Your Entire Business. One Conversation.

YC MCP Apps Hackathon | Feb 21, 2026 | Y Combinator, SF

We replaced 8 SaaS tools with 7 natural-language prompts. In 90 seconds, a founder goes from "How's my business?" to closing a $150K deal — with real pipeline data, real intent signals, and a pricing simulator that tells you whether to discount or hold firm. All inside Claude.

This is not a hackathon prototype. This is 2 years of production infrastructure — 41 customers, $15.6K MRR, 82% gross margins — exposed through MCP in a single weekend.

Why This Matters

The $1T+ SaaS market is collapsing. HubSpot down 70%. ZoomInfo down 49%. Five CEO departures in 2025. Every founder still stitches together 8-12 disconnected tools that don't talk to each other.

Meanwhile, 800M people use ChatGPT weekly. Claude is where developers live. Both have zero business context — no pipeline, no customers, no revenue.

We fixed that. 32 MCP tools. 8 business domains. 12 interactive scenes. One conversation.

The founder asks a question. The AI workspace answers with live data, interactive charts, and actionable recommendations — not generic advice, but "Reema Batta at Figma has intent score 87/100, she visited your pricing page 6 times this week, here are your talking points for the call, and here's what happens to your runway if you offer 20% off."

The Demo (7 prompts, 90 seconds)

Every scene links to every other scene. The slider talks back to the server. The Go buttons trigger real tool calls.

What Makes This Different

Bidirectional UI inside the AI workspace. The pricing simulator slider doesn't just display data — it calls app.callServerTool() to re-run financial models on every drag. The server responds, charts update, the verdict changes. This is not a static embed. This is a reactive application living inside Claude.

Real business, real data. $875K pipeline across 8 active deals. Person-level intent signals with behavioral scoring. Financial models that calculate close probability, expected value curves, and LTV impact. Not mock data for a demo — production APIs serving a live business.

32 tools, zero new infrastructure. Every tool is a thin MCP wrapper around existing Markster Panel REST APIs that serve 41 customers today. The hackathon didn't build a product. It proved a distribution model.

Architecture

+-----------------------------------------------------------+
|                   Claude / ChatGPT                         |
|                   (AI Workspace)                           |
+--------------------------+--------------------------------+
                           | MCP Protocol
+--------------------------+--------------------------------+
|              Markster MCP Server                           |
|  Express 5 + StreamableHTTPServerTransport                 |
|  32 tools + UI resource (single-file HTML bundle)          |
+------------------------------------------------------------+
|  +-----------+ +-----------+ +-----------+ +-----------+   |
|  | Customer  | | Content   | | Outreach  | |  Tasks    |   |
|  | Workspace | | Calendar  | | Pipeline  | | AI Team   |   |
|  +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+  |
|        |              |             |              |        |
|  +-----+--------------+-------------+--------------+----+  |
|  |           Markster Panel REST API                     |  |
|  |           panel.markster.io (production)              |  |
|  +-------------------------------------------------------+  |
|                                                             |
|  +-----------+ +-----------+ +-----------+ +-----------+   |
|  | Apollo    | | Impact    | | Pricing   | | Signals   |   |
|  | Enrich    | | Engine    | | Sim       | | Engine    |   |
|  +-----------+ +-----------+ +-----------+ +-----------+   |
|  Custom tools: live enrichment + financial modeling         |
+------------------------------------------------------------+
|  Interactive UI (Vite + vite-plugin-singlefile)            |
|  Chart.js | Bidirectional callServerTool() | 12 scenes     |
+------------------------------------------------------------+

12 Interactive Scenes

32 Tool Inventory

Customer and CRM (8 tools)

Tasks and AI Team (5 tools)

Content and Social (4 tools)

Outreach and Prospecting (4 tools)

Knowledge and Artifacts (3 tools)

Email Infrastructure (3 tools)

Voice and Brand (1 tool)

Deal Intelligence (4 custom tools)

Tech Stack

Sponsor Technology Integration

Quick Start

npm install
npm run build
PORT=3002 npm run start

Verify:

curl -s http://127.0.0.1:3002/
curl -sS -H 'Accept: application/json, text/event-stream' \
  -H 'content-type: application/json' \
  -X POST http://127.0.0.1:3002/mcp \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Environment Variables

Deploy

npm run deploy   # One-command deploy to Manufact Cloud

MCP Validation CLI

npm run mcp:test -- check --endpoint http://127.0.0.1:3002/mcp
npm run mcp:test -- list-tools --endpoint http://127.0.0.1:3002/mcp
npm run mcp:test -- call --tool sales_intelligence \
  --args '{"mode":"research","company_name":"Anthropic"}' \
  --endpoint http://127.0.0.1:3002/mcp

Numbers

Project Structure

.
├── server.ts                 # MCP server setup + tool registration
├── main.ts                   # Express app, auth, logo proxy, MCP routes
├── src/
│   ├── tools/
│   │   ├── markster-business-os.ts   # 28 Markster tool executors + registration
│   │   ├── daily-priorities.ts       # Action plan tool
│   │   ├── simulate-pricing.ts       # Pricing simulator
│   │   ├── deal-impact.ts            # Financial deal modeling
│   │   ├── prep-call.ts              # Call prep briefing
│   │   └── get-signals.ts            # Person-level signals
│   ├── services/
│   │   ├── impactEngine.ts           # Close probability, EV curves, LTV model
│   │   └── mockData.ts               # Pre-seeded company data, deal defaults
│   ├── ui/
│   │   └── app.ts                    # 4500+ line interactive UI (12 scenes)
│   ├── shared/
│   │   ├── marksterToolManifest.ts   # Tool catalog + schema definitions
│   │   └── markster-brand-pack.ts    # Visual system tokens
│   ├── markster/
│   │   └── client.ts                 # Markster Panel API client
│   └── types/
│       └── contracts.ts              # Zod schemas + TypeScript types
├── data/
│   ├── financials.json               # Puzzle-style financial baseline
│   ├── mock-signals.json             # Person-level intent signals
│   └── preseeded/                    # Apollo company cache (10 companies)
├── resources/
│   └── app.html                      # UI entry point (Vite builds to dist/)
└── docs/
    ├── demo-storyboard.md            # Demo script + Q&A playbook
    ├── design.md                     # Full architecture + positioning
    ├── ecosystem-poster.html         # Visual ecosystem map
    └── demo-page.html                # One-page project overview

Team

Ivan Ivanka (CEO) — on-site at YC SF. Pitch, demo, strategy. Attila Sukosd (CTO) — remote from Copenhagen. Engineering, MCP wiring, deploy.

Zero employees. Both founders code daily. First client grew 9.2x in 15 months.

markster.ai

28 production tools. Your entire business operating system. One conversation. We showed you one flow — try the rest.