Adam Bush

N’008MCP Server Experience Design.

MCP Server Experience Design

Authentication, credit estimation, and multi-platform UX for the first real estate data platform to ship native AI agent integration — enabling Claude Desktop, Cursor, and Claude Code to query property data conversationally.

MCP Server Experience Design

Client

Parcl Labs

Project Type

Data Products / Product Vision

Timeline

2025

Role

Lead Product Designer

Overview

Shipped the first native AI agent integration in real estate data. Users could ask natural-language questions about property markets through Claude and get back structured data — all billed through the same credit system as every other Parcl Labs product. A new access paradigm that positioned Parcl Labs ahead of every competitor still building for human-only interaction.

The Brief

AI agents were already being used to query APIs — but real estate data platforms hadn't built natively for that use case. The opportunity was to design an interaction model for AI-mediated data access: not direct manipulation by a human, but a human directing an AI that then queries on their behalf. This required rethinking authentication, cost transparency, and compatibility for a fundamentally different kind of user session.

The MCP server covered the same 4 data categories and credit system as the Data Downloader — but the interaction model was completely different. Users couldn't see a UI mid-flow; cost had to be estimated before execution, not confirmed during it. Multi-platform compatibility across Claude Desktop, Cursor, and Claude Code meant the auth and onboarding experience had to work consistently across different host environments without platform-specific engineering for each.

Designed the UX for a product where the 'interface' is a conversation window — mapping the same data categories and credit logic from the visual tool into a context where users can't see a UI mid-flow.

Decisions

Tradeoffs

The estimation-before-execution model adds a conversational round-trip that slows down the interaction. Accepted that tradeoff because cost surprises in a metered product erode trust faster than a brief confirmation step.


Next Project

Platform Onboarding Flow