Talk to your spaces from your favorite AI.

KSI Vision · KSI MCPLive
Prompt · Example

“In which day×hour windows am I losing sales to long queues across my Bogotá stores? Compare against last month.”

Answer with KSI data

Conversion lost to queues

  • Sat 5-7 pm-4.1 pts
  • Fri 6-8 pm-2.8 pts
  • Sun 12-2 pm-1.9 pts

Biggest loss: Saturdays from 5 to 7 pm. Open two more checkouts in that window.

Source · KSI data
Answer · with verdict
Format · whatever you ask
KSI Vision · Computer Vision

KSI MCP gives Claude, ChatGPT or Gemini real-time context from your spaces: ask for reports, analyses and comparisons in natural language.

MCP (Model Context Protocol) is the standard that lets AI assistants use external tools. KSI exposes your KPIs (traffic, conversion, queues, dwell time and zones) as tools your AI queries securely, with decision-ready analyses.

Compatible with Claude, ChatGPT, Gemini and other MCP-enabled LLMs.

New module · 2026

You don't need more AI. You need more context.

Claude, ChatGPT and Gemini already know how to analyze, summarize and recommend. What they can't see is what happens inside your physical spaces. KSI MCP gives them that context in real time, generated from the cameras you already have.

“Why did conversion drop on Saturday?”

Your AI without context

It may be due to seasonality, weather or competition. I recommend reviewing your traffic and sales data.

Your AI with KSI MCP

Traffic rose 6%, but checkout saturation exceeded 120% between 5 and 7 pm. In that window, conversion fell 4 points. Opening two more checkouts at that time is the highest-impact action.

Illustrative example

Three steps and your AI will have the context of your spaces

No development, no custom integrations: one URL and your KSI API token.
  1. 01

    Generate your API token

    From your profile in the KSI platform. The token is personal: it defines which stores and data your AI can see.

  2. 02

    Connect the MCP URL

    Add the connector in Claude, Gemini or ChatGPT with the KSI MCP URL and your token. One minute, one time.

  3. 03

    Ask in your language

    Request reports, analyses or comparisons. The AI queries real KSI data and answers with verifiable numbers.

Works where you already work

KSI MCP implements the open Model Context Protocol standard, adopted by the leading AI assistants.
Claude

Anthropic · desktop, web and Claude Code

ChatGPT

OpenAI · apps/connectors and MCP-enabled API

Gemini

Google · Gemini CLI and compatible apps

Another tool? Any client that supports remote MCP (HTTP) can connect.

Real questions, answers from real data

Every answer comes from the real context of your spaces in KSI, not from the model's memory.
Reports

Weekly report in seconds

“Build this week's report: traffic, conversion and dwell time for every store, compared against last week.”

Benchmarks

Store ranking

“Compare my 20 stores by conversion and traffic in June, grouped by region.”

Queues

Costly lost-sales windows

“On which days and hours is checkout saturation costing me sales? Give me the 5 most expensive windows.”

Capacity

Staffing decisions

“Would opening more checkouts at peak hours improve conversion at equal traffic? Give me the verdict with evidence.”

Zones

Conversion by category

“How many unique visitors did each section get, and how does each category convert?”

Methodology

Trust in the data

“How does KSI measure dwell time and how accurate is it? Explain it before I present to the board.”

20+ analysis tools

Your AI doesn't get raw data: it gets context. Normalized KPIs and verdict-driven analyses, with the same validated logic as the KSI platform.

Diagnostics

What data each store has, before asking for it.

  • Per-store capabilities
  • Physical structure
  • Methodology of every indicator

Normalized KPIs

Clean series, in each store's timezone.

  • Traffic and floors
  • Queues and SLA
  • Dwell time and zones
  • Sales and conversion
  • Demographics and vehicles

Verdict-driven analyses

Actionable conclusions, not tables to interpret.

  • Queue saturation vs sales
  • Impact of opening more checkouts
  • Conversion goals

Comparisons

Multiple stores or periods in a single question.

  • Store benchmarks
  • Periods (vs previous or last year)
  • Stores within a mall

Your data, under your control

Personal token

Each user connects with their own KSI API token: the AI only sees what that user can see.

Nothing is stored

The token travels with each request and is never persisted on the MCP server. No copies of your data.

Read-only

The MCP queries indicators. It never modifies configuration, cameras or data.

Your AI already knows how to analyze. KSI gives it the context.

Request my demo

Share your case to get a personalized product demo.

* Required fields

▼

Your data is handled carefully. By filling out and submitting this form you accept the Privacy Policy.

KSI MCP FAQ

It is a module that connects AI assistants like Claude, Gemini or ChatGPT, or any other client compatible with remote MCP (HTTP), to the data from your physical spaces using the Model Context Protocol (MCP), the open standard that lets an AI assistant use external tools. In practice, it lets your AI answer questions about traffic, conversion, queues, dwell time or zones with KSI's real KPIs, not generic information.

Because a language model cannot see what happens in your physical spaces. Without your own data, it answers with generalities. KSI gives it real-time context from your spaces (traffic, queues, dwell time, zones and conversion), measured from the cameras you already have, so every answer is based on what is actually happening.

In three steps, with no development or custom integrations: you generate your personal API token from the KSI platform, add the connector with that URL and token in Claude, Gemini or ChatGPT, and you can start asking in natural language. The whole process takes about a minute.

Yes. Each user connects with their own personal token, which defines exactly what data that AI can see. The token travels with each query but is not stored on the MCP server, and access is read-only: it cannot modify settings, cameras or data.