Skip to main content

Version control and promoting content

Flows and bits are the definition of how your organization works, and many teams want to treat them the way they treat code: kept under version control, reviewed before they change, tested away from live work, and then promoted to production. This guide describes how to do that with Floxar today.

Coming soon. Built-in versions for bits are planned: publishing a version, promoting it, and archiving old ones, inside Floxar. Until then, the approach below gives you version control and promotion with tools you already have.

The approach in short​

  1. Keep a test account where content is written, tried and validated.
  2. Keep your content in your own repository. An AI agent connected to your account reads flows in full and stores them as files; changes are reviewed there like any other change.
  3. Promote with an agent. Once a change is approved, the agent writes it into your production account.
  4. Gate production. Require flows to be approved before they can be run, and change live flows with care.

A test account​

A test account is a separate Floxar account where nothing affects your real work. Write and change content there, run it end to end with test trails, and invite the people who review it. One person can belong to both accounts and switch between them from their profile, with separate permissions in each.

Anyone can create a new Floxar account and use it as a test account; an administrator marks it as a test account in its settings, and Floxar then shows a banner so nobody mistakes it for production. Accounts are independent of each other: nothing links a test account to your production account. If you need both accounts to sit on the same infrastructure, contact Floxar before creating the test account.

Your content in your own repository​

Everything that defines a process can be read in full by an AI client or agent connected to your account: each flow with its starting step, every bit's content and interactive elements with their settings, the connections between bits including their conditions, and linked references. That makes it possible to keep your processes as files in a repository you own:

  • Your agent exports the flows you care about to the repository, one file per flow or per bit, in a structured format.
  • Changes are proposed and reviewed as changes to those files, with your usual review process.
  • The history of the repository is the history of your processes: who changed what, when, and why.

Floxar also records every change made to content inside each account, visible to administrators on the account's audit page, so you can check what happened in an account against what was approved in the repository.

To connect an AI client or register an agent, see Connecting an AI client and Registering an agent.

Promoting a change to production​

Promotion is the agent writing the approved content into the production account. Each account is connected separately, so the agent works with one connection to the test account and one to production. A few things are specific to each account, and the agent must handle them:

  • Identifiers. Bits, flows and references get new identifiers in each account. Connections and conditions refer to other bits and to answers by identifier, so the agent must keep track of which item in the test account corresponds to which in production.
  • Order. Bits come before the connections between them, a flow's starting bit before the flow, and references before they are linked.
  • Credentials. The Secrets Vault is separate for each account. Create the credentials a process needs in production and grant your agents access there; they are never copied.
  • Shared bits. A bit can be part of several flows, so promoting a change to a bit changes every production flow that uses it.

Gating production​

  • Approval before running. An administrator can require new flows to be approved before anyone can run them, so a promoted flow waits for a person to approve it in production.
  • Changing live flows. Trails in progress follow the current version of a flow, so an edit to steps not yet reached affects runs already underway. For a change that should not touch runs in progress, promote it as a new flow and deactivate the old one: running trails continue on the old flow, and new runs start on the new one.
  • Check the result. After promoting, run a test trail through the changed path in production and review it.

Why this works well with agents​

Because AI agents can read and write everything that defines a process, the same agent can keep your repository, prepare a change, write it into the test account for you to try, and promote it when it is approved. Your people review the change itself, as files and as a test run, rather than retyping it by hand. For more on putting agents to work around Floxar, see Designing AI workloads with Floxar.

Last reviewed: 2026-10-06