moltyjacs

Sign it. Prove it. JACS cryptographic provenance plugin for OpenClaw.

Which integration should I use? | Full JACS documentation

Why use JACS?

So your OpenClaw agent can be trusted -- and can trust others. JACS is like DKIM for agents: you sign what you send; recipients verify the signature against your public key. It is decentralized -- no single authority. You publish your key (DNS, optional HAI.ai); others fetch and verify. Without it, nothing you say or do can be proven. With JACS you sign messages, commitments, and state; anyone with your public key can verify they came from you and were not altered. You get proof of origin, integrity, and accountability. Other agents can discover your key via DNS or HAI.ai and verify your documents; you verify theirs with jacs_verify_auto and optional trust levels (domain, attested). Keys and signed payloads stay local; you send the same signed JSON over any channel (WhatsApp, HTTP, MCP). Use it whenever another agent or human needs to trust that you said or agreed to something.

Overview

moltyjacs adds post-quantum cryptographic signatures to your OpenClaw agent communications. It enables:

• Document signing - Sign any document with your agent's cryptographic identity
• Verification - Verify documents from other agents
• Agent discovery - Publish and discover agents via well-known endpoints and DNS
• A2A interoperability - Export Agent Cards, wrap A2A artifacts with JACS provenance, and assess remote A2A trust
• Multi-party agreements - Create and manage agreements requiring multiple signatures
• Agent state - Sign and track memory, skills, plans, configs, and hooks
• Commitments - Track agreements and obligations between agents with lifecycle management
• Todo lists - Private, signed work tracking with goals and tasks
• Conversations - Signed message threads between agents
• HAI platform features - Hello/auth checks, username lifecycle, mailbox/email workflow, key registry lookups, and benchmark orchestration

Installation

From npm

npm install moltyjacs

From ClawHub

npx clawhub@latest install moltyjacs

As OpenClaw Plugin

openclaw plugins install moltyjacs

From GitHub

openclaw plugins install https://github.com/HumanAssisted/moltyjacs

Quick Start

1. Set a password for key generation

You need a password once — during jacs init — to encrypt your new private key. Pick whichever method fits your setup:

# Local dev / headless — set an env var
export JACS_PRIVATE_KEY_PASSWORD='use-a-strong-password'

# Containers / CI — point to a secrets-mounted file
export JACS_PASSWORD_FILE=/run/secrets/jacs_password

After init, the password is only needed to decrypt your private key for signing. On macOS the OS Keychain stores it automatically, so you can skip the env var for day-to-day use. On Linux/CI, keep one of the env vars above set.

2. Reserve a username and initialize

Reserve your username at https://hai.ai/dashboard and copy the registration key.

openclaw jacs init --name myagent --key YOUR_REGISTRATION_KEY

This generates a key pair, creates jacs.config.json, registers with HAI, and assigns myagent@hai.ai. Your agent now has a cryptographic identity and is registered.

For local-only init (no registration): openclaw jacs init --name myagent --register=false

3. Sign and verify

openclaw jacs sign document.json
openclaw jacs verify signed-document.json

4. Bootstrap trust with another agent

When your agent needs to trust (or be trusted by) another agent:

1. Sender shares credentials via tools: jacs_share_public_key + jacs_share_agent
2. Receiver imports them: jacs_trust_agent_with_key with the shared agentJson and publicKeyPem

Once trust is established, use jacs_verify_auto to verify any document from that agent (keys are fetched automatically).

Direct JACS SDK Quick Start (outside this plugin)

For direct @hai.ai/jacs/client or @hai.ai/jacs/simple usage:

import { JacsClient } from "@hai.ai/jacs/client";

const client = await JacsClient.quickstart({
  name: "my-agent",
  domain: "agent.example.com",
  algorithm: "pq2025", // optional, this is the default
});

JACS Workspace Compatibility

For coordinated local development, moltyjacs now resolves @hai.ai/jacs from the sibling ../JACS/jacsnpm workspace so its A2A surface matches the current haiai wrapper contract. The JACS Node API remains async-first. All NAPI operations return Promises by default; sync variants use a Sync suffix (e.g. loadSync vs load). moltyjacs uses the async API for setup (agent.load(), createAgent()) and the sync API for hot-path operations (signRequest, verifyResponse) that must run on the V8 thread.

Recent JACS updates relevant to moltyjacs:

• Direct quickstart() usage in @hai.ai/jacs/client and @hai.ai/jacs/simple now requires identity inputs (name and domain) for first-time agent creation.
• Default algorithm across JACS is pq2025.
• Trust/bootstrap surfaces now include trustAgentWithKey / trust_agent_with_key, sharePublicKey / share_public_key, and shareAgent / share_agent.
• A2A surfaces now include Agent Card export, wrapped-artifact signing/verification, and generated well-known discovery documents.

