Product Documentation

Overview

AgenticRadar is a Shopify application that makes your store discoverable and queryable in the era of agentic commerce, where AI shopping assistants discover, evaluate, and purchase products on behalf of consumers.

AgenticRadar optimises and monitors your Shopify store for AI shopping agents: Shopify Catalog, UCP, and AI answers. It scores catalog fields, checks the store's native UCP profile and catalog MCP, and can publish agents.md plus llms.txt on your domain. The FFP content audit remains available. Theme publishing is optional. Use Restore Shopify default before you uninstall. Uninstall cannot delete templates/agents.md.liquid or templates/llms.txt.liquid, so remove those two files in the theme editor if they are still present. Shopify then serves its managed agent files.

• Available on Shopify App Store
• MCP Server endpoint: https://api.agenticradar.radrly.com/server/mcp
• 12 AI agent tools across 4 categories (catalog, cart, orders, customer)
• AgenticScore 0 to 100 composite readiness metric
• FFP Content Audit (Findability · Freshness · Precision) with AI bulk fix
• Legacy UCP Manifest (agenticradar.json) plus llms.txt
• Agent Interaction Log with JSON export; records are kept for 90 days
• Shopify Admin GraphQL 2026-04 + Storefront API

Quick start (5 minutes)

1. Install & onboard: Install from the Shopify App Store and complete the onboarding checklist. This alone adds +25 pts to your AgenticScore.
2. Create an agent token: Go to MCP Server → Create token. Copy the raw token (shown once). This is your Bearer key for AI agent authentication.
3. Publish the UCP Manifest: Go to UCP Manifest, toggle on the capabilities you want to expose, and click Publish. Adds +20 pts to AgenticScore.
4. Run a Content Audit: Go to Content Audit → Scan Products. Review FFP scores and use AI Fix or Bulk Fix to fill content gaps.
5. Connect your first AI agent: Point Claude, GPT, n8n or any MCP-compatible agent at https://api.agenticradar.radrly.com/server/mcp with your Bearer token and call tools/list to discover all 12 tools.

Who is it for?

• Shopify merchants who want their products discovered and purchased by AI shopping agents
• Store operators preparing for agentic commerce (AI agents acting on behalf of consumers)
• Technical store owners who want a live MCP endpoint connected to their Shopify data
• Brands concerned about product content quality and AI discoverability scores
• Merchants who want to see which AI agents call their store and what those agents asked

Problems it solves

1. Agentic Commerce Gap

AI shopping agents (ChatGPT Shopping, Perplexity, Claude) cannot interact with stores that only expose HTML pages. AgenticRadar adds a standards-based MCP Server with 12 tools, a UCP Manifest for auto-discovery, and an llms.txt file, making the store natively agent-navigable without custom development.

2. Content Invisibility

AI agents rank and recommend products based on structured metadata. Most stores have gaps: missing tags, thin descriptions, absent product types. AgenticRadar's FFP Content Audit identifies every gap and generates AI-powered fixes in bulk.

3. No visibility into agent activity

Merchants usually cannot see what AI agents ask about their store. AgenticRadar records every MCP tool call in the Agent Interaction Log, shows usage KPIs and lets you export the records as JSON.

AgenticScore

AgenticScore is a composite 0 to 100 metric that measures how ready your store is for agentic commerce. It is recalculated on every dashboard load and persisted to your store record.

Score formula

Score thresholds

• 70 to 100 Excellent (green): Store is agent-ready; AI agents can fully interact with your catalogue.
• 40 to 69 Improving (amber): Partial readiness; key gaps exist (e.g., no manifest or low FFP scores).
• 0 to 39 Needs Work (red): Store is not yet discoverable by AI agents.

MCP Server & AI Agent Tools

AgenticRadar runs a live JSON-RPC 2.0 MCP (Model Context Protocol) server that any AI agent can connect to. All tool calls are authenticated with per-store Bearer tokens and recorded in the Agent Interaction Log.

