SQL + AI over the data
that's hard to query.
PoutyQL is a query, AI, and export layer for systems whose native tools fall short — starting with Salesforce. Here's the problem it solves, how the engine works, and what it's built on.
See it in action
Ask in plain English or write SQL. PoutyQL generates the query, compiles it to the optimal Salesforce execution, and returns a result you can chart and export — including joins SOQL can't express.
SELECT a.Name, a.AnnualRevenue, c.Email FROM Account a JOIN Contact c ON c.Email = a.Owner_Email__c -- join SOQL can't express ORDER BY a.AnnualRevenue DESC LIMIT 5
| Name | AnnualRevenue | |
|---|---|---|
| Vertex Foods | $4.2M | dana@vertexfoods.com |
| Nimbus Group | $3.8M | priya@nimbusgroup.com |
| Sterling Media | $3.1M | omar@sterlingmedia.com |
| Quantum Cloud | $2.9M | lee@quantumcloud.com |
| Cobalt Energy | $2.4M | jo@cobaltenergy.com |
The problem we solve
Your most important data lives in systems that make it hard to ask real questions. Salesforce is the clearest example: it holds the CRM, but its query language can't express the analysis teams actually need — and the usual workarounds are expensive.
The SOQL ceiling
SOQL can't join objects that aren't linked by a lookup, has no window functions, and only limited aggregates. Real questions turn into multi-step exports or hand-written Apex.
ETL is overkill
The standard fix is piping everything into a warehouse first — cost, latency, a stale copy of your data, and a data-engineering project before you can answer a single question.
Data is fragmented
The best answers span Salesforce, product analytics, finance, and often several orgs. Stitching them together by hand is a project in itself.
How it works
From connection to chart in one place — no pipeline, no warehouse, no stale copy. Five steps, all against live data.
Connect
OAuth your Salesforce org in a click (warehouses like BigQuery next). Tokens are encrypted; queries run as the connected identity, so your org's security carries through.
Ask or write
Write real SQL in our SFQL dialect, or ask in plain English and let Claude generate the query from your live schema.
Compile
A cost-based compiler parses the query and chooses the cheapest correct way to run it against the source.
Execute
Runs against live data via the best path — and can join across multiple orgs and sources in a single query.
Explore & export
Results as a table with auto-charts, then export to Sheets, Docs, Slides, Excel, Word, PowerPoint, or a Jupyter notebook.
The design
The wedge is a compiler that turns one SQL dialect into the optimal execution for each source. It parses your query, plans the cheapest correct strategy, emits the right native query, and runs it:
Strategy-based execution
Simple filters push down to SOQL. Relationships SOQL can't express run via anonymous Apex (single- and two-phase). The heavy and cross-source work runs in an embedded DuckDB engine — the planner picks the cheapest correct path each time.
Joins SOQL can't do
Non-relationship joins, window functions, and multi-object aggregates — the queries that simply don't fit SOQL — via Apex pushdown and local execution, against live data.
Multi-org federation
Connect several Salesforce orgs, prefix tables by org, and JOIN across them. Rows are pulled into DuckDB and joined locally with row/byte guardrails and org-labeled result columns.
Security preserved
Queries run under the connected Salesforce identity, so object-, field-, and row-level security from your org is preserved. OAuth tokens live encrypted in AWS Secrets Manager and refresh automatically.
The technology
A serverless, cost-conscious stack: pay-per-use compute, no standing warehouse, and AI that runs inside our own cloud boundary.
Frontend
Identity
API & compute
Query engine
AI
Data sources
Infrastructure
Query Salesforce the way you'd query a database — joins, window functions, AI, and exports, all against live data.