Reach Dog · AI Agents for E-commerce

Find the buyer questions your products can answer. Put the evidence to work.

Start with a product. Ask the Reach Agent which buyer situations are worth exploring, or bring a need you already have in mind. Investigate the buyer language and relevant product evidence, then choose which connection deserves action. It brings the appropriate specialists together and returns a cited answer, so you can check the reasoning before changing your catalog or content.

You do not need to learn the agent architecture. The Reach Agent decides which specialists and evidence are needed.

Available on Google Cloud Marketplace. Pay As You Go usage pricing, or API Access for enterprise integration.

Run the Reach Agent View on Google Cloud Marketplace

Why This Is an Agent Workflow

Find the audience connection. Give it a useful voice.

Someone may discover an idea on social, search with a short keyword, ask an assistant a detailed question, and compare facts on a product page. The Reach Agent helps investigate the need and evidence; content tools help prepare an answer for the chosen format. Follow a worked example across those moments.

A useful answer has to separate a buyer question into What, Where, When, Who, and Why. It must compare those signals with product facts, retrieve dated market evidence, reject unsupported claims, and preserve citations. It then has to turn the supported finding into a decision the merchant can review. Repeating that work across a catalog is what makes the agent necessary.

Want to see the method before connecting a catalog?

Every How-to-Sell category guide includes a free category-specific agent blueprint. Add verified facts for one product. The agent checks six buyer questions and can create four evidence-traced drafts for every supported connection, up to 24 drafts for review.

Choose a free category agent or see the complete workflow.

The Agents

One Reach Agent. Five specialists.

Each specialist investigates a different part of the same visibility problem. The Reach Agent decides which specialist to dispatch when the question spans more than one. Every agent reads from the same map, so the answers line up across questions.

Reach Agent

Not sure which agent to ask? Describe the problem.

The Reach Agent reads your question and dispatches it to the right agent. For questions that span multiple domains, it sends the question to several agents in parallel and assembles a unified answer. You do not need to learn the product to use it.

$0.25 per question on Marketplace

How to use

What to ask: Describe your situation or goal in a full sentence. The more context you give, the better the Reach Agent classifies. "I sell outdoor dining sets and want to know where my gaps are, which comparable listings to inspect, and when demand peaks" dispatches to three agents at once.
What you get: A unified answer organized by theme (gaps, pricing, demand), not by agent. Each finding cites the data row it came from. Ends with 2-3 specific next steps.
What to do with it: Review the cited evidence and next steps together. Use the findings that fit your catalog, economics, and current priorities.

Example questions:

I'm 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's the most valuable analysis I can run on my outdoor furniture catalog?

The specialists it runs

Search Intent Agent

How are buyers describing products like yours, and where are the language gaps?

Product listings use one vocabulary. Buyers may search in another. This agent makes the gap visible by returning observed keywords and questions, candidate product matches, ghost products, and listing-language suggestions for merchant review.

$0.50 per question on Marketplace

How to use

What to ask: Name your product category or a specific buyer phrase you want to investigate. "Which of my product titles would not show up for best yarn for baby blankets" gives the agent a specific phrase to match against your catalog.
What you get: Buyer keywords and questions with available search volume, CPC, and competition. Ghost products with the current language gap, plus possible listing-language changes for review.
What to do with it: Check product fit before changing a listing. You decide whether the evidence warrants a title change, supporting content, a campaign test, or no action.

Example 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 Agent

When and where does interest around these products appear?

This agent tracks available demand signals across 12 dimensions with up to 63 weeks of history. It returns observed seasonal patterns, audience breakdowns, and trend context so you can review when and where interest appears before making an inventory or marketing decision.

$0.50 per question on Marketplace

How to use

What to ask: Name the category, audience, or time period. "Pool floats" gives the full demand profile. "Who searches for hiking boots" gives the audience breakdown. "Is demand for electric scooters growing" gives the trend.
What you get: Demand direction with specific numbers (e.g., "search volume grew 34% from week 12 to week 38"). Seasonal patterns with peak and trough weeks named. Audience breakdowns by demographic segment.
What to do with it: Use the observed timing, audience, and trend evidence as inputs to your plan. You decide whether they warrant a campaign, targeting, or inventory change.

