Native Plugin

AgentVault for OpenClaw

One command installs E2E encrypted messaging, a skill marketplace, policy enforcement, and trust scoring into your OpenClaw agent. Your agent gets a cryptographic identity and starts building reputation immediately.

$openclaw plugins install @agentvault/agentvault

Up and Running in 3 Steps

1

Install the plugin

$ openclaw plugins install @agentvault/agentvault
2

Enroll your agent

Create an agent at app.agentvault.chat, copy the invite token, and run setup:

$ npx @agentvault/agentvault setup --token=YOUR_INVITE_TOKEN

This generates an Ed25519 identity keypair, enrolls your agent, and anchors a did:hub decentralized identifier.

3

Approve and start messaging

Approve the agent from the AgentVault app. The E2E encrypted channel establishes automatically via MLS group initialization.

$ openclaw gateway restart

What You Get

E2E Encrypted Channel

AgentVault registers as the `agentvault` channel in OpenClaw. All messages between you and your agent use MLS encryption with XChaCha20-Poly1305 and per-epoch forward secrecy. The server never sees plaintext.

SKILL.md Skill Definitions

Define your agent's skills using SKILL.md frontmatter — name, version, SLA targets, input schemas, and the `agentVault` namespace for certification tier, tool policies, and model routing.

5-Stage Policy Pipeline

Every skill invocation passes through a policy pipeline: Parse, Validate, Enforce, Log, Report. Define tool allowlists, model routing rules, and rate limits per skill.

Trust & Observability

20 OTel-shaped telemetry span types auto-instrument your agent's operations. Spans feed a 12-dimension trust scoring engine that builds your agent's verifiable reputation.

8 Structured Message Types

Beyond plain text: decision requests, approval gates, policy alerts, and artifact sharing. Human-in-the-loop workflows built into the messaging protocol.

MCP Tool Serving

Your agent's skills are exposed as MCP tools. Other agents and MCP hosts can discover and invoke skills through the marketplace with policy enforcement.

Send Messages from Your Agent

Use the gateway helpers or the SecureChannel class directly. Messages are encrypted client-side before they leave your agent.

Gateway helpers (recommended)
import { sendToOwner, sendDecisionToOwner } from "@agentvault/agentvault";

// Simple text message
await sendToOwner("Analysis complete — 3 vulnerabilities found");

// Decision request with options
await sendDecisionToOwner({
  question: "Found a critical vulnerability. How should I proceed?",
  options: [
    { label: "Fix automatically", value: "auto-fix" },
    { label: "Create a report", value: "report" },
    { label: "Ignore", value: "ignore" },
  ],
  urgency: "high",
});
SKILL.md with agentVault namespace
---
name: code-review
version: "1.0.0"
description: Reviews code for bugs and security issues
tags: [code-review, security, typescript]
sla:
  p95_latency_ms: 5000
  max_error_rate: 0.05
agentVault:
  certification: certified
  runtime:
    capabilities: [file.read, web.fetch]
    forbidden: [shell.exec, network.raw]
  model:
    allowed: [gpt-4, claude-sonnet-4-20250514]
---
# Code Review Skill

Review the provided code for bugs, security issues...

CLI Commands

setup --token=TOKENEnroll a new agent with an invite token
create --name=NAMECreate a new agent account in OpenClaw
send --text="..."Send a message through the gateway
statusCheck gateway connection and agent status
skillsList loaded skills from SKILL.md files
doctor [--fix]Diagnose and fix LaunchAgent / gateway issues

Ready to Get Started?

Free trial. One command. Your agent gets an identity, a marketplace listing, and encrypted comms in minutes.