OpenClawIntegrationMCPIdentity

Giving OpenClaw Agents Identity, Permissions, and Audit Trails

Kwame Nyantakyi

April 18, 2026 · 5 min read

Your Agent Has Root Access. Now What?

OpenClaw is one of the most powerful personal AI agents available today. It runs locally on your machine, controls your browser, executes shell commands, reads and writes files, and integrates with 50+ services including Gmail, GitHub, Slack, and Spotify. It has 120K+ GitHub stars and a growing ecosystem of 15K+ community skills.

That power comes with a problem: there's no built-in way to verify which agent did what, enforce what it's allowed to do, or produce a tamper-proof record of its actions. If your OpenClaw agent sends an email you didn't intend, deletes a file it shouldn't have, or accesses a service without authorization, there's no identity trail to trace.

Today we're launching the Vorim skill for OpenClaw, giving every OpenClaw agent a cryptographic identity, scoped permissions, and a signed audit trail.

What This Integration Does

The Vorim skill adds three capabilities to your OpenClaw agent:

1. Cryptographic identity. When your OpenClaw agent registers with Vorim, it receives an Ed25519 keypair and a unique agent ID. This isn't a shared API key. It's a verifiable, per-agent identity that can be validated by any external service your agent interacts with.

2. Permission enforcement. Before your agent takes a sensitive action (running a shell command, sending an email, making a payment), it checks with Vorim whether it has the right permission scope. Seven scopes are available, from agent:read (browsing and searching) to agent:elevate (escalating privileges). If the permission check fails, the agent stops and informs you.

3. Tamper-proof audit trail. Every action your agent takes is logged with a ULID timestamp, content hash, and cryptographic signature. You always know what happened, when, and whether the agent had authority. The audit trail can be exported as a signed bundle for compliance review.

How It Works

The integration works through OpenClaw's MCP (Model Context Protocol) support via mcporter. Vorim's MCP server exposes 17 tools that your OpenClaw agent can call natively.

First, create a free Vorim account at vorim.ai and get your API key from Settings → API Keys. The free tier includes 3 agents with full SDK access — no credit card required.

Then, setup takes two commands:

# Add Vorim as an MCP server
mcporter config add vorim --stdio "npx -y @vorim/mcp-server"

# Set your API key (from vorim.ai → Settings → API Keys)
export VORIM_API_KEY=agid_sk_live_...

First run — register your agent:

mcporter call vorim.vorim_register_agent \
  name="my-openclaw" \
  capabilities:='["browse","email","shell","calendar"]' \
  scopes:='["agent:read","agent:write","agent:execute"]'

Before sensitive actions — check permission:

mcporter call vorim.vorim_check_permission \
  agent_id="agid_..." scope="agent:execute"

After every action — log it:

mcporter call vorim.vorim_emit_event \
  agent_id="agid_..." event_type="tool_call" \
  action="send_email" result="success"

Once the Vorim skill is installed, your OpenClaw agent automatically checks permissions before destructive actions and logs everything it does. No manual instrumentation needed.

Why This Matters

OpenClaw runs locally with full system access. That's what makes it powerful. It's also what makes identity and accountability essential.

Consider what your OpenClaw agent can already do: browse the web, execute arbitrary shell commands, send messages via WhatsApp, iMessage, Slack, and Discord, manage your calendar, access your email, and interact with 50+ services. All of this happens autonomously, 24/7.

Without an identity layer, there's no way to answer basic questions: Did my agent send that email? Was it authorized to run that script? What exactly did it do while I was asleep?

With Vorim, the answers are cryptographically provable. Every action is signed, every permission is enforced, and every event is logged in a tamper-proof chain.

Trust Scores

Every OpenClaw agent registered with Vorim builds a trust score from 0 to 100 based on its behavioral history. Agents that consistently stay within their permissions, complete actions successfully, and have few denials build higher trust scores over time.

This trust score is publicly verifiable. External services can check your agent's trust score before granting access, creating a reputation layer for the agentic web.

# Check your agent's trust score
mcporter call vorim.vorim_verify_trust agent_id="agid_..."

# Returns: trust_score: 85, status: active, scopes: [...]

Get Started

The Vorim skill is available now. Get started in under a minute:

1. Create a free account at vorim.ai (no credit card required)

2. Get your API key from Settings → API Keys

3. Connect to OpenClaw:

# Add the MCP server
mcporter config add vorim --stdio "npx -y @vorim/mcp-server"

# Set your API key (from vorim.ai → Settings → API Keys)
export VORIM_API_KEY=agid_sk_live_...

# Done — your agent now has identity, permissions, and audit trails

The free tier includes 3 agents with full SDK access.

We've also submitted a PR to include Vorim as a bundled skill in OpenClaw (PR #68640). Once merged, every OpenClaw user will have access to agent identity and trust out of the box.