Use case

Let your Agent operate; keep the decision surface human-readable.

Some users want their Agent to configure, test, and run content workflows. Others want to make a quick adjustment themselves. GrowFlux supports both through the same underlying contracts and a visual evidence layer.

Start a 3-day trialRead the docs

Agent-operated flow

  1. The user states the business outcome.
  2. The Agent discovers valid workflows, routines, routes, and costs.
  3. The Agent presents a plain-language plan for confirmation.
  4. The Agent runs and monitors the approved work.
  5. The user reviews visual outputs and evidence, then gives focused feedback.
  6. The Agent uses Test & Tune or a scoped configuration change instead of rebuilding everything.

Human-operated adjustment

The user can open the same workflow, adjust a simple prompt/model setting, compare the result, and save it. They do not need to understand run groups, branch topology, provider payloads, or internal IDs unless they intentionally open advanced evidence.

Control remains explicit

An Agent's ability to call a tool does not imply permission to publish, spend without confirmation, or cross a tenant boundary. GrowFlux exposes capability and confirmation semantics together.

AI Agent guide Test & Tune Security

Last updated: · Maintained by GrowFlux