Track AI Agent Costs & Decisions with Webhooks + WhatsApp Alerts

Track AI Agent Costs & Decisions in Real Time—With Webhooks + WhatsApp Alerts

This n8n workflow receives webhook events from your AI agents, calculates estimated spend from token usage and pricing, stores a persistent execution ledger, and alerts you on WhatsApp when a run exceeds your budget—plus a daily digest of activity and costs.

What this workflow does

• Logs agent events via Webhook (POST): Send one event or a batch of LLM/tool/decision/error events to the workflow.
• Validates events and estimates cost: For each event, it computes cost from input/output token counts and configured model pricing, or uses an explicit USD cost (with optional tool cost).
• Appends to a persistent execution ledger: The priced events are stored in n8n workflow static data, including per-run summaries and retention limits (max entries/runs).
• Returns a JSON webhook response: The caller receives how many events were recorded and the updated run totals.
• WhatsApp budget alerts per run: If a run exceeds the configured per-run budget and an alert hasn’t been sent yet for that run, it sends a WhatsApp message.
• Generates ledger reports (GET): A GET webhook returns a JSON ledger report filtered by run ID, agent, time window, and limit.
• Daily 24-hour digest: A scheduled run builds a 24-hour cost/activity summary and sends it via WhatsApp.

Use cases

• Monitor costs for multi-step AI agents and tool calls across environments.
• Get instant WhatsApp notifications when a run’s spend crosses a per-run budget.
• Produce on-demand JSON reports for audits, debugging decisions, or performance review.
• Send a daily “what happened and how much it cost” summary to your team.

Technical details

• Nodes/blocks: if, set, code, webhook, whats app, and sticky note.
• Integrations: WhatsApp Business credentials (WhatsApp phone number ID + recipient phone number).
• Core feature: Persistent ledger stored in n8n workflow static data with retention enforcement.

Perfect for: n8n users, automation engineers, and SaaS operators who need visibility into AI agent spend, decisions, and errors.