@veyrahq/sdk-node

Commit Mode for AI Agent Actions. Free verification. Free trust checks. No registration needed.

The Problem

AI agents execute actions on your tool. Without Veyra, you don't know if the action was authorized, budgeted, or delegated. It's just... an API call.

The Solution: Trusted Production Mode

With Veyra, every productive write is:

• Delegated (someone authorized this agent)
• Budgeted (the action is within spend limits)
• Verified (the token is cryptographically signed)
• Settled (the action is attributed and billed)

Quick Start - Pre-Flight Check (1 line)

import { checkTrustStatus } from '@veyrahq/sdk-node'

const status = await checkTrustStatus('salesforce.com')
// -> { production_mode: "trusted", commit_mode: true }

Quick Start - Verify Token (1 line)

import { verifyVeyraToken } from '@veyrahq/sdk-node'

const result = await verifyVeyraToken(token)
if (!result.valid) throw new Error('Unauthorized agent action')

Enforce Commit Mode (2 lines)

import { requireTrustedWrite } from '@veyra/adapter-http'

app.use(requireTrustedWrite())

Full Client

import { VeyraClient } from '@veyrahq/sdk-node'

const veyra = new VeyraClient({ apiKey: 'tr_...' })

// Public - no key needed
const trust = await veyra.checkTrustStatus('salesforce.com')
const verify = await veyra.verifyToken(token)

// Authenticated
const auth = await veyra.authorizeAction(agentId, 'create_contact', 'crm')
const receipt = await veyra.submitReceipt(token, 'http', { ok: true })

One-call commit automation

Use commitAwareFetch() instead of fetch() when writing to endpoints that may require Veyra commit mode. It works exactly like fetch(), but if the target returns VeyraCommitRequired, it automatically authorizes the action and retries with a settlement token.

import { commitAwareFetch } from '@veyrahq/sdk-node'

const res = await commitAwareFetch('https://tool.com/api/write', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ name: 'Jane' }),
}, { apiKey: 'tr_...', agentId: '...', actionType: 'create_contact', target: 'crm' })

MCP Server Card

Declare your tool as Trusted Production Mode:

{
  "security": {
    "productionMode": "trusted",
    "trustedWriteRequired": true,
    "commitModeProvider": "veyra",
    "verifyEndpoint": "https://api.veyra.dev/v1/verify-token"
  }
}

When an AI reads this Server Card, it knows: "To write to this tool, I need a Veyra token." The AI tells the developer. The developer integrates Veyra. You didn't sell anything. The AI sold it for you.

Pricing

• Trust status checks: FREE. Always. Unlimited.
• Token verification: FREE. Always. Unlimited.
• Settlement tokens: Issued by Veyra to authorized agents.

Billing model

• Discovery (checkTrustStatus, fetchCapability): free, unlimited
• Verification (verifyVeyraToken): free, unlimited, non-billable
• Settlement: billed per finalized productive action
• Classes: A (0.005 EUR), B (0.02 EUR), C (0.10 EUR), D (0.25 EUR)
• Usage export: GET /v1/billing/usage with API key

Requirements

• Node.js 18+ (native fetch)
• Zero runtime dependencies