AI integration · MCP

AI agents work on the scenario. The operator decides.

CityTac includes a built-in MCP server: the open standard that lets any compatible AI agent read and manage tactical data, within the permissions set by the agency.

01What MCP is

An open standard to connect AI to data.

The Model Context Protocol (MCP) is an open standard that defines how an AI agent connects to external tools and data sources. Instead of custom integrations built for each model, a single common protocol.

For CityTac, this means the tactical scenario is accessible to any compatible agent, without being tied to a single artificial intelligence vendor.

  • Open standard
  • No vendor lock-in
  • Per-tool permissions
  • Logged actions
  • Human approval

02Exposed tools

What an agent can do with CityTac.

The MCP server exposes specific tools for each phase of the work. Each tool has its own permissions: an agent can be authorized only to read, or also to propose changes.

citytac-mcp · connected8 tools
  • get_area_layers

    Returns the geospatial layers of the selected area.

  • get_alpha_zone

    Returns the current Alpha Zone with its perimeters and security rings.

  • set_alpha_zone

    Sets or updates the Alpha Zone and its security rings.

  • add_threat

    Logs a threat with location, type and risk level.

  • list_threats

    Lists the active threats in the scenario.

  • check_coverage

    Checks which access points and threat axes are covered by the scheme.

  • propose_defense_scheme

    Generates a defense scheme for the operator to review.

  • export_briefing

    Prepares a scenario summary for the chain of command.

Agent sessionExample
  1. Operator

    The march is changing route towards piazza Garibaldi. Check the impact on the current scheme.

  2. AI agent

    Reading the new route and comparing it with the Alpha Zone and the manned access points.

  3. CityTac MCP

    get_alpha_zone → 3 rings · check_coverage → 2 uncovered access points

  4. AI agent

    Two access points on the east side remain uncovered. I propose moving two posts from the outer ring and opening an alternative egress route.

  5. CityTac MCP

    propose_defense_scheme → awaiting approval

  6. Operator

    I approve moving the posts. I'll assess the egress route with command.

03Uses of AI

Where an AI agent makes a difference.

  1. 01

    Scenario analysis

    • Detects uncovered access points
    • Highlights inconsistencies between rings and posts
    • Flags blocked egress routes
    • Compares the scenario with logged threats
  2. 02

    Defense proposals

    • Prepares alternative schemes
    • Suggests repositioning posts
    • Assesses coverage
    • Explains every proposal
  3. 03

    Monitoring developments

    • Updates the picture when the situation changes
    • Recalculates critical issues
    • Alerts the operator to relevant changes
    • Keeps track of changes
  4. 04

    Briefing and communication

    • Summaries for the chain of command
    • Shift handover summaries
    • Debriefing reports
    • Clear, verifiable language

04Principles

  1. 01

    Artificial intelligence proposes. The operator decides.

  2. 02

    Every agent action is visible, explained and logged.

  3. 03

    An agent sees only what it is authorized to see.

05AI governance

Human control at every step.

  • Per-tool permissions

    Each agent is enabled only for the tools it needs: read, propose or none.

  • Mandatory approval

    Proposals remain pending until an operator approves or rejects them.

  • Complete log

    Every call to the MCP tools is logged with author, time and outcome.

  • Vendor independence

    MCP is an open standard: the agency chooses the most suitable agent and model.

Security and governance

06Frequently asked questions

The answers, in brief.

01

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that defines how AI agents connect to external tools and data. CityTac implements it with a built-in server, so the tactical scenario is accessible to any compatible agent.

02

Which AI agents can connect to CityTac?

Any agent or client that supports MCP can connect to CityTac, within the permissions assigned by the agency. The choice of model and vendor remains with the organization.

03

Can an AI agent change the scenario on its own?

No. In CityTac, an AI agent can read and propose; proposals become operational only after an authorized operator approves them.

04

How do you control what an AI agent does?

In CityTac, each agent has permissions defined per tool and every action it takes is logged: it is always possible to reconstruct who proposed what, and when.

Request a demo

See an AI agent at work on CityTac.

In the demo we connect an AI agent via MCP to a realistic scenario.

Request a demoSecurity and governance