All work

Case study · 02 of 03

Pulse

AI marketing coordinator. Edge-native MCP server with 18 tools pulling from 26 integrations. Sub-100ms globally.

Site

askpulse.io

Year

2024–2025

Role

Solo · everything

Status

Live

Overview

Pulse is a marketing intelligence layer. It connects to the 15–30 tools a typical marketing team uses (Stripe, Klaviyo, Google Ads, Meta Ads, SimilarWeb, GA4) and gives you one coordinated view of revenue, ads, email, social, SEO, and support.

Under the hood it's an MCP server. Each tool integration is a tool an AI agent can call. The dashboard is just a thin layer on top.

The problem

Marketing teams live across 15–30 tools. No unified view. The weekly status report takes hours to compile by hand. By the time you have the numbers, the week is over.

Existing aggregators (Looker, Funnel, Supermetrics) target enterprise data teams. They're slow to set up and expensive to run. They also don't speak AI.

The approach

Edge-native from day one

Built on Cloudflare Workers. Sub-100ms response from any region. No cold starts. No regional configuration to think about.

MCP-first architecture

Each tool integration is an MCP tool an AI agent can call. The dashboard wraps the same primitives. AI and humans hit the same endpoints, with the same auth.

Composio for integration plumbing

26 toolkits via Composio. Lets the build focus on the intelligence layer instead of OAuth flows and pagination patterns.

Multi-client by default

Agencies and in-house marketers both need to manage multiple brands. The data model is multi-tenant from the foundation, not bolted on later.

Stack

CLOUDFLARE WORKERS TURSO (libSQL) NEXT.JS VERCEL STRIPE COMPOSIO MCP SERVER 18 TOOLS 26 TOOLKITS

Results

18

MCP tools in production

26

Toolkit integrations

<100ms

Global response time

Next case

Ezra