You are not the report mailroom.

DRE is an open-source reporting engine. It turns every recurring report into SQL, Jinja and YAML in git, then builds the Excel pack, the bank file, a PDF for every member or the Slack number and delivers it on schedule. From your stack, without you.

pip install dre-cli
Build your first report Watch one run

Moving SAS reports? Keep your database. Explore the example. Released

Open source, GPL-3.0 · Runs on your infrastructure · Works with your scheduler

Requests
  • Can you resend last month’s pack?Finance
  • The bank needs the payments file by six.Treasury
  • Same report, but one per client?Sales ops
FinanceEmail · 1st · 07:00BankSFTP · Weekdays · 17:30#revenueSlack · Daily · 08:00Client AEmail · Set A · 2nd

reports/finance/finance.yml

output:
  format: xlsx
  destination:
    profile: mail_local
    to: finance@example.com
$ dre runOne YAML per report
MembersEmail · Released · 09:00PlatformS3 · Daily · 06:00LakehouseVolume · 1st · 07:00
Empty until
Mon 07:00Next runScheduled
One illustrative project built from DRE’s demo reports: each slot opens the real SQL, YAML and output.

Move your SAS reporting workflows to DRE. Keep your database.

A synthetic overdue-loan pack, from source inputs to checked Excel outputs. Oracle and SQL Server SQL. A generated workbook and an Excel-authored template. Released with DRE 0.10.0.

Watch the workflow and inspect the exampleDownload the project

One report definition, database SQL and two checked Excel workbook approaches. Watch the real run and inspect the code.

Return the old mailroom to sender.

Recurring reports quietly turn data engineers into couriers. DRE takes the courier work and leaves you the definitions.

  1. send_finance_v3_FINAL.py

    A script per report, openpyxl formatting by hand.

    A report is SQL for the data and YAML for the tabs, formats, totals and template.

  2. cron on someone’s VM

    Nobody knows it runs until it doesn’t.

    dre run from the scheduler you already have: cron, Airflow, Dagster or Databricks Jobs.

  3. “Who changed this number?”

    Logic lives in a workbook or a laptop.

    Every change is a reviewed commit, with history, blame and a way back.

  4. Copy, paste, attach, send

    The last mile is a person.

    Email, Slack, SFTP, S3, GCS, Azure Blob and Databricks Volumes are declared in YAML.

Open any slot. Read the code behind it.

The definition on the left, what the recipient gets on the right. Real configuration from DRE’s tested examples.

The definitionSQL + Jinja + YAML
reports/finance/finance.yml · output excerpt
output:
  - name: finance
    format: xlsx
    columns:
      revenue: {total: sum}
    style:
      header: {fill: "#176B59", font_color: "#FFFFFF"}
      banded_rows: "#F0F5F3"
    destination: {profile: local, path: "out/finance-{{ var('month') }}.xlsx"}
Inspect the SQL and full configuration
reports/finance/detail.sql
select s.*, a.account_code, revenue * 0.1 as commission
from {{ source('finance', 'sales') }} s
join {{ ref('accounts') }} a using (category)
where month = '{{ var("month") }}'
order by region, category;
reports/finance/finance.yml
queries:
  - {query: setup, tab: false}
  - query: detail
    tab_name: Detail
    columns:
      revenue: {format: "#,##0.00"}
      commission: {formula: "=ROUND({revenue}*0.1,2)", format: "#,##0.00", total: sum}
  - query: summary
    tab_name: Summary
    columns:
      revenue: {format: "#,##0.00"}
output:
  - name: finance
    format: xlsx
    columns:
      revenue: {total: sum}
    style:
      header: {fill: "#176B59", font_color: "#FFFFFF"}
      banded_rows: "#F0F5F3"
    destination: {profile: local, path: "out/finance-{{ var('month') }}.xlsx"}
  - name: presentation
    format: xlsx
    queries: [summary]
    template:
      file: templates/finance.xlsx
      bindings:
        - {sheet: Summary, cell: A1, value: "Acme Corp - {{ var('month') }}"}
        - {sheet: Summary, query: summary, anchor: A4, header: true}
    destination: {profile: local, path: "out/presentation-{{ var('month') }}.xlsx"}
The resultExcel workbook
finance-2026-01.xlsxXLSX
Monthly finance, January 2026
Total revenue3,500.00
Commission350.00
Illustrative summary of the finance demo workbook
RegionRevenue
North2,000.00
South1,500.00
Total3,500.00
DetailSummary

Illustrative preview using the tested finance demo’s data. Download the actual workbook below.

See the finance reportDownload the workbook

One command does the rounds.

  1. Define the report

    SQL selects the data. YAML describes the tabs, columns, formulas, templates and destinations. Review them together.

  2. Run it from your stack

    dre run builds and delivers. Call it from cron, Airflow, Dagster, Databricks Jobs or any orchestrator.

