ARC: agent checkout reachability

What ARC: agent checkout reachability does

ARC: agent checkout reachability provides MCP access to ARC, a hosted record of whether AI agents can buy from online stores. Results include a date and a source label, such as an agent report, an ARC test completed up to payment, or a verified purchase. The service treats missing results as unknown; lack of a record does not count against a store.

The repository contains connector metadata and setup instructions rather than an implementation that runs locally. The service is available at https://www.arcreport.ai, and its public MCP registry names are ai.arcreport/mcp for store checks and ai.arcreport/ask for questions about stores, protocols, and submissions.

How it works

Both connectors use the Streamable HTTP transport. An MCP client sends requests to the hosted endpoint, so there is no local package, process, database, or sign-in flow to maintain. The store-check connector exposes check_store, find_stores, and report_checkout as its primary operations. Reading saved checks and reporting a checkout attempt do not require sign-in.

ARC: agent checkout reachability is useful before an agent attempts a purchase, when evaluating supported commerce destinations, or when recording what happened during an automated checkout. Its evidence describes reachability and reported outcomes; it does not guarantee that a future checkout will succeed.

Setup and configuration

Add the store-check endpoint to an MCP client that supports Streamable HTTP:

https://www.arcreport.ai/api/mcp

The Ask connector uses:

https://www.arcreport.ai/api/ask

The README gives configurations for Claude Code, Claude Desktop or claude.ai, ChatGPT, Cursor, VS Code, and Cline. For example, Claude Code can register the server with claude mcp add --transport http arc https://www.arcreport.ai/api/mcp. Cursor uses an mcp.json entry with the server URL, while VS Code and Cline specify the HTTP or Streamable HTTP transport in their respective MCP configuration files.

No environment variables or local installation steps are documented. The repository also links to an A2A agent card, an OpenAPI description, llms.txt, an open dataset, and the Agent Outcome Report specification.

Tools and capabilities

• Check a named store’s saved ARC result with check_store.
• Search available store records with find_stores.
• Submit an observed checkout attempt with report_checkout.
• Use the Ask endpoint for questions involving stores, protocols, or submissions.
• Consume dated evidence with source labels for downstream agent decisions.

The open dataset is licensed CC-BY-4.0. The repository contents are licensed MIT. ARC: agent checkout reachability is therefore suited to clients that need a hosted, evidence-based lookup and reporting path, but it is not a local checkout automation framework.