Write Prompts Like Code.

Build type-safe, composable, production-ready prompts with TypeScript. Replace fragile string templates with a modern developer workflow.

Open Studio

1. Fragile String

const prompt = 
  "You are a helpful assistant\n" +
  "Context: " + data.context + "\n" +
  "User: " + req.query; // Breaks easily

2. Type-Safe API

const prompt = pf.define({
  role: 'helpful assistant',
  context: data.context,
  task: req.query
}); // Compiled & Validated

String concatenation does not scale.

As your AI features grow, managing large prompt templates with string interpolation becomes a maintenance nightmare. Unescaped characters break JSON outputs, context gets duplicated, and refactoring is impossible.

PromptForge replaces template literals with a structured, strongly-typed compiler designed specifically for Large Language Models.

const systemPrompt = `
You are an expert financial analyst.
Here is the user's portfolio data:
${JSON.stringify(userData)}
Please return a JSON array of... wait what if userData contains quotes?
`;

Why PromptForge?

Without PromptForge

With PromptForge

Manual Strings

Type-safe API

No Validation

Compile-time Validation

Provider-specific formatting

Provider Agnostic Compilation

Difficult Maintenance

Composable Prompts

The Prompt Pipeline

Write

Compile

Validate

Analyze

Optimize

Deploy

Code First.

Define your inputs, outputs, and constraints using Zod schemas. PromptForge handles the markdown generation and LLM formatting automatically.

agent.ts

import { pf } from '@promptforgee/core';

export const summarize = pf.define({
input: z.object({ text: z.string() }),
output: z.object({
summary: z.string(),
keywords: z.array(z.string())
}),
messages: (vars) => [
pf.system`You are a strict technical summarizer.`,
pf.user`Summarize this: ${vars.text}`
]
});

COMPILEDOPENAIANTHROPIC

SYSTEM: You are a strict technical summarizer.

USER: Summarize this: {vars.text}

OUTPUT FORMAT:
{ "summary": "string", "keywords": ["string"] }

Analyzer

Static analysis and real LLM evaluation grades your prompts before they hit production.

SCORE

95/100

TOKENS

124

Perfect task isolation

Structured output enforced

Optimizer

Run your code through the optimizer to strip filler words and strictly enforce prompt engineering best practices.

BEFORE

"Please somehow summarize this stuff if you can..."

AFTER

"Summarize the provided text."

Saved: 24% TokensRemoved Ambiguity

Production Ready Examples

RAG Pipeline Agentic Loop Summarization Security Audit SQL Generator Data Extraction Classification Code Review

Load these directly into Studio →

Developer Experience First.

TypeScript First
Tree Shakeable
ESM & CommonJS
Zero Dependencies (Core)
Provider Agnostic

Open Source & Community.

GitHubMIT License npmLatest Version

Ready to standardize your prompts?

Open Studio