Example questions:

When does demand for pool floats peak?
Who searches for 'best hiking boots for wide feet': age, gender, device?
Is demand for electric scooters growing or declining?

Taxonomy Gap Agent

Where are the gaps in your catalog?

The agent compares available demand signals with your current product coverage and returns candidate gaps: relevant buyer language or product territory without a current catalog match. Each result includes search volume, CPC, competition context, and product-cluster coverage for review. A gap is evidence to investigate, not a prediction that a product will sell.

$1 per question on Marketplace

How to use

What to ask: Name the category you sell in or want to expand into. "Kitchen knives" gives the full gap analysis. "Which subcategories of yoga equipment have demand but low competition" narrows to candidate expansion areas.
What you get: A ranked list of gaps scored by opportunity. Each gap shows the keyword, monthly search volume, CPC, competition level, and how many product clusters already serve that demand.
What to do with it: Review demand, competition, relevance, and existing product coverage together. CPC provides advertising-cost context; actual campaign cost depends on your auction and campaign.

Example 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 Research Agent

What do comparable listings show, and how have their prices changed?

This agent returns individual product listings, their observed prices, and available price history. Review the products and dated observations alongside brand, materials, pack sizes, and your own economics. It does not calculate a cluster price range or tell you that your price is too high or too low.

$1 per question on Marketplace

How to use

What to ask: Name a product or category to inspect, then ask for the actual listings or the tracked history of one product. Include product differences that matter to your comparison.
What you get: Individual product rows, observed prices, and available price history, with citations to the evidence.
What to do with it: Review the individual listings and their dates before deciding whether a price should change. For a separate sample of prices from shared title words, use Store Checker.

Example questions:

Show me comparable kitchen knife listings and their observed prices.
Show the tracked price history for this santoku knife.
Which product differences should I review before comparing these listings?

Campaign Architect Agent

Draft a Google Ads plan.

This agent turns commerce-map evidence into a campaign draft for merchant review: ad groups with keyword lists, match types, bid ranges, budget allocation, and negative keyword suggestions. Every keyword comes with search volume and CPC from the commerce map. The output is formatted for direct import into Google Ads Editor. Draft mode only. Nothing is live until you review and approve.

$2 per question on Marketplace

How to use

What to ask: Name the products you want to advertise. "Insulated water bottles" gives a full campaign structure. Include your budget ("$500/month") for a more tailored plan.
What you get: A draft campaign plan: ad groups by theme, keyword lists with search volume and CPC, bid ranges (low/high), budget allocation weighted by opportunity, and negative keyword suggestions. Formatted for Google Ads Editor import.
What to do with it: Review the draft before export. Check opportunity, product fit, competition, CPC, budget, and your economics before choosing which ad group to launch first.

Example questions:

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

Answers With Receipts

If a number cannot be traced to the data, you do not see it. Agent answers carry evidence citations and data-through dates. Numeric claims are checked against the data that produced them before the response is delivered. If a number cannot be grounded, the answer is rejected and run again.

Your team already has a chat window. The Reach Agent is in it.

The Reach Agent is registered in Google Gemini Enterprise. Ask it a market question where your team already works. It runs its specialists across the commerce map and answers with citations and interactive cards you can act on: tables, catalog pickers, one-tap follow-ups. No new tool to learn and no new tab to keep open. The same agent, the same map, inside Gemini.

Agents in front. Reports behind.

Each agent answers in conversation, then offers more when the data warrants. When a question crosses a depth threshold (hundreds of keywords queried, a multi-dimension analysis, or language signaling a request for the complete picture), the agent surfaces the matching Lens Intelligence Report as the next step.

The agents are the entry point. The reports are the deliverable a merchant takes to their team or executes on Monday morning. Seventeen reports are available, generated by a four-pass analysis pipeline, priced from $10 to $45 per report. One offering, two surfaces.

Technical Reference

How the platform is accessed and integrated.