Available tools (12)

Catalog & Discovery

• check_stock: Check real-time inventory for a product (by handle). Returns variant-level availability.
• get_product_context: Full product details: title, description, price, images, tags, variants.
• search_products: Full-text search across the product catalogue with keyword filters.
• get_related_products: Retrieve related or cross-sell products for a given product handle.

Cart & Checkout

• create_cart: Create a Shopify cart with specified line items and return a checkout URL. Requires Storefront API token.
• get_shipping_estimate: Estimate shipping costs and delivery time to a destination country.

Orders & Returns

• get_order_status: Retrieve order status, fulfillment, and tracking by order ID or email.
• get_return_eligibility: check whether items in an order are eligible for return.
• get_store_policies: Return store policies (returns, shipping, privacy) as structured text.

Customer & Loyalty

• get_customer_history: Fetch a customer's order history and lifetime value by email.
• get_loyalty_balance: Return the current loyalty points balance for a customer.
• apply_loyalty_discount: Apply a discount code or loyalty redemption to a cart.

How to connect

1. Create a token: Go to MCP Server in the dashboard → Create token. Set a label and optional expiry. Copy the raw token (shown once only).
2. Configure your agent: Point it to the MCP endpoint above using HTTP/SSE or JSON-RPC transport. Pass the token as Authorization: Bearer <token>.
3. Discover tools: Send a tools/list request. The server returns all 12 tools with full JSON Schema definitions for inputs and outputs.
4. Call tools: Use tools/call with the tool name and arguments. Send Mcp-Session-Id in the header for session tracking.

Agent Token Management

Tokens are scoped to mcp:read by default. You can create multiple tokens for different integrations (e.g., one for Claude, one for a custom bot) and revoke them individually. Token hashes are stored, the raw token is shown only once at creation.

UCP Manifest

The AgenticRadar UCP Manifest (agenticradar.json) is a machine-readable JSON file that describes your store's agent-accessible capabilities, so AI systems can discover the available tools and data endpoints without manual configuration. It is kept for compatibility: Shopify now serves a native UCP profile at /.well-known/ucp on your storefront, which the App checks for you.

Manifest capabilities (toggleable)

• tools: MCP tool names exposed to agents
• mcp_server_url: URL of the MCP server
• llms_txt_url: link to your store's llms.txt file
• shop: name, domain, currency, country and language, for agent context
• disclosure: tells agents that interactions are logged and that records are kept for 90 days

Publishing the manifest adds +20 pts to AgenticScore. Merchants control exactly which capabilities are visible to external agents via toggle switches in the UCP Manifest dashboard page.

Content Audit, FFP Score

The Content Audit scans your Shopify product catalogue and scores every product on three dimensions that determine how well AI agents and search engines can discover and understand them.

FFP Score = (Findability + Freshness + Precision) / 3

• Findability (F, 0 to 100): Does the product have a meaningful title (≥5 chars), product_type, and tags? Deductions: no tags (−30), no product_type (−30), weak title (−40).
• Freshness (F, 0 to 100): Is there a description, and is it long enough for agents to understand the product? Deductions: absent or <50 chars (−60), thin <150 chars (−30).
• Precision (P, 0 to 100): Are there images, a price, and at least one variant? Deductions: no images (−40), no price (−30), no variants (−30).

A product score above 80 is well-optimised for agentic commerce. Below 50 means an agent cannot confidently describe or recommend it.

AI Fix & Bulk Fix

For products with gaps, AgenticRadar uses GPT-4.1-nano to generate optimised values:

1. Analyze: Audit identifies all products below threshold.
2. Generate: AI writes missing body_html, tags, and product_type for each product (max 5 concurrent).
3. Review: All suggestions are shown side-by-side. You can edit any field.
4. Apply: One click pushes all accepted fixes to Shopify via the Admin API and re-runs the audit to update scores.

Average FFP across all products contributes up to +30 pts to AgenticScore.

Agent Interaction Log

AgenticRadar records every MCP tool call so you can see what AI agents asked and what your store returned. Records are kept for 90 days; once a day, older records are deleted automatically.

