Skip to content
Viberia Docs

MarketResearchers

MarketResearchers is the research pipeline in a single building — a Researcher who gathers source material, an Analyst who pulls insight out of it, and a Presenter who turns the analysis into a slide-style deliverable.

At a glance

Section titled “At a glance”
  • Default agents: Researcher, Presenter, Analyst
  • Best for: Market scans, competitor analysis, landscape reviews, and any research task where you want both a written analysis and a presentable summary at the end
  • Skills it ships with: /start-research, /handoff-research-to-presenters
  • Place from: Building Type Drawer → MarketResearchers

How It Works

Section titled “How It Works”

MarketResearchers structures research into three roles so you don’t get one agent trying to do everything at once. The Researcher does the legwork — searches, reads, captures notes. The Analyst sits with the captured material and produces synthesis: trends, comparisons, recommendations, the boring written core. The Presenter takes the analysis and packages it for someone else to consume: a deck, an HTML one-pager, or a short executive summary.

You start a research effort by talking to the Researcher (or by invoking /start-research, which prompts you for a topic and scope). The Researcher works through your project’s folder_path, dropping notes, references, and raw findings into structured files. When the Researcher has enough material, the Analyst takes over and writes the analysis on top of those notes. When the analysis is in place, /handoff-research-to-presenters passes the result to the Presenter, who produces the polished output.

Each role has its own conversation, and the artifact (notes, analysis, presentation) lives in the project folder on disk. You can stop at any stage — sometimes you only want the Researcher’s notes, sometimes you want the analysis but not slides — and you can iterate within a role without restarting the others.

This is not a fully automated research robot. The three agents need direction from you about scope, sources, and the question you actually want answered. The structure of the building keeps each step honest: the Researcher is graded on coverage, the Analyst on synthesis, the Presenter on clarity.

Default agents

Section titled “Default agents”

  • Researcher. The Researcher gathers source material and produces structured notes. Talk to the Researcher first, with the topic and the scope you want covered. The Researcher works through web searches (if you’ve enabled a search MCP), file system reads, and any other sources you wire in, and writes findings to the project folder. They do not draw conclusions — they collect.

  • Analyst. The Analyst reads the Researcher’s notes and produces the written analysis: trends, comparisons, frameworks, recommendations. Talk to the Analyst once the Researcher has enough material, or directly if you already have notes from somewhere else. The Analyst is the one to push back on if you disagree with a conclusion — they’re the role that owns synthesis.

  • Presenter. The Presenter turns the Analyst’s written output into a deliverable shaped for someone who didn’t run the research: a slide deck (Markdown or HTML), an executive summary, or a one-pager. Talk to the Presenter once you have an analysis you’re happy with. The Presenter can also be asked to produce alternate formats from the same analysis (slides for one audience, a brief for another).

Configuration

Section titled “Configuration”

MarketResearchers follows the standard three-agent configuration model. From the Building Window you can:

  • Override each agent’s prompt — for example, narrow the Researcher to specific source domains, give the Analyst a house framework (SWOT, JTBD, Wardley maps), or constrain the Presenter to a specific deck template.
  • Set per-agent models. A common shape is Researcher on a model with good tool use and long context, Analyst on a strong reasoner, Presenter on a fast formatter.
  • Add MCP servers. A web search MCP server on the Researcher (for example, Brave Search) is one of the most common additions. The filesystem MCP server tightens read/write control on note files.
  • Add additional directories if your project pulls from material outside folder_path — for example, a shared notes folder on disk.

The two skills under At a glance are installed by default. You can disable either per agent.

Typical workflow

Section titled “Typical workflow”
  1. Place MarketResearchers in a project. Make sure the project’s folder_path is set — research output will live there.
  2. Open the building and click the Researcher.
  3. Invoke /start-research or describe the topic in plain language. Give the Researcher a question (“What are the three main approaches to X, and who’s shipping them?”) and a scope (“Last 12 months, English-language sources, no academic papers”).
  4. The Researcher works through the question and drops findings into structured Markdown files under your project folder.
  5. Skim the notes. If coverage is thin in some area, push back on the Researcher.
  6. When the notes are good enough, ask the Analyst to write the analysis. The Analyst reads the Researcher’s notes and produces analysis.md (or whatever filename you prefer).
  7. Iterate with the Analyst until the analysis says what you want it to say.
  8. Invoke /handoff-research-to-presenters. The Presenter receives the analysis as a queued message and produces a slide deck or summary.
  9. Open the result in the Toolbox — the Markdown Editor for written summaries, the Web Browser for HTML decks.

Limitations

Section titled “Limitations”
  • The quality of the output is bounded by the sources the Researcher can reach. If you don’t wire in a search MCP, the Researcher is limited to what’s already in the project folder and what the model knows.
  • The Presenter writes Markdown or HTML deliverables. It does not export .pptx or .key directly. You can re-render its output in a presentation tool if needed.
  • MarketResearchers is a writing pipeline, not a fact-checker. The Analyst can flag uncertainty, but you are responsible for verifying claims before sharing them externally.
  • The three agents share a project folder but not a conversation. Context handoff happens through the files they write (notes, analysis) and through /handoff-research-to-presenters — make sure those artifacts are clean before each step.
  • Research that requires authenticated sources (paid databases, internal wikis) needs the right MCP server set up first; the agents have no special access by default.
Section titled “Related”