Identity infrastructure for AI agents.

AI agents are showing up everywhere: browsing sites, calling APIs, acting on behalf of users. But they have no identity, no credentials, and no accountability. We’re building the infrastructure to change that.

Get started for free How ready are you?

Before

/api/checkoutUnknown

/productsUnknown

/account/settingsUnknown

/api/searchUnknown

→

With Checkpoint

ClaudeVerified

via Sarah Chen·read-only·94

HumanVerified

GPT-4Verified

via Mike Torres·full-access·87

Custom AgentFlagged

via Lisa Park·read-only·62

How It Works

How Checkpoint protects your site

An agent shows up. You can’t tell.

Right now, AI agents browse your site using real user credentials. Your analytics counts them as users. Your server sees a browser. Your WAF sees a logged-in human. But it’s an AI agent, acting on Sarah’s behalf, using her session cookie. And you have no way to know.

Incoming Request

User-AgentMozilla/5.0 (Windows NT 10.0)

Sessionsarah.chen@company.com

IP192.168.1.42

Looks like Sarah. But is it?

Detect

Checkpoint sees what your server can’t.

One script tag. That’s all it takes. Checkpoint’s detection layer identifies AI agents in real time using signature-based analysis, not user-agent strings, not IP lists. Real behavioral signatures. You see the agent, the user who authorized it, session duration, pages visited, and a confidence score. Passive. Zero performance impact.

Agent Detected

AgentClaude (Anthropic)

Identified

Usersarah.chen@company.com

Confidence98%

Session3m 42s · 7 pages

Enforce

Your site, your rules.

Now that you can see the agents, decide what they can do. Enforce gives you a policy engine at the edge. Decisions happen before requests reach your origin. Allow trusted agents. Block bad actors. Challenge unknowns. Rate-limit scrapers. Redirect agents to dedicated endpoints. Per-path, per-agent, configured through the dashboard or API. No code changes on your end. Your application doesn’t need to know.

Enforcing: /api/checkout

Claude (Anthropic) for sarah.chen

POST /api/checkout

Identityverified

Delegationvalid until 6pm

Rate limit12/100 today

Scoperead-only

⌊checkout requires write access

Action:redirect → /api/products

Latency: 9ms

Govern

Identity that’s cryptographic, not cosmetic.

Every agent gets a DID: a cryptographic identity that works like an unfakeable agent passport. A Verifiable Credential specifies exactly what it’s allowed to do, and when that authorization expires. Every action produces a signed proof. If something goes wrong, you have a tamper-proof audit trail. And if you need to revoke access? One click. This is what trust infrastructure looks like.

Identity Verified

DIDdid:key:z6Mkf5rG...

Delegatorsarah.chen@company.com

CredentialVerifiable Credential

Permissions

Browse products

Read pricing

Process payments

Access admin

Expires2026-03-15 18:00 UTC

Reputation94/100 ★★★★☆

Audit trail3 signed proofs

One Tag. Five Minutes.

Install the Checkpoint pixel and start seeing agent traffic immediately. No SDK, no build step, no config files.

<!-- Checkpoint Pixel -->
<script>
  (function() {
    var script = document.createElement('script');
    script.src = 'https://cdn.agentshield.ai/pixel.js';
    script.setAttribute('data-project', 'your-project-id');
    document.head.appendChild(script);
  })();
</script>

Free tier. Unlimited detection. No credit card.

Get started for free

Open Standards, Not Lock-In

Checkpoint is built on MCP-I, the identity layer for the Model Context Protocol. We donated MCP-I to the Decentralized Identity Foundation because trust infrastructure should be governed by the industry's finest, not one company.

MCP-I at DIF

MCP-I extends Anthropic’s Model Context Protocol with cryptographic identity, delegation, and reputation. It adds DIDs, Verifiable Credentials, and signed proofs to every agent connection, so trust is built into the protocol, not bolted on after.

Read the spec →

KnowThat.ai

The WHOIS lookup for AI agents. Look up any registered agent’s identity, verification status, and reputation score. Think of it as a background check for AI.

Explore KnowThat.ai →

Building AI Agents? Give Them Claws.

IdentiClaw gives your agents their own cryptographic identity. No more pretending to be a browser. Deploy agents that can prove who they are, who sent them, and what they're authorized to do.

Learn about IdentiClaw

Start seeing. Then set the rules. Then build on trust.

Checkpoint grows with you. Begin with free detection, add policy enforcement when you're ready, and scale to full cryptographic governance. One platform. Open standards. Your pace.

Get started for free