TechAppForce

What is TAFI?

AI-assisted application building with a confirmation step.

TAFI reads your application and the platform's knowledge base, proposes changes, previews each one and applies it only after you confirm. It works through the platform's tool gateway, which exposes about 50 tools for entities, queries, lifecycles, screens, menus, access, workflows, notifications, reports, Virtual URLs and diagnostics.

TAFI accelerates development; people stay in control of every change. It does not change your application silently.

What TAFI is

TAFI is TechAppForce's AI engine for building apps. You describe what you need, for example “a leave tracker where employees file requests and managers approve them”. TAFI asks clarifying questions where it is unsure, drafts requirements, designs the data model and lists every change it will make. It builds only after you approve each of those steps.

TAFI does not generate a separate codebase. It writes into the same app model you use in the web app and in TAF Studio, so the result is an app your team can read, change and release like any other change.

What TAFI builds

  • Objects and fields, including the table behind each object and its default list and detail queries.
  • Access for the app's roles on each object, so the people who should see the records can.
  • Screens, generated from the object's real fields, and checked against the platform's component rules so they render rather than only save.
  • Menu entries so users can reach the new screens.
  • Lifecycles, triggers and workflows where the feature needs them. On an environment where the workflow engine is not deployed, TAFI saves the workflow and tells you it will not run yet.

Where you use TAFI

  • Inside TAF Studio. Open the TAFI panel from the title bar, the status bar or the side bar. It works on the app you are signed in to, can attach files and screenshots, and knows which file you have open. Use /build for a staged build. See The build flow.
  • From an AI client. TAFI's tools are served by an MCP gateway that clients such as Claude, Cursor or Codex can connect to. See Use TAFI from AI clients.

Inside Studio, TAFI runs on an AI engine you supply: Claude Code if it is installed and signed in on your computer, or your own key or subscription for Anthropic, OpenAI or an OpenAI-compatible endpoint. Your TechAppForce sign-in and your AI account are kept separate. A key you enter is stored encrypted on your computer and is not sent to TechAppForce.

Ask, Plan and Agent

A mode chip under the message box sets what TAFI is allowed to do:

  • Ask answers questions about your app and changes nothing. “List this app's workflows” or “Why can't managers see timesheets?”
  • Plan works out the steps and shows them. You can edit the plan, then choose to build it.
  • Agent makes the change: it edits your local files and, through the gateway, the live app.
Important

Agent mode changes the live app. TAFI is instructed to preview each live change and wait for your OK, but in Agent mode nothing in Studio forces that pause. Use Plan mode first for anything risky, and prefer a development app.

Staying in control

  • Before TAFI edits files it shows how many files it wants to change, with a before-and-after diff.
  • After each turn it lists the files it changed and the items it changed on the platform, each with a link to open it in its designer.
  • Undo all changes reverts TAFI's edits to your local files. For live changes, Studio keeps a copy of what TAFI changed, and you can ask TAFI to put it back.
  • /review checks that the session's changes really work, including that each screen renders, and flags anything half-applied.
  • When TAFI wants a tool its mode does not allow, it asks you to allow or deny it. Shell commands stay blocked unless an administrator enables them.

Where to go next