Your prompt library, organized.
Keep every prompt in one structured, searchable place. Import thousands at a time, track each one through a real workflow, and drive your generation pipeline through an API — then publish the packs you want to share.
Looking for prompts to use? Browse the free marketplace →
What it does
Everything a prompt library needs, and nothing it does not.
A hierarchy that survives ten thousand prompts
Category → Sub-category → Bundle → Pack. Four levels, so a prompt has a real address instead of a tag soup. Every level is yours to name, and nothing is capped.
An API that drives generation
Lock the next pending prompt, generate against it, complete it. Your worker never picks the same prompt twice, and a crashed worker releases its lock automatically.
Bulk import and export that do not fall over
CSV or JSON, streamed rather than buffered, inserted in batches. Move thousands of prompts in or out without the request timing out or the memory spiking.
A status workflow you can filter on
Pending, Queued, Generated, Inactive. Know what is waiting, what a worker holds right now, what is done, and what you have parked — at a glance and in a query.
Metadata that is a vocabulary, not free text
Type, Engine, Source and Distribution are shared, seeded dimensions you can extend. Tags stay free-form for the long tail. Filter by any of them, alone or combined.
Workspaces with real roles
Invite people as Admin, Editor or Viewer. An Editor can add and edit but cannot delete — the exact shape you want for a freelancer.
Structure
Four levels, so a prompt has an address.
Flat lists work until they do not. A tag-only system collapses somewhere around a hundred prompts, because two people tag the same thing differently and neither can find it a month later.
A Category is the broad domain. A Sub-category narrows the topic. A Bundle is your operational grouping — a client, a campaign, a batch. A Pack is the set you actually work with, and the unit you publish if you choose to.
Every level is free text you control, and none of it is capped. Filters compose across all four, plus status, type and every lifecycle dimension.
Automation
An API built for a generation pipeline, not just for reads.
Most tools give you a read endpoint and call it an API. The problem with running prompts at scale is not reading them — it is making sure two workers never grab the same one, and that a crashed worker does not strand it forever.
So the flow is a lease, not a fetch:
POST /api/v1/prompts/lockClaims the next Pending prompt and marks it Queued, so two workers never collide.…run your model…Whatever you use — the library does not care which one.POST /api/v1/prompts/completeMarks it Generated. If your worker dies instead, the lock expires and the prompt returns to Pending on its own.
Personal API keys authenticate your scripts. Keys are stored only as a hash — we cannot read them back — and you can revoke one at any time.
API=https://psapi.degird.com
KEY=sk_1a2b3c4d…
# 1 — claim the next prompt waiting for work
curl -X POST "$API/api/v1/prompts/lock" \
-H "Authorization: Bearer $KEY" \
-H "Content-Type: application/json" \
-d '{"script_id":"worker-1","type":"Text"}'
# → { "id": "a1b2c3d4", "prompt": "Write an…" }
# 2 — run your model, then hand the prompt back
curl -X POST "$API/api/v1/prompts/complete" \
-H "Authorization: Bearer $KEY" \
-H "Content-Type: application/json" \
-d '{"prompt_id":"a1b2c3d4"}'Your data
Get everything in, and everything back out.
Import
CSV or JSON. The file is streamed rather than read into memory and written in batches, so a large file imports without timing out. The hierarchy is created as it goes, and lifecycle metadata round-trips intact.
Export
Everything, or one pack, as CSV or JSON — streamed the same way. Exports are safe to open in a spreadsheet: any cell that could be read as a formula is escaped, which is a real vulnerability most exporters ship with.
Teams
One library, several people, no accidents.
| Can… | Owner | Admin | Editor | Viewer |
|---|---|---|---|---|
| Read and export everything | ||||
| Create and edit prompts | — | |||
| Delete prompts | — | — | ||
| Run bulk operations | — | — | ||
| Publish a pack | — | — | ||
| Edit the store, invite people | — | — | — |
Editor is the role that matters. It is the one you hand a freelancer: they can add and improve prompts all day and cannot delete a single one. Removing someone revokes their access immediately, along with any API keys they created in your workspace.
The other half
Private by default. Public when you say so.
Nothing in your library is visible to anyone else unless you publish it, and publishing takes two deliberate switches — one to open your store, one per pack. Both have to be on. Nothing is public by accident.
When you do publish, the pack gets its own indexable page on the marketplace, a permanent URL that survives you renaming things, and a place in the automated discovery rails. Placement cannot be bought — see how packs get featured.
Before anything goes live, the publish step scans for credentials and blocks the publish if it finds one — and shows you every field of yours that is about to become public.
The two gates
Your store is published
Off by default. While it is off, your storefront 404s and every pack in it is invisible.
This pack is published
Per pack, on top of the store switch. Both must be on for anything to appear.
Questions
What people ask before signing up.
What is a prompt library?
A single place to keep every prompt you use, organized well enough to find one months later, with the metadata you need to know where it came from and what it is for. Once you pass a few dozen prompts, notes files and spreadsheets stop working — you cannot filter them, two people cannot edit them safely, and nothing can read them programmatically.
Is it free?
Yes. The prompt library, the public marketplace and every pack on it cost nothing. There is no paid tier, no seat pricing and no card. Reading prompt bodies on the marketplace requires a free account; browsing the catalogue does not.
Can I use my prompts outside Prompt Share?
That is the point. Personal API keys let your own scripts read and update prompts programmatically, and bulk export gives you everything as CSV or JSON whenever you want. Your prompts are yours — nothing is locked in.
How do I move prompts I already have?
Bulk import accepts CSV and JSON. The file is streamed rather than loaded into memory and inserted in batches, so a file with thousands of rows imports without timing out. The category, bundle and pack hierarchy is created as it goes, so you do not have to build it first.
Can my team work in the same library?
Yes. Invite people by email and give each a role — Admin, Editor or Viewer. They work on your data, not their own copy, and their role decides what they may do: Editors can create and edit but not delete, Viewers can read and export, and only the owner can invite people or change the store.
What is the difference between the library and the marketplace?
The library is private by default — nobody sees your prompts unless you decide otherwise. The marketplace is the public side: you open a store, choose which packs to publish, and they appear for everyone. Publishing takes two deliberate switches, and nothing is ever public by accident.
Do I have to publish anything?
No. Plenty of people use the library and never open a store. Publishing is entirely opt-in, one pack at a time, and unpublishing is immediate.
Ready to organize your prompt library?
Free forever, no card, and your prompts export in one click if you ever want to leave.
Create your free account