Agent Control Layer

Control what your
AI agents can do.

Define policies. Evaluate every tool call. Maintain a tamper-evident audit chain. Inner I Secure brings enterprise-grade governance to AI agents without the complexity.

Try the live demo → See how it works
40%
of enterprise apps will run AI agents by 2026
0
governance layers exist in most agent deployments today
<5ms
policy evaluation latency at scale
HMAC-SHA256
tamper-evident hash chain for every decision

The problem

AI agents are deployed faster than they're governed.

"We deployed an AI agent to handle customer refunds. It approved $40K in fraudulent requests before we caught it. There was no policy enforcement layer — just the agent and its judgment."
— Enterprise ops lead, anonymized

Features

Everything you need to govern AI agents.

Four integrated modules that work together to give you complete visibility and control.

Policy Tester

Instantly evaluate any tool call against your active policies. Pass in the agent ID, tool name, parameters, and get an instant allow/deny verdict with reasoning.

POST /api/secure/evaluate

Policy Editor

Define named JSON policies with allow/deny rules, conditions, and descriptions. Version them, activate/deactivate them, and see which tools they govern — all in a single UI.

CRUD at /api/secure/policy

Audit Chain

Every policy evaluation is recorded in an append-only HMAC-SHA256 hash chain. Verify the chain's integrity at any time. No tampering goes undetected — ever.

GET /api/secure/chain/verify

TypeScript SDK

Drop into your agent with one import. Full type definitions, async/await API, and a 5-minute integration guide. Works with any agent framework.

npm install @innerinet/secure

How it works

From zero to governed agent in minutes.

1

Define your policy

Write a JSON policy in the editor. Use allow/deny rules, conditions on tool names, parameter constraints, and agent IDs. Activate it in one click.

allow: ["send_email", "read_file"]
2

Integrate the SDK

Add @innerinet/secure to your agent. Call secure.evaluate() before every tool invocation. That's it — the rest is handled automatically.

const result = await secure.evaluate(call)
3

Evaluate and audit

Every call is logged to the hash chain. Inspect the chain, verify its integrity, and export the full audit trail whenever you need it — compliance, incidents, or debugging.

GET /api/secure/chain

Developer experience

Integrate in minutes, not days.

The TypeScript SDK handles everything — policy fetching, evaluation, and chain recording — so you can focus on your agent, not the governance layer.

agent.ts — TypeScript SDK integration
// Import the SDK
import { SecureClient } from '@innerinet/secure';

// Initialize with your API key
const secure = new SecureClient({
  apiKey: process.env.INNER_I_SECURE_KEY,
  endpoint: 'https://shapeos-3.polsia.app/api/secure',
});

// Before every tool call, evaluate the policy
async function runToolCall(agentId, tool, params) {
  const result = await secure.evaluate({
    agentId,
    tool,
    parameters: params,
  });

  if (!result.allowed) {
    // Policy denied — log and halt
    console.error(`Policy blocked: ${result.reason}`);
    return { error: 'blocked_by_policy', reason: result.reason };
  }

  // Proceed — the call is recorded in the audit chain
  return await executeTool(tool, params);
}

// Verify chain integrity at any time
const verification = await secure.verifyChain();
console.log(verification.valid ? 'Chain intact' : 'CHAIN TAMPERED');

Pricing

Start free. Scale as you grow.

No locked-in commitments. Upgrade when you need more evals, persistence, and enterprise features.

Free
$0 / month
  • 50 policy evaluations / month
  • 3 active policies
  • 24-hour audit chain retention
  • Policy Tester UI
  • Basic TypeScript SDK
Start free →
Enterprise
Custom pricing
  • Everything in Pro
  • Self-hosted deployment option
  • Custom policy schemas
  • SLA + dedicated support
  • Audit export (SOC 2 / GDPR reports)
Contact sales →

Ready to govern your agents?

The live demo is already running. No setup, no sign-up required.

Launch the live demo →