Skip to main content
The Settings page at /settings is where you control the three pillars of Tavryn’s governance model: your on-chain treasury wallet, the deterministic spending policy rules that constrain the agent’s financial authority, and the notification webhook that delivers approval links and agent alerts to your team. Changes you save here take effect immediately — the next time the agent evaluates a contract, it reads these values directly from the policy engine.
The AI negotiation agent has zero ability to approve its own transactions. The Deterministic Procurement Policy Engine is implemented in pure TypeScript with no LLM involvement. When you change a threshold here, you are modifying hard arithmetic boundaries — not suggestions that an AI can reason around.

Section 1: Arc Treasury Settlement Wallet

The Arc Treasury Settlement Wallet card displays the Circle Developer-Controlled Smart Contract Account (SCA) provisioned for your organization during onboarding. This is the on-chain address from which USDC is escrowed and released when the agent settles a negotiated contract. From this card you can:
  • View your wallet address — a full Ethereum-format address on Arc Testnet (Chain ID 5042002)
  • Copy the address to your clipboard with one click
  • Open ArcScan — click the ArcScan link to view your wallet’s full transaction history and current USDC balance on the Arc Testnet block explorer at testnet.arcscan.app
  • See your recorded USDC balance — the balance Tavryn last polled from the USDC precompile (0x3600000000000000000000000000000000000000) via a direct JSON-RPC call
Network details:
If your USDC balance drops below 100, Tavryn automatically sends an alert to your configured notification webhook with a direct link to faucet.circle.com to top up. Select the Arc network in the faucet and paste your wallet address.

Section 2: Deterministic Procurement Policy Engine

The Deterministic Procurement Policy Engine section lets you configure the hard financial boundaries that govern autonomous agent behavior. Three numeric thresholds and a set of category controls define exactly how much the agent can do without asking a human.

Spending thresholds

Autonomous Ceiling

max_auto_transaction — The maximum single transaction the agent can execute without any human signature. Transactions at or below this value are auto-approved if all other policy checks pass.Default: $2,000

Min Savings Threshold

min_savings — The minimum projected annual savings required before the agent opens a negotiation at all. Deals where the potential saving falls below this figure are skipped.Default: $200

Human Approval Required

human_approval_required_above — Transactions exceeding this amount always generate a 1-tap HMAC approval link for a supervisor, even if other conditions are met autonomously.Default: $2,000
To update a threshold, edit the value in the corresponding input field and click Save Policy & Settings at the bottom of the page.

Allowed procurement categories

Toggle the category buttons to control which contract types the agent is permitted to negotiate. At least one category must remain enabled. Active categories appear highlighted. The agent will only open negotiations for contracts whose category field matches one of the enabled categories. Contracts in disabled categories are ignored during the daily scan.
Contracts imported today support the software, cloud, and contractors categories. The additional categories — infrastructure, legal, and marketing — are available in the policy engine for organizations that track those spend types and are being expanded to full contract support.

Category budgets

For each enabled category, you can set an annual spend ceiling in the Category Budgets fields. The policy engine blocks any negotiated renewal that would push total committed spend in that category above the ceiling — even if the individual transaction is within the auto-approval threshold. This gives you department-level budget enforcement at the procurement layer. For example, setting a cloud budget of $50,000 means the agent will never autonomously commit to a cloud renewal that pushes cumulative cloud spend above that figure for the period.

Section 3: Notification Webhook

The Per-Business Notification Webhook section configures where Tavryn sends real-time alerts. The agent uses this endpoint to deliver:
  • Approval requests — 1-tap HMAC links when a transaction requires supervisor sign-off
  • Waste detection alerts — notifications when the agent identifies a contract entering the 45-day renewal window with significant projected savings
  • Settlement confirmations — confirmation messages when USDC escrow is successfully released
  • Low balance warnings — alerts when your treasury USDC balance drops below 100

Configuring the webhook

Enter a valid HTTPS endpoint URL in the Webhook Endpoint URL field. Tavryn supports:
  • Slack Incoming Webhooks — https://hooks.slack.com/services/...
  • Discord Webhooks — https://discord.com/api/webhooks/...
  • Custom HTTP endpoints — any HTTPS URL that accepts a POST with a JSON body

Testing the webhook

Before saving, click Test Webhook to send a test payload to your endpoint. Tavryn will POST a sample notification message and display a success or error message inline. Fix any connection issues before saving.
If no webhook URL is configured, approval request links are only accessible from within the Tavryn dashboard. Configure a webhook to ensure supervisors receive out-of-band notifications and can act on time-sensitive approvals before the 48-hour HMAC token expiry.

Saving your settings

After making any changes across the three sections, click Save Policy & Settings at the bottom of the page. Tavryn sends a single PUT request to /api/business/[id]/settings containing the updated policy object and webhook URL. A success confirmation banner appears when the save is complete. Settings are scoped per organization. If your Tavryn instance manages multiple organizations, use the ?businessId=[id] query parameter to load and edit a specific organization’s settings.