Questions and answers

Flowcall is a decision-operations tool: it turns recurring data work into pipelines that run on a schedule, checks the data before anything downstream runs, routes decisions to a person for approval, and records every decision with its outcome. It runs on your own Windows machine or server; your data never leaves it.

Analysts, operations teams and small data teams who run recurring reports and decisions today with SQL, Excel, Python scripts and email. It is designed for people who can write a query or a short script but do not want to run infrastructure. It is not built for platform engineering teams that need a distributed orchestrator.

Both, in that order. The pipeline part (SQL, Python, commands, API calls, mailboxes, quality checks) gets the data ready. The decision part (approval steps with the data in front of the approver, a decision log, outcome checks, a hit rate) is what makes it different from an orchestrator.

No. Flowcall runs entirely on your own computer or server. The only outbound connections are the ones your tasks make (your warehouse, an API you call) and, if you enable them, email or Teams alerts and the optional AI assistant. The assistant can run on a local model so even that stays inside your network.

For analysts and small teams, yes. Flowcall installs on Windows in minutes with one file, needs no Docker or Kubernetes, and has a visual builder instead of Python DAG files. It is not a replacement for Airflow at platform scale: it runs on one machine without distributed workers, so large engineering teams should keep Airflow, Prefect or Dagster.

If your Airflow mostly runs a handful of scheduled SQL and Python jobs for an analytics team, migrating is a day's work and you gain quality checks, approvals and a decision log. If you have hundreds of DAGs, backfills and Kubernetes workers, upgrade to Airflow 3 instead; Flowcall is not built for that scale.

For scheduled data work, yes, at a fraction of the price: Flowcall is priced per company, not per seat, and the Starter plan is free. Alteryx has hundreds of drag-and-drop transforms; Flowcall expects you to write the transform in SQL or Python and gives you the scheduling, checks, approvals and decision tracking around it.

Those are orchestrators for engineers who write pipelines as code. Flowcall is for the person who receives the pipeline's output and has to decide something with it: it adds approval steps, a decision log with outcomes, and receipts. It also runs on Windows without containers, which none of the three do natively.

Not yet. Flowcall runs SQL on a schedule with quality checks, but it does not have dbt's model graph, ref() lineage or incremental models. A dbt-style modeling layer is on the roadmap; today, teams that use dbt keep it and use Flowcall for the scheduling, checks and decisions around it.

It replaces the fragile parts: the manual refresh, the copy-paste, the formula nobody understands, and the knowledge that lives in one person's head. It keeps Excel as the front end: upload .xlsx files as inputs and download any result as an Excel workbook.

SQL against your warehouse (SQLite, Postgres, SQL Server, MySQL, Snowflake, BigQuery, Databricks), Python, Windows or shell commands, API calls, reading a mailbox, moving files, quality checks, predictions with a trained model, optimization, and approval steps.

A quality-check task is a list of plain rules: rows >= 1, not_null: account_id, unique: invoice_id, range: amount 0 50000, allowed: region NA, EMEA, anomaly: amount. If a rule fails, the pipeline stops there and, if alerts are on, a person gets an email or Teams message with the failing rule.

The pipeline pauses, the request appears in an inbox with the actual data table, and a named person approves or rejects with a comment. Rejection stops the pipeline; a time limit expires the request. Every decision is recorded with who, when, on what data, and a dollar value if you set one.

A downloadable record of one decision: the table the approver saw, who decided, when, their comment, the value at stake, and what happened afterwards. It carries a fingerprint so it can be shown to an auditor or finance reviewer unchanged.

An approval step can name a follow-up pipeline and a rule. That many days after approval, Flowcall runs the pipeline and judges the decision (for example, it worked if sla_breaches.csv has zero rows). The decision log then shows a hit rate: what share of your team's approved decisions actually worked.

Yes, for tabular data. A Predict task learns from a CSV, holds out 20% to score honestly, tries linear and boosted-tree models, keeps the better one, and saves it with its scores. A second Predict task scores new rows. It is built for analysts; a data science team with custom feature engineering will still want their own tooling.

Yes. An Optimize task makes one decision per row under limits written in plain lines (maximize: value, total <= 10000, budget: cost <= 250000, each <= capacity, group region total <= 4000) and reports which limits held the answer back.

Yes. It reads your pipelines and run logs, explains failures, and drafts or fixes tasks as proposals you apply with a button. It runs on Hugging Face (the default: paste one free token), Anthropic's API, or a local model (Ollama, vLLM, LM Studio) so nothing has to leave your network.

On Windows 10 or 11 (64-bit), download the installer and run it: everything Flowcall needs is included, and it does not need administrator rights. On Mac or Linux, or if your IT team prefers it, the zip runs with Python 3.11 or newer. No Docker, no database server, no cloud account.

Yes, natively; Windows is the primary platform. Command tasks run in Command Prompt, SQL Server connects through ODBC with Windows authentication, and Task Scheduler can start Flowcall at login.

Yes. Owners manage people, editors build and run pipelines and decide approvals, viewers can look and download. New people get a temporary password and choose their own on first sign-in. Single sign-on is on the Enterprise roadmap.

Yes, with Docker or as a plain Python process behind HTTPS. A hardened mode disables Python and command tasks so a server only runs SQL, API and check steps.

Every save is a version you can restore. Pipelines can be pushed to and pulled from a GitHub repository as JSON files, and a GitHub Actions workflow validates them on every pull request and deploys on merge.

Passwords are hashed with scrypt. Secrets such as database passwords and API keys are encrypted on disk under a key only your user account can read, substituted into tasks only at run time, and masked out of every log line.

It ships with the standard defenses: hashed passwords, encrypted secrets, session and CSRF protection, login lockout, strict browser security headers, role-based access, and an audit log. It has not yet had a third-party penetration test or SOC 2 audit; those are planned before enterprise contracts.

Anyone with the Editor or Owner role can add Python and command tasks that run as the Flowcall service account. Treat editor access like access to the machine, or run the server in hardened mode where those task types are disabled.

By email, at the address on your invoice, answered within one business day. We do not offer phone support and we do not schedule support calls: one written thread keeps the exact error text and the history, which is how a small vendor answers everyone the same day.

You can request a full refund within 7 days of first use of a paid plan. After 7 days of use we do not issue refunds, but you can cancel at any time and keep access until the end of the period you already paid for. If you were charged but never installed, we refund it as a courtesy whenever you ask.

From Settings → Plan → Manage billing inside Flowcall, or by email. Cancelling stops the next renewal; your key stays valid to the end of the paid period plus a short grace window.

So you can evaluate Flowcall for as long as you like before paying, instead of relying on a short trial and a refund. Starter includes five pipelines, quality checks and the local AI assistant, with no time limit.

Starter is free: five pipelines, quality checks, 500 MB of files, local AI assistant, one machine. Team and Enterprise are monthly subscriptions priced per company, not per seat; current prices are on the pricing page. Cancel any time.

The Starter plan is free without a time limit. Paid plans can be cancelled any time; the key stays valid to the end of the paid period plus a grace period.

A subscription payment is refunded in full if you ask within 7 days of purchase and have used the software for 7 days or less. After 7 days of use, the current period is not refundable. Cancel any time and keep access to the end of the paid period.

No. All support is in writing through the support email, on business days. Team plan requests are answered within two business days, Enterprise within one. Written support means every answer is on record and can be checked.

Yes, for Enterprise. Ask for a quote and we issue the license key by hand when the invoice is paid.