@jsonfirst/implants

One line of code to make any LLM structured, reliable, and hallucination-resistant — works with ChatGPT, Claude, Gemini, Llama and Mistral.

Install

npm install @jsonfirst/implants

Usage

const implants = require('@jsonfirst/implants');

// List available implants
implants.list();
// → ['universal', 'chatgpt', 'claude', 'gemini', 'llama', 'mistral']

// Get raw JSON string (to paste into your LLM)
implants.get('universal');

// Get as parsed object
implants.getJSON('claude');

// Get all implants
implants.getAll();

Available Implants

| Name | Target | Chat | API | |------|--------|------|-----| | universal | All LLMs | ✓ Paste at start of conversation | Custom Instructions or System Prompt | | chatgpt | ChatGPT | ✓ Paste at start of conversation | Settings → Personalization → Custom Instructions | | claude | Claude | ✓ Paste at start of conversation | Projects → New Project → Set Instructions | | gemini | Gemini | ✓ Paste at start of conversation | Gems → Create a Gem → Instructions | | llama | Llama / Ollama | ✓ Paste at start of conversation | system_prompt in API config | | mistral | Mistral / Le Chat | ✓ Paste at start of conversation | System prompt in Le Chat or Mistral API |

No configuration needed. For instant use, paste any implant directly at the start of your conversation. For persistent governance across all sessions, use the API column to inject at the system level.

What it activates

• Anti-hallucination — zero invented names, prices, IDs or metrics
• Structured prose — no imposed JSON output, no filler openers
• Risk guard — blocks delete/deploy/transfer until explicit confirmation
• Mode confirmation — LLM confirms JSONFIRST_MODE Activé at startup

Protocol

• Spec: JSONFIRST v2.0
• Reference: https://jsonfirst.com
• License: MIT