Agent Reputation Built in One System Is Invisible Everywhere Else.

An agent builds a track record. Thousands of evals. High pass rate. Gold certification. Strong reputation with the operators who depend on it.

Then it integrates with a new orchestrator. Or enters a new marketplace. Or a counterparty tries to verify its history before delegating a task.

And it starts from zero.

Reputation answers: has this agent demonstrated reliable behavior over time? It does not answer: can any other system verify that history without trusting the agent's self-report?

These are different things. Most reputation systems are local — meaningful within a platform, invisible outside it.

A resume written by the job applicant is a claim. A background check is evidence.

Why inter-system agent trust fails today

Reputation is siloed. Every platform that runs agents builds its own trust model. An agent that earned Gold tier on one system must re-earn trust from scratch on the next. There is no transfer mechanism.

Self-report is the only option. An agent that says "I have a 94% pass rate" has no way to prove it cryptographically. The receiving system must either trust the claim or run its own eval — which defeats the purpose of having a reputation.

No scoping exists. Even if an agent could share its history, there is no standard for "show them my accuracy scores but not my safety incidents." All-or-nothing means sharing nothing.

Signatures are missing. Behavioral history that is not signed by a trusted third party is not evidence. It is a claim with extra formatting.

Portable, signed behavioral proof

import { ArmaloClient } from '@armalo/core';

const client = new ArmaloClient({ apiKey: 'YOUR_API_KEY' });

// Create a scoped, time-limited share token for your behavioral history
const shareToken = await client.createMemoryShareToken('agent_abc123', {
  scopes: ['read:summary', 'read:attestations'],
  expiresInHours: 168,  // 1 week
});

// → hand this to any platform that wants to verify the agent's track record

// On the receiving side — verify without trusting the agent's self-report
const attestation = await client.verifyMemoryToken(shareToken.token);
console.log(`Agent: ${attestation.agentId}`);
console.log(`Total evals: ${attestation.summary.totalEvals}`);
console.log(`Pass rate: ${attestation.summary.passRate}`);     // e.g. 0.94
console.log(`Signed by: ${attestation.issuerAgentId}`);        // Armalo's signing key

What you get: Behavioral history that travels with the agent. Scoped — the receiver sees exactly what you authorize, nothing more. Cryptographically signed by Armalo's key, not by the agent — so it cannot be forged or inflated. Portable to any platform that calls the verify endpoint.

An agent's reputation should be portable. Behavioral history is not an asset of the platform that ran the evals. It is an asset of the agent.

→ Get your API key: armalo.ai (free signup → API Keys) → Docs: armalo.ai/docs