v1.0.0

01020304

ENGINEER BETTER AI

Open SourceType-SafeComposable

Engineer
Better
AI Prompts

PromptForge helps developers build, validate, optimize, and manage prompts with a powerful, type-safe toolkit.

explain-concept.prompt.ts

123456789101112131415161718

import { pf } from 'promptforge'
import { z } from 'promptforge/schema'

export const explainConcept = pf.define({
  input: z.object({
    concept: z.string().min(1),
    audience: z.enum(['child', 'expert']),
    context: z.string().optional()
  }),
  output: z.object({
    explanation: z.string(),
    confidence: z.number().min(0).max(1)
  }),
  messages: (vars) => [
    pf.system`You are an elite tutor.`,
    pf.user`Explain the concept of: ${vars.concept}`
  ]
})

Type-safe

Validated

Optimized

128 tokens

A premium tool developed by

OMNIKON

THE ISSUE

String Concatenation Spaghetti.

Currently, most developers build AI prompts using massive, messy string templates. This approach is highly problematic in production environments:

01
No Type SafetyIt is easy to forget a variable or inject a malformed object, confusing the LLM.
02
Token BloatDevelopers repeat rules across prompts. The LLM processes "filler words" and duplicate constraints, wasting tokens and driving up API costs.
03
Hard to ScaleYou cannot easily swap out parts of a string or share logic across a massive codebase safely.

THE USE CASE

Real Software Engineering.

PromptForge solves these issues by treating prompt engineering as software engineering, providing a production-ready toolkit:

01
Immutability & ComposabilityBuild libraries of reusable rules (e.g., security constraints) and inject them safely without string concatenation.
02
Pre-Flight ValidationIntercept missing variables or conflicting formats before making expensive network requests.
03
Heuristic OptimizationOur optimizer mathematically strips filler words and deduplicates rules, lowering API bills and latency.

Type-Safe by Default

Define inputs and outputs with Zod for end-to-end type safety.

Composable Prompts

Compose reusable blocks and nest prompts with ease.

Smart Optimization

Reduce tokens, improve clarity, and strengthen instructions.

Seamless Integrations

Export to OpenAI, Anthropic, Gemini, and more.

Command Line Mastery

Validate, analyze, and optimize your prompts directly from your terminal using the PromptForge CLI.

Terminal

$ promptforge analyze src/prompts/email.ts

✓ Analyzed src/prompts/email.ts (152 tokens)

ℹ Insight: Context variable 'user_name' is defined but unused.

$ promptforge optimize src/prompts/email.ts

✓ Optimization complete!

Tokens reduced: 152 ➔ 114 (25% reduction)

Removed redundant instruction: "Make sure to output valid JSON" (Schema handles this).

The PromptForge Architecture

@promptforgee/core

Core SDK

The primary builder API. Create schemas, compose messages, and validate configurations in TypeScript.

@promptforgee/cli

CLI Toolkit

Terminal commands to initialize templates, run audits, and analyze files locally.

@promptforgee/analyzer

Static Analysis

Validates variable interpolation, detects hallucination risks, and ensures structural integrity.

@promptforgee/optimizer

Heuristic Engine

Mathematically reduces token bloat by stripping filler words and deduplicating constraints.

The Omnikon Ecosystem

PromptForge is part of a broader suite of professional developer tools built by Omnikon, designed for resilience and productivity.

Ready to Engineer Better Prompts?

Install the core toolkit today and bring type safety to your AI pipeline.

Read the Docs

npm install @promptforgee/core

Engineer Better AI Prompts