platform / mcp

Your semantic data layer, in every LLM.

The Reflective Data MCP server gives any MCP-compatible AI tool access to your governed data: the same definitions as your dashboards, wherever your team works.

request-demo see-the-platform

claude — mcp: reflective-data

how did paid search revenue trend last quarter?

using reflective-data · revenue_daily
✔ metric: revenue (completed orders)
✔ grouped by channel, by week
paid_search  +24% vs previous quarter

// illustrative

trusted by data teams at

camino/financial

✓ Enterprise-ready. Already trusted by multiple Fortune 100 companies.

see it

One governed layer behind every AI tool

Connect once and every MCP-compatible tool gets the same trusted definitions, whether it is a chat assistant, a Slack bot or a scheduled report.

mcp — one layer, every tool

ClaudeChatGPTGeminiSlack botYour own agentsScheduled emails

any MCP-compatible tool

reflective data

MCP server

one connection, governed access

your definitions

Semantic layer

dbt models, metrics, dimensions

your data

Your warehouse

reporting tables, compact and fast

slack — #marketing

What was revenue by channel last week?

via reflective-data MCP · semantic layer

Revenue last week was €1.42M, with paid search the largest channel. Metric: revenue (completed orders). Same definition as your dashboards.

weekly-email.md

Weekly stakeholder email

Scheduled run asks the MCP for the previous week’s metrics
Numbers match the dashboards, because they share definitions
Written and sent without anyone assembling it by hand

// illustrative

what you get

MCP server, in detail

any-llm.md

Works with any LLM

Connect Claude, ChatGPT, Gemini or your own agents. Any tool that supports MCP can use the same data layer.

same-data.md

The same governed data

Answers come from your semantic layer, so they match your dashboards, your Explore charts and the in-platform agent.

work-where.md

Work where you work

Use your data in an AI assistant, in Slack, in your editor or inside internal tools, without opening another app.

no-raw.md

No raw API wiring

You do not connect every source platform to every AI tool. You expose one governed layer once.

automation.md

Powers automation

Use it to drive scheduled reports and summaries, such as weekly stakeholder emails written from live data.

one-integration.md

One integration, many tools

New AI tools and models inherit the same definitions the moment they connect.

how it works

From setup to answers

01 /

Build the layer

Your dbt models and semantic layer define the metrics and dimensions. We can build it with you.

02 /

Connect your tools

Add the MCP server to your AI assistant, Slack bot or internal agent. We help you set it up.

03 /

Ask anywhere

Your people and agents ask questions in the tools they already use.

04 /

Get governed answers

Every answer uses the same definitions as the rest of the platform.

in practice

How Barbora uses it

Barbora, one of the largest e-commerce businesses in the Baltics, uses the Reflective Data MCP server in Slack and directly in Claude. It also powers their weekly stakeholder emails. Because everything reads from one semantic layer, the answer in Slack, the number in a dashboard and the figure in the email all match.

// read how we built it: why a semantic layer is essential for AI

faq

Frequently asked questions

What is MCP?

The Model Context Protocol is an open standard that lets AI tools connect to data and other tools in a consistent way. If an LLM or assistant supports MCP, it can use the Reflective Data MCP server.

Which LLMs does it work with?

Any that support MCP, including Claude, ChatGPT and Gemini, as well as your own agents and internal tools.

What data does it expose?

Your semantic data layer: the dbt models, metrics and dimensions you have built. It does not hand AI tools raw source data or direct connections to your source platforms.

Why not connect each source platform directly?

Every platform defines users, sessions and revenue differently, and each has its own limits. Going through one governed layer gives consistent answers and one place to control access.

Is there documentation yet?

Not yet. The page is here so you know it exists. Contact us and we will set it up with you.

Do I need the full platform?

You need a semantic layer for the MCP server to expose. If you do not have one, we can build it, or you can use the hosted dbt in the platform.