Roadmap

PromptHub's vision: store every LLM prompt as a versioned, typed, testable artifact your application fetches at runtime – so prompt iteration doesn't require a deploy.

Here's the state of each pillar today.

✅ Shipping

• Prompt creation. POST /prompts endpoint and dashboard editor.
• Handlebars template engine with type-aware variable detection.
• Typed variables (string, number, boolean, array, object, enum, binary).

🚧 In active development

• Runtime fetch API. GET /prompts/{name} to fetch the current template at runtime from your application. The first interesting integration point – nothing else really works without it.
• Version history. Append-only changes with named tags (production, staging).
• Prompts list endpoint. So the dashboard list view actually shows your prompts.

📋 Planned

• SDK generation. A prompthub generate command produces type-safe client libraries (Go, TypeScript first; Python and others later). Type-safety means a call like prompthub.welcomeEmail({ user: 42 }) is a compile error if user is a number but the prompt expects a string.
• A/B testing. Multiple active versions of a prompt; the SDK picks one based on a deterministic hash of the request context.
• Evaluation framework. Define golden examples per prompt; run them against new versions to score regressions before promoting.
• Global variables. Org-wide shared values usable inside any prompt's template.
• Team collaboration. Per-prompt access control, review workflows, audit trails.

Get early access

Interested in any of the above? Reach out via the dashboard – we prioritise based on who's asking for what.