MCC AI

An internal AI operations console for a Medicare agency — one place that unifies call analytics, an AI assistant, code agents, automated reporting, and CRM data for the whole ops team.

What it is

Running a Medicare agency means living across a dozen tools — a call-center platform, a CRM, spreadsheets, Slack, commissions. MCC AI pulls all of it behind a single login and turns it into dashboards, an AI assistant, and reports that generate themselves.

It's built as an integration hub: everything the operations team touches in a day, unified into one console — and increasingly, automated by AI.

What it does

Real-time KPI dashboard

Live agency KPIs and charts — today, this week, this month — pulled together from the call center and CRM into one operations view.

Call-center analytics

Convoso call data turned into agent-success and daily performance reports, with campaign-level breakdowns the team actually uses.

AI business assistant

A Claude-powered chat that answers questions over the agency's data, streaming token-by-token with saved, resumable sessions.

Concurrent code agents

Launch and monitor Cursor cloud coding agents right from the console — up to four stream live at once, with pinning and per-tab state so an in-progress run never gets torn down.

Automated reporting

Scheduled background workers generate hourly and daily reports on their own and deliver them straight into Slack.

CRM & commissions

Reads the agency's CRM and tracks commissions alongside everything else, so the whole operation lives behind one login.

Architecture

Next.js console (Vercel)

Next.js 16 with Tailwind, shadcn/ui, and Recharts. Operators sign in with a single shared passcode via NextAuth — no per-user password screen — and land in an authenticated dashboard.

FastAPI + MongoDB (Railway)

A FastAPI backend with JWT-authed REST and SSE streaming for chat, organized into clean route, connector, reporting, and storage layers over MongoDB.

Typed integration layer

Each external service has its own typed connector, isolated from route logic — so Convoso, Slack, Claude, Cursor, and Google can evolve independently.

Scheduled workers

Separate Railway worker processes run cron-style jobs — hourly Convoso pulls, daily report generation, and recurring Slack reports — independent of the web API.

MCP server

An embedded Model Context Protocol server exposes the agency's data to agents and tools through a clean, typed interface.

CRM querying

A read-only, connection-pooled link into the PolicyDen CRM (Postgres) with cached queries, kept safely separate from the agency's source-of-truth database.

Connected systems

The console is an integration hub — each service plugged in through its own typed connector.

ConvosoCall-center analytics
Anthropic ClaudeAI assistant
Cursor Cloud AgentsCoding agents
SlackCommands & report delivery
Google WorkspaceSheets & Calendar
PolicyDen CRMRead-only policy data
Built with
Next.js 16TypeScriptFastAPIMongoDBPostgresAnthropic SDKConvoso APISlack SDKGoogle APIsCursor Cloud AgentsRechartsRailwayVercel

Want to know more?

MCC AI is an internal tool, so there's no public demo — but ask Iris and she can walk you through how it works, or point you to my other work.

Talk to Iris →