Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.marcenta.ai/llms.txt

Use this file to discover all available pages before exploring further.

AI Agents are automated workflows that carry out multi-step marketing analysis tasks on your behalf. You configure an agent run with your inputs, and the agent works through a series of steps—fetching data, applying research strategies, filtering and grouping results—then delivers a structured artifact you can review and use. Agents handle the repetitive, time-consuming parts of marketing research so you can focus on decisions.
AI Agents require the Orbit plan. If your organization is on a lower tier, you’ll see an upgrade prompt when you navigate to the Agents page.
Select Agents in the left sidebar. The page shows a list of your configured agent schedules along with their status and the result of the most recent run.

Available agents

Keyword Research Agent

The Keyword Research Agent discovers and organizes keyword opportunities for your marketing program. It pulls from your existing keyword data, researches new terms based on your inputs, filters for quality, and groups keywords by theme into a structured plan you can review and export.

Run the Keyword Research Agent

1

Open the agent configuration page

From the Agents page, select Keyword Research Agent (or navigate directly via the sidebar). The configuration form opens.
2

Choose a keyword category

Select the category of keywords you want to research. Available categories include branded terms, competitor terms, product-focused terms, and others. Selecting a competitor category requires you to add at least one competitor name.
3

Configure seed inputs

Depending on the category, provide seed keywords, brand names, or competitor domains. The agent uses these as starting points for its research strategy.
4

Set priority and schedule

Choose when you want the agent to run:
  • Now — starts immediately with the highest priority
  • Today — runs during the current day
  • Overnight — runs during off-peak hours
You can also set a recurrence interval (in days) to run the agent automatically on a schedule.
5

Start the run

Select Run Agent. The agent is queued and begins processing. You can watch progress in real time on the run detail page.

Monitor a run in progress

Open the run detail page by selecting any active run from the Agents list. The page shows the current step the agent is working through:
Step labelWhat’s happening
Analysing existing dataThe agent is reviewing your current keyword coverage
Planning keyword researchThe agent is building a research strategy
Finding keywordsThe agent is discovering new keyword candidates
Evaluating resultsThe agent is scoring and filtering candidates
Filtering keywordsThe agent is removing low-quality terms
Grouping by themeThe agent is organizing keywords into clusters
Ready for your reviewThe agent has paused and is waiting for you to review intermediate results
Building your keyword planThe agent is finalizing the artifact

Review and approve

Some agent runs include a review checkpoint. When the agent reaches a paused state (labeled “Ready for your review”), you’ll see an option to Approve or Reject the intermediate output before the agent continues to the next step. This gives you control over which keywords or strategies advance to the final artifact.
If you want to adjust the direction of the research before continuing, use the Revise option on a paused run. You can override inputs and restart the agent from any earlier step, carrying forward the work already completed.

Artifacts

An artifact is the structured output produced at the end of a successful agent run. For the Keyword Research Agent, the artifact is a keyword plan organized by theme, with each keyword group containing:
  • Keyword terms discovered and selected
  • Volume and relevance signals
  • Thematic grouping labels
Access artifacts from the run detail page once a run completes. Artifacts are stored and accessible from your run history.

Run history

Every run is recorded with its status, inputs, timing, and output. From the Agents page, select any schedule to see the history of runs associated with it. Select an individual run to view its full detail, including which step it completed last and any error messages if a run failed.

Retry a failed run

If a run fails, open the run detail page and select Retry. The agent restarts from the step where it failed, carrying forward all completed work so you don’t lose progress.
Each run supports up to 10 retry attempts and 10 revision cycles. If a run has reached its retry limit, create a new schedule and run.