Building a server monitoring SaaS on Cloudflare Workers — architecture, decisions, mistakes

Building a Server Monitoring SaaS on Cloudflare Workers: Architecture, Decisions, and Mistakes

TL;DR Pulse is a server monitoring SaaS created by a solo founder for Latin American small businesses. It operates entirely on Cloudflare Workers + Turso (libsql) + R2, with a Go agent installed via a single curl command. The platform costs around $5 per month for the first 100 customers and achieves sub-second alert latency at the edge.

Stack Overview:

1. Cloudflare Workers for compute

2. Turso (libsql) for multi-tenant database

3. R2 for agent binary delivery

4. Workers AI (Llama 3.3) for alert interpretation

5. Resend for magic-link email

6. Telegram Bot API for alerts

7. Stripe for billing

The core idea behind Pulse isn't the stack itself, but the tradeoffs made during its development. This article will cover five key decisions that shaped the platform, along with code snippets from the actual repository.

Problem Statement:

Small and medium businesses in Latin America struggle with the pricing models of enterprise-grade observability tools. Local e-commerce businesses with 15 servers cannot justify monitoring bills that scale unpredictably with log and metric ingestion. Free/self-hosted alternatives require operational skills that a two-person infrastructure team cannot allocate.

Pulse was created to address this problem, providing a fixed monthly price without per-metric fees and ensuring native LATAM channels, such as Telegram alerts and Spanish-language dashboards. The platform aims to achieve sub-second alert latency from threshold breach to notification while being manageable by a single founder who can keep it running indefinitely.