  3. Repeat without rewriting

    Variables for the period, sets for each client or region, machine-readable results for automation.

Built your tables with dbt?

dbt transforms data in the warehouse. DRE turns it into the files and messages people receive, with familiar SQL, Jinja and YAML, reusable macros, ref() and git habits.

DRE and dbt, side by side
Recorded from the tested finance example: Excel detail, commission formulas, totals and a filled template.

Recorded from the tested finance example: Excel detail, commission formulas, totals and a filled template.

Inspect the code and download the workbook
Recorded from a real run: 400 statements made and held, then released to the test inbox with a live progress bar.

Recorded from a real run: 400 statements made and held, then released to the test inbox with a live progress bar.

Inspect the code and the runDownload one member’s statement (PDF)

400 statements. Not one sent unchecked.

One run builds a branded PDF for every member, from an HTML template, and an email that carries only that member’s statement. Then it holds them. A checker looks over the run and releases it, and DRE sends exactly what was rendered, at a pace your mail server can take.

  • for_eachOne PDF and one email per row. The queries run once; each member gets only their own rows.
  • hold: trueNothing leaves the run folder until someone runs dre release. --preview renders three members first and sends nothing.
  • rate: 50/secondSends are spaced evenly, with a live progress bar. A failed row doesn’t stop the rest, and --retry-failed redoes only those.

Mirage Super is a fictional fund: every name and amount is made up.

Any warehouse in. Any inbox out.

Sources, formats and destinations are independent plugins. One report can read several sources and deliver to several places in one run.

Read from

  • ClickHouse
  • Databricks
  • DuckDB
  • JDBC
  • PostgreSQL
  • BigQueryalpha
  • Snowflakealpha

Build as

  • CSV
  • Delimited text
  • HTML
  • PDF
  • Fixed-width
  • Parquet
  • Excel
  • Headline messages

Deliver to

  • Databricks Volumes
  • Email
  • FTP / FTPS
  • Amazon S3
  • Google Cloud Storage
  • Azure Blob
  • SFTP
  • Slack
  • Google Chatpreview
  • Microsoft Teamspreview
  • Local files

Browse the plugin registry

Let an agent write the report. Review it like code.

Reports are plain text, so coding agents can help. DRE’s skills take an agent from “help me with dre” to a delivered report, one question at a time, and the SQL and YAML stay in your project.

terminal
npx skills add get-dre/dre#skills-latest
Explore agent skills

Give AI answers you can trace MCP · RC preview

DRE MCP is a separate companion. It brings reviewed reports and documented tables to AI, and marks what is verified, derived or exploratory.

MCP preview: reviewed report context, derived answers and exploratory SQL with clear provenance.

MCP preview: reviewed report context, derived answers and exploratory SQL with clear provenance.

Read the MCP documentationRequest access

MCP has its own early-access licence. DRE core stays open source.

Know what you’re adopting.

DRE is under active development, before 1.0. Check its capabilities and limits before it joins your stack.

What is DRE?

DRE (Declarative Reporting Engine) is a free, open-source command-line tool for reports as code. You write each report as SQL, Jinja and YAML files in git; DRE runs the SQL against your database, builds the file (an Excel workbook, CSV, fixed-width, Parquet, HTML or PDF) or a headline message, and delivers it by email, SFTP, cloud storage or Slack. What reports as code means.

Does DRE replace dbt?

No. dbt transforms data inside your warehouse and builds tables. DRE starts where dbt ends: it reads those tables and turns them into the files people and systems receive. Many teams will use both.

Does DRE replace my BI tool?

No. Dashboards stay in your BI tool. DRE is for recurring reports and files that have to arrive somewhere in an exact shape: a monthly finance workbook, a fixed-width file for a bank, an extract for another system. How DRE compares with the usual ways.

Does DRE schedule reports?

DRE is not a scheduler: it runs from the one you already have, such as cron, Airflow, Dagster, Databricks Jobs or GitHub Actions. You can name schedules in YAML so the orchestrator only says which one to run, and dre schedule ls shows every upcoming run.

Is DRE free?

Yes. DRE is open source under GPL-3.0, free for any use, commercial use included. A commercial licence is only needed to embed DRE in a product you distribute without the GPL’s obligations. Licensing FAQ.

What is DRE MCP?

DRE MCP is a separate companion in release-candidate preview. It brings reviewed reports and documented tables to AI, distinguishing verified report outputs, derived answers from saved results, and unverified exploratory queries with SQL to review. Access is available by request, under its own early-access EULA; DRE core remains open source under GPL-3.0. MCP demo and licensing details.

Close the mailroom.

DRE core is free and open source under GPL-3.0. Run it on your infrastructure, keep the definitions in git.

Build your first reportView the source on GitHub
Licensing and commercial use
pip install dre-cli

Python · macOS · Linux · Windows · All installation methods