Gatlas

Connect · Model · Govern · Ask

AI-native data infrastructure for entire portfolios.

Connect every entity's tools, model them into a documented semantic layer, and let people and AI agents query trusted answers from anywhere: Claude, MCP, SQL, and Slack.

Claude · gatlas MCP

Consolidated gross margin by entity, last quarter?

Across 4 entities, blended gross margin was 58.4% in Q1. Holding led at 71%; DTC Brand trailed at 48%.

Holding
71%
Retail Co
55%
DTC Brand
48%
Services Co
62%
grounded in · Consolidated Portfolio P&L
Bulletin № 01Filed · gatlas.ai

By the numbers

One layer. Every entity. Nothing shared you didn't mean to.

14
live connectors
1
semantic layer, every entity
0
credentials you collect
9
MCP tools, all scoped
One layer between your data and your agents.Get startedSee pricing

Connectors

Every source your portfolio runs on.

Operators connect their own tools with a setup link — no shared credentials. Gatlas normalizes each entity, then unifies it into one semantic layer.

E-commerce

  • Shopify
  • Stripe
  • Traede

Accounting

  • QuickBooks
  • e-conomic
  • Xero

Productivity

  • Google Sheets
  • Notion

Analytics

  • Google Analytics 4
  • Google Ads
  • PostHog

Database

  • PostgreSQL / Supabase

CRM

  • HubSpot
  • Attio

More coming soon · NetSuite · Sage · Exact Online · Facebook Ads · Microsoft Dynamics 365 · Salesforce · and more

The stack

One layer, built in six.

01Connectors
02Per-entity warehouse
03Semantic views
04AI context
05MCP server
06People & agents

How it works

From scattered tools to trusted answers.

  1. 01 · Connect

    Operators connect their own tools.

    Send a setup link. Each portfolio company connects Shopify, Stripe, QuickBooks, e-conomic and more — without ever sharing a credential with you.

    setup · retail-co.gatlas.ai

    Connect your tools

    3 of 4 connected · no credentials shared with gatlas

    Shopifyconnected
    Stripeconnected
    QuickBooksconnected
    Google Analytics 4Connect
  2. 02 · Model

    Model raw data into a semantic layer.

    Per-entity warehouses become documented views — P&L, margin, customers — each enriched with AI context and column docs, the way your analysts would explain them.

    view · monthly_pnl
    monthrevenuemargin
    2026-01€482k57%
    2026-02€515k59%
    2026-03€548k61%

    AI context

    Monthly P&L per entity. Revenue is net of returns; margin excludes intercompany.

    revenue · netmargin · ex-IC
  3. 03 · Govern

    Scope exactly what agents can see.

    Issue MCP keys scoped to specific entities, views, and tools. Read-only by default, every query logged. Trust without a security review.

    MCP key · Portfolio CFO
    mcp_live_••••••7a2 governed
    4 entities12 views
    query_viewallowed
    query_sqlallowed
    list_entitiesallowed
    trigger_syncblocked
  4. 04 · Ask

    Ask from anywhere, in plain language.

    Your team and your AI agents query the same governed views — over MCP, SQL, or Slack — and get trusted answers across every entity.

    Claude · gatlas MCP

    Consolidated gross margin by entity, last quarter?

    Across 4 entities, blended gross margin was 58.4% in Q1. Holding led at 71%; DTC Brand trailed at 48%.

    Holding
    71%
    Retail Co
    55%
    DTC Brand
    48%
    Services Co
    62%
    grounded in · Consolidated Portfolio P&L

What's in the box

More than a pipeline.

Delegated setup links

Send a link; each portfolio company connects its own tools. You never collect or store a credential.

Cross-entity consolidation

Roll the whole portfolio up into one set of numbers — margin, cash, customers — deduped across entities.

Per-entity isolation

Each company in its own schema, with row-level security.

Semantic views + AI context

Documented and defined for people and agents alike.

Scoped keys + audit log

Entity, view, and tool scopes. Every query logged.

Reverse ETL

Push trusted data back into the tools teams use.

Anywhere you work

Ask from anywhere. Same trusted answer.

In Claude, over MCP, in a SQL workspace, or right inside Slack — every surface reads the same governed views, so the number doesn't change with the tool.

ClaudeMCPSQL workspaceSlack

Slack · #finance

@gatlas consolidated cash position right now?

€2.4M across 4 entities. Holding holds 61%; DTC Brand is lowest at €180k.grounded in · Bank Activity

Why gatlas

Built for portfolios, and for agents.

Multi-entity
GatlasNative org → entity hierarchy, schema-isolated
The usual stackOne workspace; entities faked with table prefixes
Onboarding data
GatlasDelegated setup links — operators connect their own tools
The usual stackCollect and store everyone's credentials yourself
Semantic layer
GatlasDocumented views with AI context and column docs
The usual stackRaw tables and tribal knowledge
AI access
GatlasMCP-native; agents query governed views
The usual stackBolt an LLM onto a SQL endpoint and hope
Governance
GatlasKeys scoped to entities, views, and tools; every query logged
The usual stackAll-or-nothing database credentials
Consolidation
GatlasCross-entity rollups built in
The usual stackMonthly spreadsheet exports, by hand
First answer
GatlasMinutes, in plain language
The usual stackFile a ticket with the data team

From the field

Operators and CFOs, in their words.

I used to export spreadsheets from six accounting systems every month. Now I ask one question and the numbers reconcile to my accountant's file.
CFO, multi-entity retail group
Our portfolio companies connect their own tools in an afternoon — we never see their passwords.
Operating partner, investment firm
Claude answers across every entity now, and I can see exactly which view each number came from.
Finance lead, holding company

Point your agents at data they can trust.

Spin up a workspace, send your first setup link, and ask your portfolio a question in plain language.

14 live connectorsOne semantic layer per entityMCP-nativeRead-only by defaultEvery query loggedCross-entity consolidationAnswers in plain language
14 live connectorsOne semantic layer per entityMCP-nativeRead-only by defaultEvery query loggedCross-entity consolidationAnswers in plain language
14 live connectorsOne semantic layer per entityMCP-nativeRead-only by defaultEvery query loggedCross-entity consolidationAnswers in plain language
14 live connectorsOne semantic layer per entityMCP-nativeRead-only by defaultEvery query loggedCross-entity consolidationAnswers in plain language