Fields logged per interaction

• session_id: MCP session identifier (from Mcp-Session-Id or X-Session-Id header, or request body)
• tool_name: Which of the 12 tools was called
• latency_ms: Response time in milliseconds
• resulted_in_purchase: true / false / null (updated when a matching Shopify order is created)
• request and response: the tool arguments and the data returned (this can include customer data such as email address or order details)
• error: error message if the call failed
• created_at: Timestamp

KPIs (last 30 days)

• Total agent calls
• Unique sessions
• Conversion rate: purchases / sessions × 100
• p50 response latency
• Top-called tools chart

Export

Click Export interaction log (JSON) to download the interactions from the selected period (up to the last 90 days) as a JSON file with store metadata, the total count and per-record detail (session, tool, latency, purchase outcome, error and timestamp).

llms.txt Generator

AgenticRadar generates a per-store llms.txt file, an AI-readable plain-text overview of the merchant's store. The file is served publicly at https://{your-shop}.myshopify.com/apps/agenticradar/llms.txt via Shopify App Proxy.

• Generated using GPT-4.1-nano from live store data (products, policies, brand identity)
• Regenerated on demand from the UCP Manifest page
• URL is automatically included in the published UCP Manifest (llms_txt_url field)
• Lets any LLM-based agent understand your store without querying the MCP endpoint

Data handling

What we collect

• Store details (domain, name, owner name and email, currency, country, language) and the Shopify OAuth tokens
• Product titles, handles, descriptions, tags, product type (for FFP scoring and llms.txt)
• Agent interaction records: session ID, tool name, request and response, latency, purchase outcome, timestamp
• UCP Manifest capabilities JSON and MCP configuration

Retention periods

• Agent interaction records: 90 days, deleted automatically once a day
• Shopify access tokens: deleted immediately when you uninstall the App
• All store data, including audit results and interaction records: deleted immediately when Shopify sends the shop/redact webhook (48 hours after uninstall)
• Interaction records about a specific customer: deleted when Shopify sends the customers/redact webhook

For full details, see the Privacy Policy and Data Processing Addendum (DPA).

Legal references

Support & FAQ

Contact us at hello@agenticradar.io. You can also use the AI Support Chat in the app dashboard for instant answers.

Frequently asked questions

Q: How do I connect Claude or ChatGPT to my store?

A: Create an agent token in the MCP Server page. Then configure your agent with the endpoint URL (https://api.agenticradar.radrly.com/server/mcp) and pass the token as Authorization: Bearer <token>. Send a tools/list request to discover all 12 available tools.

Q: How long are agent interaction records kept?

A: For 90 days. Once a day, AgenticRadar deletes records older than 90 days. Use Export interaction log (JSON) if you want to keep your own copy.

Q: Does create_cart require a Storefront API token?

A: Yes. The create_cart tool uses the Shopify Storefront API (not Admin API) to create customer-facing carts. You need to add your Storefront API token in Settings → MCP Configuration for this tool to work. The other 11 tools use the Admin API and work without it.

Q: How does the FFP Bulk Fix work?

A: Run a Content Audit first. Select products with low scores, click Bulk Fix → Generate. GPT-4.1-nano generates improved body_html, tags, and product_type for each product simultaneously (max 5 at once). Review all suggestions, edit any field, then Apply to push fixes to Shopify and re-audit.

Q: Can I revoke an agent token?

A: Yes. Go to MCP Server → Tokens, find the token by its label, and click Revoke. It is immediately invalidated, any agent using it will receive a 401 Unauthorized response.

Q: What data is included in the interaction log export?

A: The JSON export includes store metadata, the total interaction count and per-record detail: session_id, tool, latency_ms, resulted_in_purchase, error and timestamp. Request and response data is shown in the App but not included in the export.

Q: Is my store's product data sent to OpenAI?

A: Only when you use AI-powered features: Content Audit suggestions and Bulk Fix (we send title, description, product type and tags), llms.txt generation, the AI assistant and AI Visibility checks. Customer data from MCP tool calls is not sent to OpenAI.