2 Workspace

2.2 Chat

2.2.1 What Chat is

Chat is the main way to work with agents in Marcto. You can use it for temporary questions, or use page chat from brand, product, and project detail pages so agents work from the current object context.

Marcto chat is not only conversation. It can combine model selection, personas, skills, attachments, resources, brands, products, and projects to produce research, copy, creative directions, analysis reports, task plans, and saved outputs.

2.2.2 Two common chat modes

Mode	Best for	Notes
General chat	Temporary questions, brainstorming, explanations, lightweight tasks	Not tied to a specific brand or project by default
Page chat	Formal work inside brand, product, or project detail pages	Carries current page context automatically

For formal marketing assets, start from page chat whenever possible.

2.2.3 What to prepare before chatting

Before sending a serious task, prepare:

• Goal: what the agent should produce.
• Business object: the correct brand, product, or project.
• Channel: Shopify, TikTok, Meta Ads, email, or another context.
• Constraints: voice, length, format, compliance boundaries, banned phrases.
• References: images, documents, reviews, competitor links, previous outputs.
• Output format: table, checklist, script, report, JSON, Markdown, etc.

2.2.4 Basic steps

1. Open general chat, or open page chat from a brand, product, or project detail page.
2. Choose a persona or team if needed. If unsure, use the default assistant.
3. Select a skill if you want the agent to follow a fixed workflow.
4. Add attachments or choose resources from the library.
5. Write the task goal and output format.
6. Ask the agent to summarize context before generating.
7. Revise, follow up, or expand the result in stages.
8. Save final work to the project, resource library, brand, or product.

2.2.5 Writing a good prompt

A useful prompt usually has five parts:

Goal: create 10 TikTok short video scripts for the current project.
Context: use the current brand, linked products, and project brief.
Audience: US women aged 25-35 who commute.
Constraints: keep the voice light and trustworthy, avoid exaggerated claims, no medical promises, each script under 30 seconds.
Output: use a table with hook, scene, voiceover, subtitle, CTA, and review risk.
Process: summarize the context you can see before generating.

2.2.6 Recommended flow

Do not ask for a final draft in one step. Use a staged flow:

1. Ask the agent to summarize current context.
2. Ask for missing information and confirmation questions.
3. Ask for an execution plan.
4. Generate structure or angles first.
5. Choose a direction, then generate copy or creative plans.
6. Ask for fact, voice, and compliance checks.
7. Save the final result.

2.2.7 Attachments and resources

Attachments are good for one-time reference. Resources are better for reusable materials.

Material	Recommended use	Examples
Product images	Attach or save as product resources	Hero image, lifestyle image, specs image
Review screenshots	Attach and explain the purpose	Extract pains, create FAQs
Brand book	Save as brand resource	Voice, logo, color rules
Competitor links	Add to project assets or chat	Compare positioning and creative angles
Previous outputs	Save as resources	Reuse approved copy or reports

After adding files, explain how each should be used so the agent focuses on the right material.

2.2.8 What to do with generated results

Do not leave useful output only in chat history.

• Final copy, scripts, reports: save to project resources.
• Brand voice, banned phrases, durable insights: confirm and save to brand.
• Product benefits, FAQs, specs: confirm and save to product.
• Campaign plans, asset lists, retrospectives: save to project.
• Temporary drafts and failed directions: skip saving unless they are useful later.

2.2.9 Common issues

If output is generic, the goal, object, channel, or output format is usually unclear. Add brand, product, project, and channel context, then ask the agent to restate what it sees.

If the agent does not use uploaded files, ask it to list visible attachments and resources before generating.

If output involves ads, claims, pricing, customer messages, budget, or compliance-sensitive content, review it manually before use.

If the answer is too long, ask for staged output: angles first, scripts second, review checklist third.