrow job give you a better one — expand, source, harden, fact-check, polish — and each stage sees everything the ones before it produced.
Every change is attributable
Open any stage to see exactly what it changed and why: the assumptions it made, the corrections it applied, the sources it checked. Nothing is folded in silently.
A bad stage costs you that stage
If a provider is down or rate-limited the run keeps going, the result is labelled a partial line, and the stage says what went wrong. You still leave with a usable prompt.
Grounded, not recalled
The source stage searches live and returns links. Where it finds nothing it says so rather than inventing a reference — an empty source list beats a plausible fake one.
Refine with notes
Not quite right? Say what should change — shorter, drop the caveat, aim it at beginners — and the finished prompt goes back down the line carrying your notes.
Saves to your Library
Each finished prompt is written to your Library as one document, or copied in a single click.
How a session goes
- 1
Type the rough version of what you want — a single line is enough.
- 2
OpenAI expands it into a real brief: objective, context, constraints, output format.
- 3
Grok gathers the facts, terms of art and references a good answer needs, and cites them.
- 4
Claude pins down versions, units and edge cases — but only when the request is technical.
- 5
Perplexity checks it against reality and the law, then Gemini resolves every note into one prompt in one voice.
What people use it for
- Turning a one-line idea into a brief a model can actually act on
- Technical requests where a vague word produces a confidently wrong answer
- Prompts you will reuse, and want written properly once
- Anything you would otherwise rewrite three times before the answer was usable
Which key it uses
PromptLine calls OpenAI, xAI (Grok), Anthropic (Claude), Perplexity or Google (Gemini). On a paid plan, add a key for any of them under Account → API keys and that provider bills you directly — no credits spent. Without a key, runs spend from your included plan credits instead, and you are told when that happens.
Questions
- Why five models instead of one good one?
- Because “improve this prompt” is not one job. Expanding a brief, finding sources, hardening technical detail and checking for legal problems reward different models, and a single model asked to do all four mostly makes the prompt longer. Each stage here has one narrow job and can see what came before it.
- What does a run cost?
- Five credits — one per stage, or four when the technical pass is skipped as unnecessary. If a stage fails, you are not charged for it.
- Do I need to add API keys?
- No. Every stage runs through Bizwax on platform credits, and PromptLine never asks for or stores a key. If you have added your own key for a provider, that stage runs on it and costs you no credits — the provider bills you directly instead.
- What does it cost to run?
- On a paid plan, if you add your own provider API key the provider bills you directly and Bizwax charges no credits. Without a key, runs spend from the credits included with your plan — and the tool tells you when that happens. Organizations can supply one shared key for the whole team.
- Where does my work go?
- Everything you generate is saved to your Library automatically, so it survives closing the tab. Library is available on Pro and Business plans. You choose what, if anything, to share to Creations.