Convas

Agent

Agent overview

The Convas Agent is your conversational partner: it reads and writes brief, outline, design, and canvas on one project—from sources to slides to edits.

Agent chat and task progress · coming soon

What it is

The Agent is how you work in Convas. Describe what you need in the chat panel; it reads and writes across brief, outline, design, and canvas on the same project—from organizing sources to generating slides to editing a single page.

Brief, outline, design, and canvas are four views of one project—switch anytime, with no required order. The Agent always works on the same deck throughout.

How it works

Think of three parts: your instructions (chat, @ references, attachments), what the Agent can do (read sources, write outlines, render, edit pages, and more), and the model you pick (speed, quality, Credits). For larger jobs it may plan first—see Planning & execution.

Capability index

Browse by task type for details and examples:

How to collaborate

  • Describe goals in natural language: audience, tone, and style.
  • @ references: point at assets, slides, or design styles for precise context.
  • Attachments: send PDFs and images; the Agent can parse and persist them.
  • Canvas edits: click-to-edit and refine are often faster for copy and chart tweaks (code mode).
  • Model choice: different models trade off speed, quality, and Credits.

Chat vs canvas

  • Better in chat: outlines, overall style changes, bulk rendering, document parsing, narrative structure, exporting PDF / PPTX.
  • Better on canvas: a few words on one slide, chart values, single-page layout (code mode).
  • Output mode matters: code mode supports refine and Tweak; image mode relies more on chat regeneration.
Plans, parallel rendering, background batches, and limits—see Planning & execution.
On this page