Use case · API, MCP & Webhooks
Build real-time product safety workflows
Query SafeCart from your code or AI assistant, then react as soon as safety data changes. Use the REST API and MCP server for on-demand checks, and signed webhooks for immediate, verifiable recall and product-match events.
Product-safety automation needs both answers on demand and events at the moment they matter. SafeCart exposes its monitoring layer through a public REST API (OpenAPI 3.1), an MCP server and outbound signed webhooks. Your software or AI agent can check GTINs, search official EU Safety Gate notifications and summarise a portfolio, while your operational systems receive new recall and product-match events without polling. Every result remains grounded in official data rather than model guesswork.
What you can do with API, MCP & Webhooks
Query in natural language or code, then push safety events into your own operational workflows.
Check products against Safety Gate
Ask your assistant to check one or many GTINs/barcodes against EU Safety Gate recall data and get back which are recalled, and why (the check_product_safety tool).
Search live recall notifications
Query Safety Gate notices by keyword, brand, country, category and date range straight from the chat (the search_recalls tool).
Summarise your monitoring status
Get a plain-language summary of a portfolio's Product Safety Passports, flagged products, connected stores and recall campaigns (the get_monitoring_status tool).
Scan a storefront on demand
Point the assistant at a store URL and have it scan for recalled products via the SafeCart scanner (the scan_store tool).
React to safety events in real time
Receive match.flagged and recall.published events at your HTTPS endpoint, so takedown, seller-notification and case-management workflows can begin without polling.
Test before you depend on it
Send a webhook.test event from Account Settings, verify your signature handling and inspect the delivery log before switching on a production workflow.
How it works
The building blocks that make SafeCart callable by code and AI — and able to notify your systems in real time.
MCP server for AI assistants
Connect SafeCart to MCP-compatible clients such as Claude Desktop, Claude Code and Cursor, exposing recall, Product Safety Passport and monitoring tools your assistant can call directly.
Public REST API (OpenAPI 3.1)
Every capability is also a documented REST endpoint, so you can build the same checks into scripts, scheduled jobs and your own agents. Browse it in the API reference.
Scoped API keys
Organization- and portfolio-scoped keys with rate limits and tiered quotas keep programmatic and AI access controlled, metered and auditable.
Grounded in official data
Answers come from SafeCart's ingestion of the official EU Safety Gate (SGIG) API with validated, canonical GTIN matching — so AI output stays traceable, not invented.
HMAC-signed event webhooks
Receive versioned match.flagged and recall.published HTTPS events. Each payload is signed with HMAC-SHA256 so your system can verify that it came from SafeCart.
Reliable delivery & audit trail
Failed deliveries retry automatically with backoff. Test events, attempts, status and the latest error stay visible in the delivery log for each endpoint.
Frequently asked questions
What is the SafeCart MCP server?
It is a Model Context Protocol server that exposes SafeCart's recall and product-safety tools to AI assistants. Once connected, an assistant like Claude can check GTINs against the EU Safety Gate, search recall notifications, list or create Product Safety Passports, scan a store, and report your monitoring status — all using your scoped API key.
Which AI tools can connect to it?
Any MCP-compatible client. The server is documented for Claude Desktop, Claude Code and Cursor, and works with other clients that support standard MCP servers.
Do I need the MCP server, or can I just use the API?
Both reach the same data. Use the MCP server when you want an AI assistant to answer in natural language; use the REST API (OpenAPI 3.1) when you want to build the checks into your own code, agents or scheduled jobs. The full endpoint reference lives under Resources → API Reference.
Does the AI decide whether a product is recalled?
No. Recall matches come from SafeCart's ingestion of the official EU Safety Gate (SGIG) API with validated, canonical GTIN matching. The AI assistant queries that data and presents it — the determination is grounded in the official source, not model guesswork.
Which webhook events can I receive?
SafeCart sends match.flagged when a monitored product matches a safety notice and recall.published when a newly ingested notice matches products in your organization. You can also send webhook.test from Account Settings while building your integration.
How do I verify a SafeCart webhook?
Each delivery includes a timestamp and an HMAC-SHA256 signature calculated over the timestamp and raw request body. Verify the signature before parsing or acting on the event; the API reference includes the payload schemas and verification example.
Explore other use cases
Manufacturers
Own the primary GPSR duty for your products — safety, traceability and recall response.
View use caseImporters
Placing non-EU products on the market — verify compliance, keep records and monitor recalls.
View use caseDistributors & retailers
Reselling to EU consumers — check required markings and act on Safety Gate recalls in your catalogue.
View use caseQuery when you need answers. React when safety changes.
Use one product-safety layer across your code, AI agents and operational workflows — with scoped access, signed events and a delivery trail.