Documentation

Using the Reach Agent

Ask one question in plain language. The Reach Agent reads it, dispatches the specialists that can answer it, and returns a single response with citations to the data behind every number. This page covers how to ask, what comes back, and how to read it.

There is nothing to configure

Every agent follows the same pattern. You ask a question in plain language, the agent queries the commerce map, and you get back a structured response with specific numbers, citations, and next steps. No setup, no configuration, and no data to upload first.

You do not need a catalog. Market questions about a category, a brand, an occasion, or a season are answered from the graph alone, which is how analysts and brand teams use the platform. Connecting a catalog adds a second layer: the same market answers, resolved against your own products.

One front door, five specialists

You are never required to know which specialist owns which data. Ask the Reach Agent and it decides. The specialists are documented below so you can direct a question at one of them when you already know what you want.

Reach Agent

The front door. Routes your question to the specialists that can answer it.

What to include in your question

Describe your situation or goal in a full sentence. "Outdoor furniture" is vague. "I sell outdoor dining sets and want to know where my gaps are, whether I am priced right, and when demand peaks" is specific enough for the Reach Agent to dispatch three specialists and combine the answers.

What the response looks like

A unified answer organized by theme rather than by agent, with evidence citations and next steps where supported.

What to do with the output

Review the evidence and next steps. If the response offers a Lens Report, treat it as an option for deeper analysis.

Sample questions
  • I am launching a line of insulated water bottles under $30. What does the landscape look like?
  • Which voice searches around my catalog are still unclaimed?
  • What is the most valuable analysis I can run on my outdoor furniture catalog?

Search Intent

The words and questions buyers use, and which of your products those searches do not reach yet.

What to include in your question

Name your product category, a specific product, or the buyer language you want to investigate. Giving the agent an actual consumer phrase to match against produces a sharper answer than a category alone.

What the response looks like

Buyer keywords and questions with search volume, CPC, and competition for each. Ghost products are named alongside the listing language that does not match. Title suggestions include the exact consumer phrase and its monthly search volume.

What to do with the output

Update product titles and descriptions with the phrases the agent surfaced. Build FAQ or landing page content that answers the voice questions. Start with the ghost products carrying the highest search volume.

Sample questions
  • What questions do people ask about travel mugs on voice search?
  • Which of my product titles would not show up for "best yarn for baby blankets"?
  • Show me the language buyers use for organic dog food.

Demand Intelligence

When demand peaks, who is searching, and where a market is heading.

What to include in your question

Name the category, audience, or time period you want to understand. A category name returns the full demand profile. Naming an audience returns the segment breakdown. Asking about direction returns the trend.

What the response looks like

Demand direction and magnitude with available supporting numbers. When the data supports a seasonal pattern, the response identifies peak and trough weeks. Available audience context may also be included.

What to do with the output

Use seasonal patterns to decide when campaign, content, and inventory work should begin, accounting for the merchant's own lead time. Review audience and trend findings alongside the merchant's own catalog and plans.

Sample questions
  • When does demand for pool floats peak?
  • Who searches for "best hiking boots for wide feet"?
  • Is demand for electric scooters growing or declining?

Taxonomy Gap

What buyers search for that you have no product for.

What to include in your question

Name the category you sell in or want to expand into. Narrowing to a subcategory returns the best expansion opportunities rather than the full picture.

What the response looks like

A ranked list of gaps scored by opportunity. Each gap carries the keyword, monthly search volume, CPC, competition level, and how many product clusters already serve that demand.

What to do with the output

Review search volume, competition, relevance, and the number of product neighborhoods already serving the need together. A gap with stronger demand and lighter relevant coverage may deserve earlier review, but no single threshold defines an underserved market. CPC provides advertising-cost context for the term; actual campaign cost depends on the merchant's auction and campaign.

Sample questions
  • Where are the gaps in my outdoor furniture catalog?
  • Which kitchen knife categories are still open to me?
  • Which subcategories of yoga equipment have demand but low competition?

Pricing Intelligence

Where each product sits against the market, and what to change.

What to include in your question

Name the product or category you want priced. Including a price point anchors the analysis.

What the response looks like

Comparable product rows, observed price ranges, market position, and available price history, with citations to the evidence.

What to do with the output

Review comparable products and price position alongside brand, product differences, margin, and strategy. The merchant decides whether to change a price.

Sample questions
  • How do my kitchen knife prices compare to the market?
  • Am I overpricing my santoku knives?
  • Where is my margin most at risk from mispricing?

Campaign Architect

A Google Ads campaign structure built from your catalog.

What to include in your question

Name the products or category you want to advertise. Stating a budget produces a more tailored allocation.

What the response looks like

A draft plan with ad groups organized by cluster or theme, keyword lists with search volume and CPC, bid ranges, budget allocation weighted by opportunity, and negative keyword suggestions. Formatted for Google Ads Editor import.

What to do with the output

Review the draft. Remove keywords that do not fit your catalog and adjust bids to your margin targets. Review opportunity, product fit, competition, CPC, budget, and merchant economics before choosing which ad group to launch first.

Sample questions
  • Draft a Google Ads campaign for my insulated water bottles.
  • What are the lowest-competition, highest-volume keywords for hiking boots?
  • Build me an ad group structure for pet grooming supplies.

Every answer shows its work

Responses include a Data Evidence block listing the specific rows the agent cited, labeled so numbers in the answer trace back to the supporting data. A Data Freshness stamp states when the underlying data was last updated, so you can review how current the analysis is before acting.

If an agent offers a Lens Intelligence Report at the end of a response, it is an option for deeper analysis. The offer names the report and its cost. Seventeen reports are available, and each is generated by a four-pass analysis pipeline.

One rule governs the Campaign Architect specifically. It produces drafts for your review. It does not push to Google Ads, commit spend, or modify a live campaign.

Three ways in

MethodWho it is for
Google Cloud Marketplace consolePay As You Go subscribers. Ask questions directly, no integration work.
A2A JSON-RPC APIAPI Access subscribers integrating the agents into their own systems.
Agent cardsPublished at /.well-known/agent-card.json and /.well-known/agents.json for automated discovery.

The agents are hosted on Google Cloud Run.

Getting help

Questions about agent responses, billing, API access, or data coverage go to support@reach.dog. Include the question you asked and the evidence references from the response so we can trace the answer back to the underlying rows.