Getting Started

Deploy your generated system prompt on any LLM platform.

Agent Persona Forge produces a universal system prompt and exports it in the native format of your chosen provider. Pick your provider in the wizard, then follow the matching guide below.

Anthropic Claude / Claude Code

1. Generate your CLAUDE.md

Pick the Anthropic Claude provider in the wizard and download CLAUDE.md.

2. Drop it at your project root

Save next to package.json. Claude Code auto-detects it on launch. For Claude API, paste the file contents as the system parameter.

3. Verify with /memory

Run /memory in Claude Code — your file should be listed. Commit to git so the whole team gets the same agent.

OpenAI Assistants API

1. Generate your assistant_config.json

Pick OpenAI Assistants in the wizard. Download assistant_config.json.

2. Create the assistant

import OpenAI from "openai";
import config from "./assistant_config.json";

const openai = new OpenAI();
const assistant = await openai.beta.assistants.create(config);

3. Use the assistant ID

Save assistant.id and reference it from any thread. Update by calling openai.beta.assistants.update(id, config).

ChatGPT Custom GPT

1. Generate your instructions

Pick ChatGPT Custom GPT. Download custom_gpt_instructions.txt.

2. Paste into the Configure tab

Go to chatgpt.com → My GPTs → Create or Edit → Configure. Paste into the Instructions field (8000-char limit).

3. Save and test

Click Update. Ask the GPT a domain-specific question to verify it adopted the persona.

Google Gemini

1. Generate your gemini_system_instruction.txt

Pick Google Gemini in the wizard.

2. Pass as systemInstruction

import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI(API_KEY);
const model = genAI.getGenerativeModel({
  model: "gemini-2.5-pro",
  systemInstruction: SYSTEM_INSTRUCTION_TEXT,
});

3. Or paste in AI Studio

In Google AI Studio or Vertex AI, paste into the System Instructions field of your prompt.

Mistral AI

1. Generate your system_prompt.txt

Pick Mistral AI in the wizard.

2. Send as the system message

messages: [
  { role: "system", content: SYSTEM_PROMPT },
  { role: "user", content: userInput }
]

3. Or use in le Chat

Create a new agent in le Chat and paste the prompt as its instructions.

Generic / Custom LLM

1. Generate the universal system_prompt.txt

Pick Generic / Custom LLM. You get portable Markdown that works in any chat-completion API.

2. Use as the system message

For Llama, Qwen, DeepSeek, Ollama, vLLM, LangChain, LlamaIndex, or any custom agent framework: pass the file contents as the system / role-system message.

3. Iterate

Edit the file like source code. Commit it to git alongside your agent code.

Generate yours now

60-second wizard. Production-ready output. Multi-format export.

Open the generator →