CLI Commands

HAI.ai registration

To get an attested trust level, register your agent with HAI.ai during init: run openclaw jacs init --name <name> --key <key> with a registration key from your HAI dashboard. Registration uses JACS-signed authentication (no API key needed). Alternatively, use the jacs_hai_register tool with a registrationKey after init. After registration, use openclaw jacs attestation to check your (or another agent's) attestation status, and openclaw jacs claim <level> to set or view your verification claim. verified now requires DNS TXT hash verification (domain configured + published hash matches your public key). See Configuration and Security for related options.

Agent Tools

When used with an AI agent, these tools are available:

Core signing and verification

A2A interoperability

Discovery and trust

HAI platform integration

Agreements

Agent state

Commitments

Todo lists

Conversations

MCP and Message Transport

jacs_start_conversation and jacs_send_message create signed JACS message payloads. They do not perform delivery/transport by themselves.

Use this pattern for agent-to-agent messaging:

1. Create/sign payload (jacs_start_conversation or jacs_send_message)
2. Deliver the returned signed JSON over your chosen channel (MCP, HTTP, queue, chat bridge, etc.)
3. Verify on receipt (jacs_verify_auto, jacs_verify_standalone, or jacs_verify_with_key)

For custom Node MCP servers, JACS supports transport-level integration via @hai.ai/jacs/mcp:

• createJACSTransportProxy(...) for automatic signing/verification at transport boundaries
• registerJacsTools(...) to expose JACS operations as MCP tools
• Expanded trust/bootstrap MCP/LangChain tools include jacs_share_public_key, jacs_share_agent, and jacs_trust_agent_with_key

This OpenClaw plugin does not automatically intercept all host MCP traffic; use explicit JACS tools or host transport middleware/adapters.

Well-Known Endpoints

Your agent exposes these endpoints:

• GET /.well-known/agent-card.json - A2A Agent Card for discovery
• GET /.well-known/jwks.json - A2A/JACS JWKS for verifier interoperability
• GET /.well-known/jacs-agent.json - JACS agent descriptor
• GET /.well-known/jacs-extension.json - JACS A2A extension descriptor
• GET /.well-known/jacs-pubkey.json - Your public key
• GET /jacs/agent - Current self-signed JACS agent document
• GET /jacs/status - Health check
• POST /jacs/verify - Public verification (this agent)
• GET /jacs/attestation - Full attestation status (trust level, HAI registration, DNS verification)

Recipients can verify any JACS document at https://hai.ai/jacs/verify (paste a link with ?s= or the base64). Use jacs_verify_link or the verification_url from jacs_sign when sharing signed content with humans.

Signing is internal only; no external sign endpoint is exposed (to protect the agent's identity).

Configuration

Configure via openclaw.plugin.json:

{
  "keyAlgorithm": "pq2025",
  "agentName": "My Agent",
  "agentDescription": "Description",
  "agentDomain": "agent.example.com"
}

autoSign and autoVerify are accepted for backward compatibility but are deprecated no-ops in moltyjacs.

agentId is set automatically when you run openclaw jacs init and is not edited in the config file.

JACS key filenames are read from jacs.config.json:

• jacs_agent_public_key_filename (default: jacs.public.pem)
• jacs_agent_private_key_filename (default: jacs.private.pem.enc)

Environment variables

Set one of these for init and for Linux/CI signing. On macOS the OS Keychain is used automatically at runtime when neither is set. If both are set, initialization fails to avoid ambiguity.

Password files must be owner-only on Unix (chmod 600).

Key Algorithms

• pq2025 (default) - Post-quantum ML-DSA-87
• ring-Ed25519 - Ed25519
• RSA-PSS - RSA with PSS padding

Cross-Language Compatibility

Documents signed by moltyjacs (Node.js) can be verified by Rust or Python agents, and vice versa. Cross-language interop is tested on every commit with Ed25519 and post-quantum (ML-DSA-87) algorithms. See the JACS cross-language tests for details.

Security

• Private keys are encrypted with AES-256-GCM
• Key derivation uses PBKDF2
• Default algorithm (pq2025) provides quantum resistance
• DNS records enable DNSSEC-backed identity verification

Development

# Install dependencies
npm install

# Build
npm run build

# Watch mode
npm run watch

# Run unit tests (uses mocked JACS module)
npm test

# Run integration tests (requires real @hai.ai/jacs native binary)
npm run test:integration

# Test local installation
openclaw plugins install . --link
openclaw plugins list

Publishing

CI publishes on push of a tag v* (e.g. v0.8.0). Publish @hai.ai/jacs from the JACS repo first (tag npm/v*), then tag and push moltyjacs so the build can resolve the dependency.

To npm

npm run build
npm publish

To ClawHub

npm run clawhub:publish

Or publish to both npm and ClawHub:

npm run publish:all

License

MIT License - see LICENSE

Links

• HAI.AI
• JACS Documentation
• Decision Tree
• OpenClaw
• ClawHub
• GitHub