Access methods

GCP Marketplace console

The customer subscribes on Google Cloud Marketplace and lands on console.reach.dog through account linking. Agents are accessed via a chat surface in the browser. No install. No SDK. No local dependencies.

A2A JSON-RPC API

For API Access subscribers. Direct calls to https://rd-gap-agent-64739540958.us-east1.run.app/a2a/{agent_name} with API key authentication. Protocol: A2A JSON-RPC 2.0.

Agent Cards

Agent discovery follows the standard A2A specification. Cards are served at /.well-known/agent-card.json and /.well-known/agents.json on the agent service. Schema version: A2A v0.2.x / v0.3.

Google Cloud Marketplace plans and pricing

The Marketplace listing has two plans. Pay As You Go is visible to all customers. API Access is sales-led, activated through a private offer.

Feature	Pay As You Go	API Access
Reach Agent + 5 specialist agents (Intent, Demand, Gap, Pricing, Campaign)	Yes	Yes
17 Lens Intelligence Reports (8 to 20 page consulting-grade reports)	Yes	Yes
Reach Agent (multi-domain question routing)	Yes	Yes
Catalog Onboarding (URL scan or CSV upload)	Yes	Yes
Daily Free Recharge (10,000 tokens per 24 hours)	Yes	Yes
API Key Access (programmatic A2A JSON-RPC integration)	No	Yes
Dedicated Support (priority support channel)	No	Yes
Monthly base fee	None	$10,000 / month, no contract

Lens Intelligence Reports

Seventeen reports cover the analytical surface of the platform: gap analysis, pricing intelligence, ghost product recovery, demand mapping, audience segmentation, content calendars, PPC campaign architecture, and more. Each report is generated by a four-pass pipeline that produces an executive summary, a detailed analysis with exhibits, a prioritized action plan, and an appendix. Page count ranges from 8 to 20 pages depending on report type.

For Marketplace billing, the seventeen reports are grouped into four metrics by price point. Internal accounting tracks the per-report cost; the customer invoice reports the bucketed metric.

Marketplace metric	Reports included	Price per report
Catalog Brief	Catalog Brief (orientation report)	$10
Intelligence Report - Standard	Voice Search, Demand Map, Market Audience, Modality Strategy	$20
Intelligence Report - Advanced	Ghost Products, Product Scorecard, Audience & Content, Market Overview, Market Gaps, Pricing Intelligence, Competitive Landscape, Competitor Pricing, Cannibalization, Content Calendar	$30 (blended)
Intelligence Report - Strategic	Opportunities, PPC Campaign Intelligence	$45

The commerce map

The commerce map backing the agents was built from more than 360 million product listings from the open market. Listings are organized into 29.6 million semantic product neighborhoods that group products by what they actually are, not by how they are categorized in any single merchant taxonomy.

On the buyer side, the map includes 101.3 million analyzed keywords and 70.5 million natural-language buyer questions, each analyzed for the meaning they carry. The agents query this map in real time through BigQuery and return responses with citations to the underlying rows. Every claim in an agent response is traceable to the data row that produced it.

Integration

A2A JSON-RPC 2.0 protocol for all agent calls
Agent Cards published at /.well-known/agent-card.json and /.well-known/agents.json
API key issued at API Access subscription time, scoped per customer
Tier-based rate limits apply per plan
Hosted on Google Cloud Run in us-east1
Capacity Drawdown program eligible (customers can pay from committed Google Cloud spend)

Contact

Usage documentation: Using the Reach Agent. Support and product questions: support@reach.dog.

API Access is sales-led and activated through a private offer on Marketplace. To request a private offer, email support@reach.dog with your Google Cloud billing account and expected monthly volume.

Turn a supported question into a draft.

In the merchant workspace, choose catalog products, a buyer-question context, a channel, and a saved writing voice. Generate a draft, then review its facts before using it.

See the hosted content workflow or read a finished example.

Start smaller with a free title check or a one-product preview and emailed How-to-Sell guide. Compare the choices. Hosted merchant work uses tokens; the per-question prices above describe Marketplace usage. See merchant-app pricing.