Your SQL results,
delivered.

Rowbird runs your queries on a schedule and delivers the results where your team already reads: a table in the email, or CSV, Excel and PDF in Slack, Telegram, Discord, a webhook or a bucket. Add a condition and it becomes an alert.

$docker run -p 8080:8080 ghcr.io/rowbird/rowbird

Watch the 21-second tour Try the demo Star on GitHub

sales-by-region.sql

SELECT region, COUNT(*) AS orders,
       ROUND(SUM(total), 2) AS revenue
FROM orders
WHERE created_at >= {{yesterday}}
GROUP BY region ORDER BY revenue DESC

0 7 * * 1-5 · weekdays at 07:00

Daily sales: 5 regionsfrom Rowbird07:00

Daily sales

Good morning! Here are yesterday's sales.

region	orders	revenue
East	6	548.35
West	5	500.24
South	8	495.87
Central	5	382.59
North	1	85.10

daily-sales.csv.xlsx.pdf

Watch it work.

21 seconds, recorded on a real Rowbird

Fig. 1From a question in plain English to the 07:00 email.

Somebody on your team is a human cron job.

A cron script only one person understands.
A BI tool licensed for one table.
Copy. Paste. Send. Tomorrow again.

Rowbird is the small tool that does exactly this job, well, on your own server. It is not a dashboard. It is the report that shows up.

How it works

Three steps, and then it runs without you.

Write it, or ask for it.

Write SQL with schema autocomplete and a live preview. Or describe what you need, and the optional assistant (OpenAI, Anthropic, Gemini, Ollama or any compatible endpoint) proposes the query as a change you can read before you apply it.

Supports PostgreSQL · MySQL · MariaDB · SQL Server · SQLite, with TLS and SSH tunnels.

The Rowbird assistant proposing SQL for orders and revenue per region yesterday, rounded to cents, with an explanation and a diff. — Fig. 2 · A real proposal from the assistant, shown as a diff

The report schedule: weekdays at 07:00 in America/Sao_Paulo, the cron expression, and the next five runs. — Fig. 3 · Weekdays at 07:00, with the next five runs

Put it on a schedule.

Cron with a time zone, a visual builder, and the next runs spelled out before you save. 07:00 stays 07:00 across daylight saving. Retries, misfire and overlap policies, and a report that keeps failing pauses itself and tells you why.

0 7 * * 1-5  at 07:00, Monday through Friday

Deliver it where people read.

A clean table in the body of the message, files attached, or a download link that expires, needs a sign-in if you want, and keeps a log of who opened it. Files too big for a destination fall back to a link on their own.

CSVExcelJSONPDFHTML tableMarkdownText

The email Rowbird sends: Daily sales, with a table of orders and revenue per region and a link to the run. — Fig. 4 · The actual email, with its attachments

The first page of the PDF Rowbird generates for the report. — Fig. 4 · The actual email, with its attachments

Destinations

Wherever your team already reads.

One report can go to several places, each in its own mode. Templates set the subject and the message, with values escaped for each destination.

Destinations and the delivery modes each supports
Destination	Inline	Files	Link
EmailSMTP, HTML table plus plain text	Supported	Supported	Supported
SlackBot token or incoming webhook	Supported	Supported	Supported
TelegramBots, files up to 50 MB	Supported	Supported	Supported
DiscordWebhooks, Markdown tables	Supported	Supported	Supported
WebhookJSON, signed with HMAC-SHA256	Supported	Not supported	Supported
S3AWS, MinIO, R2, B2, Wasabi, Spaces	Not supported	Supported	Not supported
Uptime KumaPush monitor, up or down per run	Supported	Not supported	Not supported

Alerts

Reports that only speak up when it matters.

Add a condition and a report becomes an alert. Combine rules with all or any. When Rowbird itself has a problem, it tells you too: an alert channel with a fallback, grouped messages, a note when things recover, and a heartbeat for your uptime monitor.

01
has_rows
Only when there is something to say.
02
row_count > 50
When the list gets long.
03
revenue < 1000
When a number crosses a line.
04
changed
Only when the result changed since the last time.

Production access, handled with care

It reads your database. It never writes to it.

Read-only by default: Read-only transactions, server-side timeouts, row limits and one statement per query. Rowbird warns you when a login can write.
Parameters are bound, always: Built-in dates like {{yesterday}} and your own parameters reach the database as driver parameters, never pasted into SQL.
Secrets stay secret: Credentials are encrypted with AES-256-GCM and never shown again: not in the UI, the API, the logs or an export.
Your people, your rules: Admin, editor and viewer roles, TOTP and passkeys, OIDC single sign-on, scoped API keys and a security log.
AI that never sees a row: The optional assistant gets your question and the schema. Rows never leave, and nothing runs until you apply it.
No telemetry: Rowbird talks to what you configure. The one exception, a daily update check, turns off with a switch.

Config as code

Reports you can review in a pull request.

Export any report as YAML, keep it in Git, and apply it with rowbird apply, which prints the plan before it changes anything. Or point Rowbird at a directory and it applies it at every start. Secrets never go in the file.

reports/daily-sales.yaml

apiVersion: rowbird.dev/v1
kind: Report
metadata:
  name: daily-sales
spec:
  query: { ref: sales-by-region }
  schedule: { cron: "0 7 * * 1-5", timezone: America/Sao_Paulo }
  condition: { match: all, rules: [{ type: has_rows }] }
  deliveries:
    - channel: sales-slack
      mode: attachment
      formats: [xlsx, pdf]
    - channel: finance-email
      mode: link
      link: { expiresIn: 7d, requireLogin: true }

What Rowbird is not.

Not a BI tool. No chart builder, no dashboards to maintain.
Not an ETL tool. It does not move data between databases.
Not an uptime monitor. It reports to yours, like Uptime Kuma.
Never a writer. Your data stays exactly as it was.

Install

One binary. One minute.

A single Go binary or a small distroless image for Linux, macOS and Windows, on amd64 and arm64. SQLite inside by default, PostgreSQL when you run several instances. Releases are signed and ship with an SBOM.

Docker

Open localhost:8080 and create the first admin.

$docker run -p 8080:8080 -v rowbird-data:/data ghcr.io/rowbird/rowbird:1

The demo

Sample data, a fake inbox and four reports, ready to run.

$git clone https://github.com/rowbird/rowbird && cd rowbird/deploy/demo && docker compose up

Production

Compose with Caddy and HTTPS, systemd, or Kubernetes.

docs.rowbird.dev/install

Stop pasting. Let the bird carry it.

GitHub Read the docs

Your SQL results,delivered.