AstraCipher

Cryptographic Identity & Trust Protocol for AI Agents

The "SSL certificates" for the AI agent economy. Open-source protocol that gives every AI agent a verifiable, cryptographic identity.

The Problem

AI agents are operating across enterprise systems with zero identity verification. No one can answer:

• Which agent performed this action?
• Was it authorized?
• Can we prove compliance to regulators?

MCP servers expose powerful tools, but any agent can call any tool. There's no authentication, no authorization, no audit trail.

The Solution

AstraCipher is a W3C-standards-based protocol that provides:

• Decentralized Identifiers (DIDs) --- Unique, cryptographic identity for every agent (did:astracipher:mainnet:abc123)
• Verifiable Credentials --- Signed attestations of capabilities, permissions, and trust levels
• Trust Chains --- Delegated authority with depth limits (Creator -> Authorizer -> Agent -> Sub-agent)
• Post-Quantum Cryptography --- ML-DSA-65 + ECDSA P-256 hybrid signatures (FIPS 204 compliant)
• Compliance Modules --- Generate regulatory-ready reports for 10+ frameworks worldwide

Why Now

• 850M+ AI agents expected by 2030 (Gartner)
• MCP adopted by Anthropic, OpenAI, Google, Microsoft --- but has no identity layer
• AAIF (Linux Foundation + Anthropic) defines agent interoperability --- AstraCipher provides the missing identity primitive
• EU AI Act enforcement begins 2025-2026, requiring traceability for high-risk AI systems
• NIST AI RMF and ISO 42001 becoming enterprise prerequisites

Quick Start

CLI

# Install the CLI
npm install -g @astracipher/cli

# Initialize AstraCipher in your project
astracipher init

# Generate post-quantum key pair
astracipher keygen --algo hybrid

# Create an agent identity (DID)
astracipher create --name "my-data-agent" --key .astracipher/keys/agent.pub.json

# Issue a credential
astracipher issue \
  --did did:astracipher:testnet:abc123 \
  --capabilities read,write \
  --trust-level 8 \
  --validity 365d

# Verify a credential
astracipher verify --credential ./credential.json

SDK (TypeScript)

import { AstraCipherClient } from '@astracipher/core';
import { HybridKeyManager } from '@astracipher/crypto';

const keyManager = new HybridKeyManager();
const keyPair = await keyManager.generateKeyPair('hybrid');

const client = new AstraCipherClient({ keyManager });
const did = await client.createDID('my-agent', keyPair);
const credential = await client.issueCredential(did, {
  capabilities: ['read', 'write'],
  trustLevel: 8,
});
const result = await client.verifyCredential(credential);

MCP Integration

Any MCP-compatible AI agent (Claude, GPT, etc.) can use AstraCipher tools:

{
  "mcpServers": {
    "astracipher": {
      "command": "npx",
      "args": ["@astracipher/mcp-server"]
    }
  }
}

Available MCP tools:

• create_agent_identity --- Create a DID for an agent
• verify_agent --- Verify an agent's credential
• check_permissions --- Check agent permissions for a resource
• inspect_credential --- View credential details

Architecture

+----------------------------------------------------------+
|                    AstraCipher Protocol                     |
+---------------+----------------+-------------------------+
|  @astracipher/  |  @astracipher/   |  @astracipher/            |
|    crypto     |     core       |   compliance-*          |
|  (PQC keys,   |  (DIDs, VCs,   |  (DPDP, EU AI Act,     |
|   signing)    |  trust chain)  |   G

... [View full README on GitHub](https://github.com/san-techie21/astracipher#readme)