PromptVault
A versioned prompt registry for teams to write, review, and ship LLM prompts the way they ship code.
PromptVault is a versioned prompt registry: a place to write, review, and ship LLM prompts the way teams ship code. You edit prompts in the dashboard, read them from application code with the SDK, and serve the live or staged version on demand.
Edit here, runs there. PromptVault keeps prompt source and version history in one place while your observability, eval, and audit tools stay connected through integrations.
What you'll find in these docs
| Section | Read this if you want to... |
|---|---|
| Getting started | Set up your workspace and make your first SDK call. |
| Dashboard | Learn the UI: prompts, drafts, versions, activity, evaluator, members. |
| API reference | Read prompts and version history directly over the HTTP API. |
| SDK | Fetch prompts from application code with @promptv/sdk. |
| Integrations | Connect PromptVault to Langfuse, Helicone, Braintrust, PostHog, GitHub, or a webhook. |
How it fits together
Dashboard
- write and review prompts
- manage versions and staging
- connect integrations
SDK
- fetch live or staged prompts
- interpolate variables
- cache prompts for low-latency reads
Your app
- calls the SDK
- sends the returned prompt to your LLM provider
- attaches version metadata to observability toolsVersioning model in 30 seconds
- A prompt is identified by a workspace-unique slug.
- Versions are append-only and labelled
v1,v2, and so on. - Each prompt has a live version and can optionally have a staging version.
- Drafts let you save work without changing what production serves.
pv_live_keys read live prompts.pv_test_keys read staging when available and fall back to live.
Everything in the dashboard and SDK exists to make those version moves explicit and reversible.