---
title: "Models, plan mode & chat sessions"
description: "Pick models, plan before building, fork sessions, and attach images to guide the AI."
url: https://evonx.ai/docs/evolve/models-and-sessions
locale: en
updated: 2026-09-04
---

# Models, plan mode & chat sessions

Pick models, plan before building, fork sessions, and attach images to guide the AI.

The chat has more controls than the message box suggests. This page collects the
ones that change how the AI works for you: model choice, plan mode, sessions, and
attachments.

## Choosing a model

The model selector in the chat lets you pick which AI model handles your request.
The default is a sensible balance; switch when a task calls for it — a stronger
model for a gnarly architectural change, a faster one for quick iterations.

> **Note**
> Model availability can vary by plan, and different models consume credits at
> different rates — see [pricing](/pricing).

## Plan mode

Plan mode makes the AI stop after planning: it lays out the steps it would take
without building anything. Use it when the request is large, risky, or vague in
your own head — reviewing a plan costs seconds, reviewing a wrong build costs a
revision round. Approve the plan and execution proceeds as usual.

> **Note**
> Plan mode is rolled out progressively and may not be enabled for every workspace
> yet. If you don't see it, ask the AI to "plan first, don't build yet" — the habit
> works either way.

## Sessions: parallel and forked conversations

A business isn't limited to one conversation:

- **Multiple sessions** keep unrelated topics separate — one session about
  invoicing, another about the mobile layout.
- **Forking** copies a session at its current point so you can explore an
  alternative direction without losing the original conversation's context.

## Attachments and feedback

- **Attach images** to any message: screenshots of bugs, photos of paper forms to
  digitize, competitor screens, sketches. Visual context routinely beats long
  descriptions — and for pointing at your own app,
  [annotate the preview](/docs/preview-and-debug/annotate-preview).
- **Rate responses** with the feedback controls on messages. It calibrates the
  assistant for your workspace and flags bad answers to the platform.

## Related

- [Effective prompting on Evonx](/docs/evolve/effective-prompting) — what to put
  in the message itself.
- [Evolution Threads explained](/docs/evolve/evolution-threads) — where chat ends
  and structured work begins.