Connect your warehouse in 60 seconds

Contextary connects to your existing data warehouse — no data migration, no ETL pipelines, no new infrastructure. Just enter your connection details and Contextary auto-discovers every table, column, and data type.

Auto-discovers all tables and columns

Detects column types, nullable fields, and relationships

Estimates row counts for each table

Maps your schema structure automatically

BigQuery Snowflake PostgreSQL Redshift

Your data stays in your warehouse. Contextary only reads schema metadata — table names, column names, and types. It never copies or stores your actual data.

Schema Discovery Complete

production · BigQuery

accounts

24 columns

opportunities

31 columns

invoices

18 columns

subscriptions

22 columns

events

15 columns

5 tables discovered 110 columns mapped

opportunities

Sales opportunities from Salesforce. Each row is one deal. Grain: one row per opportunity_id.

stage_name VARCHAR

Current sales stage. Values: Prospect, Qualified, Proposal, Negotiation, Closed Won, Closed Lost.

gotcha Stage 0 = unqualified, exclude from pipeline totals

amount INTEGER

Deal value in US cents. Divide by 100 for dollar amounts.

gotcha Stored in cents, not dollars

close_date TIMESTAMP

Expected or actual close date. Use for fiscal quarter calculations (FY starts Feb 1).

2

Document what your data actually means

This is where the magic happens. Your team adds the context that only humans know — the business rules, the gotchas, the metric definitions that live in people's heads and Slack threads.

Table descriptions

What each table represents, its grain, update frequency, owner

Column annotations

What columns actually mean, not just their technical names. "status = 'complete' means paid AND shipped"

Metric definitions

MRR, churn, NRR, pipeline value — defined once with exact formulas so every tool uses the same number

Gotchas

Reusable data hazards: "exclude trial accounts from churn", "Stage 0 isn't sales-qualified", "amounts are in cents"

Relationships

Which tables join on what keys, cardinality, FK relationships

Business context

Company info, fiscal year, currency, naming conventions

AI-assisted annotations

Don't want to document everything manually? Contextary's AI can suggest descriptions and annotations based on your schema. Review, tweak, accept.

Think of it as a living data dictionary that your AI can actually read.

3

Every AI question gets the full picture

Once your handbook is set up, Contextary and Claude work together behind the scenes. Contextary provides context, helps Claude write accurate SQL, and tells your warehouse what query to run — all automatically. No manual prompting, no copying docs into chat.

1

You ask Claude: "What's our pipeline by stage this quarter?"

2

Claude calls Contextary to understand the relevant tables and business rules

3

Contextary returns your team's knowledge:

table descriptions column meanings join keys gotchas metric definitions

4

Contextary with Claude writes accurate SQL using the correct joins, filters, and business rules

5

Contextary tells your warehouse what query to run and returns the results

6

Claude delivers the answer — as text, a table, a chart, or a full dashboard

The difference context makes

Same question, same AI, same data. The only difference is whether AI has your team's knowledge.

Without Contextary

Guesses column meanings from names alone

Picks wrong join keys between tables

Includes test data and trial accounts

Uses wrong metric definitions

Reports amounts in cents, not dollars

With Contextary

Knows exactly which columns to use and what they mean

Uses the correct join keys every time

Applies correct business rules and exclusions

Uses your team's exact metric definitions

Converts cents to dollars, handles edge cases automatically

4

Trust but verify

Contextary includes a built-in verification tool where you can test AI-generated queries against your real data. Ask questions, see the SQL, check the results, and give feedback — all before anything hits a dashboard or board deck.

Ask questions in plain English and see the generated SQL

Run queries against your actual warehouse

Review results and flag incorrect answers

Export results as CSV

Build confidence in AI accuracy over time with query history

Every query is logged with the question, SQL, results, and your feedback — so your team can track AI accuracy over time.

Verification Console

Question

"What's our MRR for Q1?"

Generated SQL

SELECT SUM(amount) / 100 AS mrr_dollars
FROM subscriptions
WHERE status = 'active'
  AND is_trial = false
  AND period_start BETWEEN
    '2026-02-01' AND '2026-04-30'

Result

$847,200 324 active subscriptions

Context Applied

cents → dollars exclude trials FY starts Feb 1

Ready to give your AI the context it needs?

Free to get started. Set up in minutes, not sprints.

Start for Free Explore Features

How Contextary works