AI agents

Built with agents.
Built for agents.

Two different readers, one page. Left: you or your AI agent integrates Ordent. Right: Ordent tells you who is on the other end.

Build with agents

Install from your editor in one prompt.

Ordent ships an MCP server, an llms.txt, rules files for Cursor and Claude Code, and typed SDKs. Paste one line, describe your app, and your agent wires the events, the webhook and the widget.

install
npx ordent init --agent
 
# or, in Cursor / Claude Code:
"Add Ordent risk tags to signup and withdrawal, Recommend mode, crypto preset."

Agents that run your business can ask. us

An agent that handles support, approves refunds or manages a storefront can call Ordent at runtime with a scoped key: check_user, explain_decision, report_outcome. Read-only by default, no policy changes from an agent.

runtime · MCP
// an agent that approves refunds asks before it acts
const r = await mcp.call("ordent.check_user", { user: "usr_0f3a" });
if (r.decision !== "allow") return escalate(r.explain);
Decide about agents

Human, bot, or agent., on every request

Every event carries actor.human, actor.bot or actor.agent. For agents, whether the request is signed under an agent protocol and whether we can attribute it to a human principal.

response · tags
"tags": [
  { "tag": "actor.agent", "confidence": 0.96 },
  { "tag": "agent.verified", "confidence": 0.99,
    "action": "apply_agent_policy" },
  { "tag": "agent.principal_known", "confidence": 0.91 }
],
"decision": "allow"  // verified · within spend limit

Decide what each agent may do.

Agent policies live next to your user policies: allowlist known agents, set spend and velocity limits per agent, step up the principal instead of the agent, and block unsigned automation where it matters. The weekly report shows agent traffic as its own line.

agent policy
# agent policy (excerpt)
actor.agent + !agent.verified      → block withdrawal
agent.verified + spend_24h > 2000  → step_up principal
agent.verified + principal_unknown → review

Detection of AI agents is a signal many vendors offer. What comes after it, attribution and policy, is the product.

Start today

Your risk team starts here.

$99 a month, 10,000 events, sandbox before you pay, money back in 14 days if it doesn't fit.

Start for $99Install with your AI agent