Ask in English. Get SQL your analysts would sign off on.
Querylight reads your schema, your dbt docs and your past queries, then writes SQL that uses the right tables, joins and metric definitions.
Readable SQL, with the reasoning left in
Every query comes back commented, so a reviewer can see why each join is there.
-- weekly active teams by plan, last 8 weeks
-- active_team: >= 3 members with an event that week (metrics.yml)
select
date_trunc('week', a.event_at) as week,
p.plan_name,
count(distinct a.team_id) as active_teams
from marts.fct_team_activity a
join marts.dim_plans p on p.plan_id = a.plan_id
where a.event_at >= current_date - interval '8 weeks'
and a.active_members >= 3
group by 1, 2
order by 1, 2;Built for data teams
Context a generic chatbot doesn't have
Schema-aware
Knows your tables, columns, keys and which models are deprecated.
Metric definitions
Uses the definitions in your semantic layer instead of inventing its own.
Read-only by design
Runs with a read-only role and never writes to your warehouse.
Cost guard
Estimates scanned bytes and warns before an expensive query runs.
Works where you query
In the browser, your editor, or the CLI.
Learns from review
Accepted fixes become examples for the next question.
From question to result, step by step
Relative time spent in each stage of a typical question.
Rollout
How teams adopt it
Connect read-only
Add a warehouse role with read access and point Querylight at your dbt project.
Connect read-only
Add a warehouse role with read access and point Querylight at your dbt project.
Analysts review
Your data team checks generated queries and corrects joins; corrections become examples.
Analysts review
Your data team checks generated queries and corrects joins; corrections become examples.
Open to the business
Product and finance ask their own questions, with analysts reviewing flagged queries.
Open to the business
Product and finance ask their own questions, with analysts reviewing flagged queries.
Fewer ad-hoc tickets
The data team spends its time on models instead of one-off pulls.
Fewer ad-hoc tickets
The data team spends its time on models instead of one-off pulls.
Pricing
Plans for every size of data team
Questions from data teams
Type a word and the list narrows in real time — no page reload, no dead ends.
Your first question takes a minute
Install the CLI, connect a read-only role and ask away.