Working with the AI engineering team

On every plan except Deploy, GetFlowing's AI engineering team plans, writes, tests and changes your apps. You ask for what you want in plain English, and you decide what ships.

Ask in the app's chatCopy Anchor Link

Each app has a chat. Ask for a feature, a fix, an investigation or a change, and the agent does the work in the app's repository.

• For a new app, the team can write a plan first for you to check. See Planning a new app.
• Each change is made on a branch, as a pull request you can read and review. A self-heal fix after a failed deploy is the exception: it goes straight to the app's branch. See Self-healing deploys.
• When you merge from the chat, the change deploys. Say Merge PR #N and deploy, or merge several in order: "Merge PR #9 then #1 and deploy". See Deploy on push.
• Before the app is released, each stage of a bigger build can go on the app for you to try.

After an import, the chat suggests a first step, such as Add a Dockerfile, Write a .env.example, Run the tests or Explain how this app is put together. Choosing one fills in the message for you to send.

You can also start work from Slack, or from a connected AI assistant. See Slack and AI assistants.

How much it asks youCopy Anchor Link

By default the agent asks before risky changes, such as a database migration, a dependency change, or a change to how the app is built and deployed. You can have it ask before every action, or approve everything. See Agent controls.

One agent, or a teamCopy Anchor Link

By default you work with one agent. Owners and admins can switch on the full team in Settings, under Workspace, on the Agent and deploys tab, under AI agent team. The team is led by a team lead, with a project manager, a developer, a DevOps engineer, a QA engineer and a product designer. The team plans, advises and asks before any real action runs.

With the team on, What you call your team lets you rename its members, and How your team works with you sets how long and how technical its replies are.

Editing code yourselfCopy Anchor Link

The app's Build tab has a built-in code editor, with the chat beside it, so you can change a file yourself and carry on with the agent. On Scale and Enterprise, Instant preview and VS Code in the browser are included too. See Preview, plan and code.

Fixing failed deploysCopy Anchor Link

When a deploy fails because of your code, the agent can fix it and deploy again by itself. See Self-healing deploys.

How agent work is countedCopy Anchor Link

Agent work is measured in two ways, and both are shown under Usage & billing.

Agent work units. Each piece of work the agent finishes uses one unit: creating an app, changing one, running tests, analysing logs, and the like. Deploys, rollbacks, imports and merges use none. Each plan includes a number of units a month:

• Starter: 20 per month
• Pro: 75 per month
• Business: 200 per month
• Scale: 500 per month

AI credits. The AI models the agent uses are paid for in credits, at cost. Each plan with the agent includes a monthly allowance, and a big piece of work uses more than a small one.

Chat replies, plans and builds all use credits. If you run out of units or credits, the agent's replies are limited and it takes on no new work until the next month. You can add capacity from Usage & billing at any time: bought credits never expire, and are used after your monthly allowance. See Plans, credits and limits.

Background and scheduled jobsCopy Anchor Link

Background and scheduled jobs are included from Pro up. There are no controls for them in the dashboard yet: ask the agent in the app's chat to build the work into your app.