How to Build an ROI Model for Agentic AI in Local Government

CityDesk.ai Makes Tyler Technologies Agentic

Our MCP server acts as a friendly interpreter between everyday language and Tyler Technologies’ Data & Insights platform, formerly Socrata.

At CityDesk.ai, we started with a simple observation: cities publish massive amounts of valuable data through Tyler Technologies’ Data & Insights platform (formerly Socrata), but most people don’t know how to access this treasure trove of information.

We built a bridge to connect everyone to this data.

The Accessibility Gap

Tyler Technologies has done incredible work helping hundreds of cities digitize and publish their data. These platforms host everything from restaurant inspections to crime statistics, building permits to 311 requests—a genuine public resource.

The challenge isn’t the platform—it’s that accessing this data traditionally requires:

  • Knowing which specific dataset contains the information you need

  • Understanding database structures and field names

  • Writing SQL queries or working with large spreadsheets

  • Having the technical skills to merge and analyze datasets

This creates an unfortunate gap between the data being available and being accessible to the average person who could benefit from it.

We built our MCP (Model Context Protocol) server to bridge this gap. Now you can ask questions in plain English and get answers immediately, while the sophisticated data infrastructure that Tyler provides continues working behind the scenes.

What We’ve Built

Our MCP server acts as a friendly interpreter between everyday language and Tyler Technologies’ Data & Insights platform. When you ask a question on CityDesk.ai, here’s what happens:

1. Smart Discovery

We search across all available datasets to find what’s relevant to your question. Ask about restaurants? We find health inspections. Ask about safety? We locate crime data, streetlight outages, and traffic incidents. We understand context and relationships between datasets that aren’t obvious from their titles—and help you discover datasets you might not have known existed.

2. Automatic Schema Translation

Every city structures their data differently—which makes sense given their different needs and priorities. Our MCP server reads and understands these structures on the fly. We figure out which columns contain dates, which have addresses, which use codes versus plain text. You don’t need to know that Cambridge calls it “violation_level_1” while Boston uses “risk_category”—we handle that translation seamlessly.

3. Intelligent Query Construction

We convert your natural-language question into the precise database queries that Tyler’s platform understands.
“Show me trends in bike accidents near universities” becomes a sophisticated query that:

  • Identifies university locations

  • Calculates proximity buffers

  • Filters accident data for bicycles

  • Aggregates by time period

  • Performs the geometric calculations

4. Human-Readable Results

Raw data is just the beginning. We transform query results into useful answers—charts, maps, summaries, whatever makes sense for your question. No more staring at spreadsheets trying to spot patterns.

Real Examples from Real Users

Since launching, we’ve seen people ask questions we never imagined:

Urban Planning: “Which blocks have both the highest number of building permits and the most 311 complaints?” —helped identify areas experiencing rapid change and potential growing pains

Small Business: “What’s the foot traffic like near vacant commercial properties on Mass Ave?” —combined multiple datasets to help evaluate potential locations

Journalism: “Do rodent complaints spike after major construction projects?” —correlated building permits with health department reports over time

Civic Advocacy: “How do response times for pothole repairs vary by neighborhood income?” —merged infrastructure requests with demographic data

Why We Use MCP

We chose to build on the Model Context Protocol because it’s becoming the standard for AI–data interactions. This means:

  • Any AI can use it: While we showcase it on CityDesk.ai, any AI assistant that supports MCP can connect to our server.

  • It’s extensible: As cities add new types of data or change their systems, we can adapt without breaking existing functionality.

  • It’s fast: MCP’s efficiency means we can search hundreds of datasets and return results in seconds, not minutes.

Beyond Cambridge: Any City, Any Platform

While we started with Cambridge, Massachusetts (home to over 200 datasets), our MCP server works with any city using Tyler’s Data & Insights platform. This includes:

  • Boston’s 147 datasets covering everything from lobbying to leaf collection

  • Seattle’s extensive transit and environmental data

  • Chicago’s impressive collection of over 600 datasets

  • San Francisco’s comprehensive permit and inspection records

  • Hundreds of other cities, counties, and states

Tyler Technologies has built an amazing foundation—we’re just making it conversational. The same natural-language questions work across all of them. Our system automatically adapts to each city’s unique structure and terminology.

What This Means for Cities

We see ourselves as amplifying the investment cities have already made in their data infrastructure. Cities and Tyler Technologies have done the hard work of collecting, organizing, and publishing this data. We’re helping that investment reach its full potential by making it accessible to everyone.

For City Staff: Reduce FOIA requests by enabling self-service data access for citizens.
For Journalists: Transform complex data investigations into quick, documented queries.
For Residents: Access the wealth of information your city has already made available.
For Researchers: Leverage sophisticated data without learning multiple query languages.

Technical Details (for the Curious)

Our MCP server:

  • Leverages the full power of SODA (Socrata Open Data API) across all endpoint types

  • Takes advantage of Tyler’s geographic/spatial query capabilities

  • Manages rate limiting and pagination automatically

  • Caches schema information for performance

  • Provides and maintains detailed attribution for all data sources

We’ve open-sourced the core MCP server at github.com/citydesk/opendata-mcp so other developers can build on this approach or adapt it for their needs.

What’s Next

We’re just getting started. Coming soon:

  • Cross-city comparisons: Ask how your city compares to others on any metric

  • Subscription alerts: Get notified when data matching your interests updates

  • Custom visualizations: Build dashboards from natural-language descriptions

  • API access: Use our MCP server in your own applications

Try It Now

Head to CityDesk.ai and ask:

  • “What are the most dangerous intersections for cyclists?”

  • “Which restaurants failed inspections this month?”

  • “How have building permits changed in my neighborhood over 5 years?”

  • “What are the top 311 complaints near parks?”

Or ask whatever you’ve always wondered about your city but never knew how to find out.

The Bigger Picture

We believe in democratizing access to the excellent data infrastructure that already exists. Tyler Technologies and cities have built something remarkable—a comprehensive, standardized way to publish civic data. Our MCP server is simply the next step: making that data conversational and accessible to everyone.

When anyone can interrogate civic data as easily as they search the web, we create better-informed communities, more accountable governments, and smarter cities.

*Questions? Want to connect your city’s data? Reach out at hello@citydesk.ai