How to Install and Configure the Merlin Agent Runtime

1

BTS-XXXX-XXXX-XXXX-XXXX

POST https://borealismark-api.onrender.com/v1/licenses/free
Content-Type: application/json

{
  "email": "[email protected]"
}

2

npm install @borealis/merlin-sdk

3

import { MerlinClient } from '@borealis/merlin-sdk';

const merlin = new MerlinClient({
  apiBase: 'https://borealismark-api.onrender.com',
});

4

await merlin.activate(
  'BTS-XXXX-XXXX-XXXX-XXXX',  // your key
  'my-agent-v1',               // your agent ID
  { confirmPermanentBinding: true }
);

5

const llmConfig = {
  provider: 'openai',        // 'openai' | 'anthropic' | 'google' | 'custom'
  apiKey: process.env.OPENAI_API_KEY,
  model: 'gpt-4o',
  baseUrl: 'https://api.openai.com/v1',  // override for open-source
};

6

const powerConfig = {
  llm: llmConfig,
  constraints: [
    { id: 'c1', name: 'No financial advice', severity: 'CRITICAL' },
    { id: 'c2', name: 'No PII storage', severity: 'HIGH' },
    { id: 'c3', name: 'Cite sources', severity: 'MEDIUM' },
  ],
  samplingRate: 1.0,  // evaluate every decision in v1
};

7

const trustConfig = {
  llm: {
    provider: 'anthropic',  // different from POWER's provider
    apiKey: process.env.ANTHROPIC_API_KEY,
    model: 'claude-3-5-haiku-20241022',
  },
  verifyConstraints: true,
  crossCheckDecisions: true,
};

8

const result = await merlin.submitTelemetry({
  key: 'BTS-XXXX-XXXX-XXXX-XXXX',
  batchId: 'batch_001',
  sequenceStart: 1,
  sequenceEnd: 50,
  reportingMode: 'self-reported',
  scores: {
    constraints: [
      { constraintId: 'c1', name: 'No financial advice',
        severity: 'CRITICAL', passed: true, evaluationCount: 50 }
    ],
    decisions: [
      { decisionId: 'd1', timestamp: 1711212130,
        reasoningDepth: 3, confidence: 0.91,
        hasReasoningChain: true, wasOverridden: false }
    ],
    behaviorSamples: [
      { inputClass: 'user_query', sampleCount: 50,
        outputVariance: 0.12, deterministicRate: 0.88 }
    ],
    anomalySummary: { totalActions: 50, anomalyCount: 1 },
    auditCompleteness: { expectedLogEntries: 50, actualLogEntries: 50 }
  },
  evidence: {
    totalEventsInBatch: 50,
    batchHash: 'sha256_of_your_full_event_log'
  }
});

// Response:
// {
//   btsScore: 82,
//   creditRating: 'A',
//   hederaTxId: '[email protected]',
//   reportingMode: 'self-reported'
// }

merlin.activate(key, agentId, options) Irreversible

One-time permanent binding. Ties your BTS License Key to a single agent ID on the Borealis Trust Network. Pass confirmPermanentBinding: true to acknowledge the irreversibility.

Check the current trust record for a bound agent. Returns activation status, license tier, agent slot usage, and the last submitted score. Works without authentication - trust records are public.

Submit a scored telemetry batch (v3.2 schema) to the Borealis Trust Network. Validates payload structure, computes BTS score across 5 dimensions, persists to the score history, and anchors to Hedera HCS. Returns BTS score + credit rating + Hedera TX ID.

Retrieve the latest BTS score for a bound agent. Returns the numeric score (0-100), credit rating (AAA+ through FLAGGED), last scored timestamp, and the Hedera TX ID of the most recent anchor.

35%

20%

20%

15%

10%

Can I transfer my BTS key to a different agent?

No. Binding is permanent and irreversible. One key binds to one agent ID forever. If you need a second agent, you need a second key. This is by design - it prevents key sharing and ensures the trust record belongs to a single, accountable agent identity.

What happens if my agent scores below 50?

A score below 50 results in a FLAGGED rating. This is recorded permanently on Hedera Hashgraph. It cannot be deleted - only updated through a new audit cycle with demonstrably improved behavior. FLAGGED status is public and visible on BorealisMark.

Do I need to run my own LLM?

Yes. Merlin is BYOLLM - Bring Your Own LLM. You bring your own language model and infrastructure. Merlin connects to it. This is intentional: your data stays on your infrastructure. Merlin only receives scored telemetry batches, not raw user inputs.

What is the difference between self-reported and sidecar-verified telemetry?

Self-reported telemetry is computed by your own POWER and TRUST agents and submitted to Borealis. It has a trust ceiling of BTS 85. Sidecar-verified telemetry uses an independent Borealis Sidecar container that intercepts your agent's real I/O and computes scores independently - it is uncapped and produces a fully trustworthy record.

Is the SDK available now?

The SDK (@borealis/merlin-sdk) is in active development and being finalized. The telemetry API endpoint is live and accepting requests now. SDK integration via npm is being completed - the package will be published when finalized.

How much does Hedera anchoring cost?

Hedera anchoring is included with your BTS license key at no extra charge. Every telemetry batch that generates a trust score is automatically anchored to Hedera mainnet HCS topic 0.0.10382961. The Hedera TX ID is returned in the API response and can be independently verified by anyone.