Documentation

Inntris

Runtime verification and cryptographic proof for AI agent actions.

Inntris verifies agent actions before execution, signs decisions with agent identity, and produces a tamper-evident receipt for every decision.

Get Started See live verification

What Inntris Is

A policy decision point and evidence system for AI agent actions

Inntris is not observability, logging, or prompt guardrails. It is a policy decision point and evidence system for AI agent actions.

Verifies actions before execution — financial transactions, API calls, data access, and more
Signs every decision with verifiable agent identity (Ed25519)
Enforces spending limits, rate limits, and action type restrictions before an agent can act
Records tamper-evident evidence for every approval, block, and exception

Built for teams that need stronger control and clearer proof than logs, screenshots, or after-the-fact review.

What We Do

Four pillars of trust for autonomous AI systems

Cryptographic Identity

Every agent gets a unique Ed25519 keypair

Unforgeable digital signatures for every action
Nonce-based replay attack prevention
Automatic signature verification on every request

Policy Enforcement

Configurable guardrails for every agent

Daily and per-action spending limits
Rate limiting to prevent runaway agents
Action type restrictions (allow/block lists)

Tamper-Evident Audit Trail

Verifiable records of every agent decision

Append-only audit logs (immutable)
Complete payload capture with signatures
Designed for enterprise audit requirements

Blockchain Anchoring

Tamper-proof verification on Base L2

Merkle tree batching (hourly anchors)
On-chain proof verification
Independent auditability by anyone

How It Works

Simple integration, powerful protection

Register Your Agent

Create an organization and register your AI agent. Inntris generates a unique Ed25519 keypair — you keep the private key, we store the public key.

POST /admin/organizations
{
  "name": "Acme Corp",
  "contact_email": "security@acme.com"
}

POST /admin/agents
{
  "name": "PaymentBot",
  "description": "Handles customer refunds"
}

Configure Policies

Set spending limits, rate limits, and allowed action types. These guardrails protect you from runaway agents and unauthorized actions.

PATCH /admin/agents/{agent_id}
{
  "daily_limit_usd": 10000,
  "per_action_limit_usd": 1000,
  "rate_limit_per_minute": 60,
  "allowed_actions": ["financial_transaction", "email_send"]
}

Integrate with MCP

Add the Inntris MCP server to your AI agent. The inntris_guard tool automatically intercepts critical actions and verifies them before execution.

// claude_desktop_config.json
{
  "mcpServers": {
    "inntris": {
      "command": "python",
      "args": ["-m", "mcp_server.server"],
      "env": {
        "INNTRIS_API_URL": "https://api.inntris.com",
        "INNTRIS_API_KEY": "<your-api-key>",
        "INNTRIS_AGENT_ID": "<your-agent-id>"
      }
    }
  }
}

Verify & Audit

Every action is cryptographically signed, verified against policies, and logged to an immutable audit trail. Merkle roots are anchored to Base L2 hourly.

APPROVED

Action verified & logged

BLOCKED

Policy violation

RATE LIMITED

Too many requests

Verification Receipt

Every decision produces a signed receipt you can inspect, share, or verify on-chain.

{
  "verdict": "approved",
  "agent_id": "uuid",
  "action_type": "financial_transaction",
  "approval_token": "base64_signed_token",
  "trust_score": 85,
  "audit_id": "uuid",
  "timestamp": "2025-01-15T10:30:01Z",
  "anchored": true,
  "transaction_hash": "0x517853a7400bffc3446fc73711a0cee2f45c82fc1b89d37e76aa3797eb951a77",
  "root_hash": "e56891f1de39aca50725f0e36ee4b1c4fe1c50966f69a1b368e1d691c2466149"
}

Receipts are stored in the tamper-evident audit trail and Merkle roots are anchored to Base L2 hourly. Any record can be independently verified by audit ID or transaction hash.

See live verification

Our Goal

Building the trust infrastructure for the agentic future

Universal Standard

We aim to become the universal verification standard for AI agents — the "VISA network" for autonomous systems. Any agent, any platform, one trust layer.

Enterprise Ready

From startups to regulated enterprises, Inntris scales with your needs. Built for auditability. Designed for enterprise control requirements. Optional on-premise deployment available for compliance-heavy industries.

Liability Clarity

When an AI agent takes action, you need proof of what happened and whether it was authorised. Inntris gives you a signed, verifiable record for every decision — before the action runs, not reconstructed after the fact.

Agent Ecosystem

Compatible with Claude, GPT, Gemini, and any MCP-compatible agent. Native integration with the Model Context Protocol for seamless verification.

Start Here

The fastest path to verification coverage is pull request verification.

Add inntris-verify to any GitHub repo. Every agent-generated PR gets a cryptographic receipt — signed by agent identity and anchored to Base L2.

Install GitHub Action See live verification

Works with GitHub PR workflows, MCP-compatible agent systems, and custom agent stacks via direct API integration.

Explore Inntris

Get started with the platform