JACS: JSON Agent Communication Standard

Welcome to the JSON Agent Communication Standard (JACS) documentation! JACS is a comprehensive framework for creating, signing, and verifying JSON documents with cryptographic integrity, designed specifically for AI agent communication and task management.

What is JACS?

JACS provides a standard way for AI agents to:

  • Create and sign JSON documents with cryptographic signatures
  • Verify authenticity and integrity of documents
  • Manage tasks and agreements between multiple agents
  • Maintain audit trails of modifications and versioning
  • Ensure trust in multi-agent systems

As a developer, JACS gives you the tools to build trustworthy AI systems where agents can securely exchange tasks, agreements, and data with verifiable integrity.

What Can You Do with JACS?

Sign AI agent output and verify every LLM response with cryptographic proof. JACS enables agent-to-agent trust through non-repudiation -- every action is signed and independently verifiable. Add MCP authentication to your tool servers, integrate LangGraph signing into your workflows, and build multi-agent systems where no output goes unverified. Whether you need to prove who said what or establish a chain of custody for AI-generated content, JACS provides the cryptographic foundation.

Key Features

  • 🔐 Cryptographic Security: RSA, Ed25519, and post-quantum cryptographic algorithms
  • 📋 JSON Schema Validation: Enforced document structure and validation
  • 🤝 Multi-Agent Agreements: Built-in support for agent collaboration and task agreements
  • 🔍 Full Audit Trail: Complete versioning and modification history
  • 🌐 Multiple Language Support: Rust, Node.js, and Python implementations
  • 🔌 MCP Integration: Native Model Context Protocol support
  • 📊 Observability: Built-in logging and metrics for production systems

Available Implementations

JACS is available in three languages, each with its own strengths:

🦀 Rust (Core Library + CLI)

  • Performance: Fastest implementation with native performance
  • CLI Tool: Complete command-line interface for agent and document management
  • Library: Full-featured Rust library for embedded applications
  • Observability: Advanced logging and metrics with OpenTelemetry support

🟢 Node.js (@hai.ai/jacs)

  • Web Integration: Perfect for web servers and Express.js applications
  • MCP Support: Native Model Context Protocol integration
  • HTTP Server: Built-in HTTP server capabilities
  • NPM Package: Easy installation and integration

🐍 Python (jacs)

  • AI/ML Integration: Ideal for AI and machine learning workflows
  • MCP Support: Authenticated MCP server patterns
  • PyPI Package: Simple pip install integration
  • Data Science: Perfect for Jupyter notebooks and data pipelines

Quick Start

Choose your implementation and get started in minutes:

Rust CLI

cargo install jacs --features cli
# Upgrade to latest: cargo install jacs --features cli --force
jacs init  # Create config, keys, and agent

Or step by step:

jacs config create
jacs agent create --create-keys true

Node.js

npm install @hai.ai/jacs

import { JacsAgent } from '@hai.ai/jacs';

const agent = new JacsAgent();
agent.load('./config.json');

Python

pip install jacs

import jacs

agent = jacs.JacsAgent()
agent.load("./config.json")

When to Use JACS

JACS is ideal for scenarios where you need:

  • Multi-agent systems where agents need to trust each other
  • Task delegation with verifiable completion and approval
  • Audit trails for AI decision-making processes
  • Secure data exchange between AI systems
  • Compliance requirements for AI system interactions
  • Version control for AI-generated content and decisions

Why JACS?

🎯 Agent-Focused Design

Unlike general-purpose signing frameworks, JACS is specifically designed for AI agent communication patterns - tasks, agreements, and collaborative workflows.

🚀 Production Ready

With built-in observability, multiple storage backends, and comprehensive error handling, JACS is ready for production AI systems.

🔒 Future-Proof Security

Support for both current (RSA, Ed25519) and post-quantum cryptographic algorithms ensures your system remains secure.

🌐 Universal Compatibility

JSON-based documents work everywhere - store them in any database, transmit over any protocol, integrate with any system.

🧩 Flexible Integration

Whether you're building a simple CLI tool or a complex multi-agent system, JACS adapts to your architecture.

Getting Started

  1. Core Concepts - Understand agents, documents, and agreements
  2. Quick Start Guide - Get up and running in minutes
  3. Choose Your Implementation:

Community and Support

  • GitHub: HumanAssisted/JACS
  • Issues: Report bugs and feature requests
  • Examples: Complete examples for all implementations
  • Documentation: This comprehensive guide

Ready to build trustworthy AI systems? Let's get started!