Local-first · No account · Works offline

A real prompt manager for the prompts you keep retyping.

Chat history is not a library. Bookmarks are not a library. A Google Doc full of pasted blocks is not a library. This is. Save a prompt once - with variables, references, tags, and a system message - and reuse it for years. Nothing leaves your browser.

Save your first prompt in about 30 seconds. No account.

  • 0 accounts
  • 1 JSON file
  • 19 destinations
Doc review - clarity and tone variables: 3 · refs: 2 · tokens: 248
// System
You are a senior editor. Be direct. No hedging.
Quote the exact line you'd cut, then propose the replacement.

// Prompt
Review {{ document }} for {{ audience | technical buyers }}.

Score against {{ ref: style_guide_v4 }} on:
 · Clarity (one idea per sentence)
 · Tone (declarative, no superlatives)
 · Specificity (concrete numbers vs vibes)

Output:
 · 3 lines to cut, with replacements
 · 1 paragraph that's working - say why
 · A {{ ref: voice_examples }} match score, 0-10
Local IndexedDB Every prompt, every tag, every reference lives in your browser's storage. No server holds a copy.
19 AI destinations Send the same prompt to ChatGPT, Claude, Gemini, Perplexity, and 15 others with one click each.
Plain JSON vault One file you choose where to keep. Inspect it in a text editor. Diff it in Git. Email it to yourself in five years.

Why this matters

A prompt library beats chat history. Every time.

Chat history is a transcript. A library is an instrument. One records what you said. The other compresses what you've learned into something you can pull off a shelf and aim at a new problem in five seconds. The difference is the difference between a journal and a toolkit.

What chat history can't do the gap, in concrete terms
// Search in ChatGPT history
"weekly status" -> 47 results across 8 months.
None of them are the version you actually settled on.
The good one was a Tuesday in March. You'll re-write it.

// Search in Prompt Organizer
"weekly status" -> 1 result.
The version you settled on, with variables filled in,
a system prompt attached, and 3 references it pulls from.
Press Run. You're done in 4 seconds.
  1. Your prompts have structure. Use it.System message, body, variables, references, tags. Each one is a first-class field, not a paragraph in a wall of text. You'll edit them like you edit code.
  2. Variables make a prompt reusable.{{ project }}, {{ audience }}, {{ list_price | 18900 }}. Save the shape once, change the values forever. The prompt for one client becomes the prompt for forty.
  3. References make a library compound.{{ ref: style_guide_v4 }} pulls in another saved artifact. Your style guide references your tone notes. Your tone notes reference your client glossary. A graph emerges. It works without you.
  4. Tags make it findable in year three.You won't remember what you titled it. You'll remember it was a writing prompt for a healthcare audience. Two tags. One filter. Found.

Type once. Reuse for years.

The actual loop is shorter than you'd expect. The first time a prompt earns its keep is the moment you stop retyping it. From there, the library starts pulling its own weight.

Minute 1You save the prompt you keep retyping.Probably your weekly status or your doc-review prompt. Paste, add a title, pick two tags, save. It's in IndexedDB before you finish reading this sentence.
Day 3You add variables.The bit that changes between uses becomes {{ project }} or {{ audience }}. The prompt body stops being one-off and starts being a template you fill in.
Week 2You add a reference.Your style guide goes in as a saved artifact. The prompt body becomes "review {{ document }} against {{ ref: style_guide }}". Now the prompt updates itself when the guide does.
Month 3You stop opening a blank ChatGPT tab.You open Prompt Organizer instead, pick the prompt, fill the variables, hit Send. Four seconds. The cold-prompt habit dies on its own.
Your library, week 634 prompts · local only
PromptWeekly status update - exec voiceSummarize {{ project }} for {{ audience | execs }}.
PromptDoc review - clarity and toneReview {{ document }} against {{ ref: style_guide_v4 }}.
RefStyle guide v4Voice: direct, declarative. No exclamations.
PromptCold email - intro to {{ persona }}Open with the metric they care about most.
PromptMeeting recap - decisions onlyStrip the chatter. Lead with the verb.
PromptCode review - security passScan {{ diff }} for auth, escape, SQLi.

34 prompts · 6 refs · 11 tagsstays in this browser unless you export it

Where would you keep these otherwise?

An honest comparison. Each of these works for something - none of them works for the prompts you actually want to refine and reuse.

Prompt Organizer local · no account Notion / Google Docs cloud · org-shared ChatGPT folders / Claude projects vendor-locked
Variables as a real fieldFirst-class, with defaultsHand-typed placeholdersRe-typed each conversation
System prompt as a real fieldSeparate, reusableMixed into one paragraphProject-level, vendor-shaped
References to other promptsYes - {{ ref: ... }}Manual copy-pasteNo
Send to other AI tools19 destinations, one clickCopy-paste onlyTrapped in one vendor
Where it actually livesYour machineVendor serverVendor server
Visible to IT adminNo - local dataYes, in admin consoleYes, by policy
Survives the vendor going awayIt's a JSON file. Yours.Lost / locked behind loginLost / locked behind login
CostFree · $39 once for Pro exports$10-20/seat/month$20-30/seat/month

Save it once. Use it for years.

The prompts you refine three times are worth keeping somewhere designed for the job. Open the workbench, save your first one in 30 seconds, and let the library compound from there. Nothing leaves your browser. Your prompts are yours.