MCP ServerNo login required

Connect your AI agent to Sourceable

Sourceable runs a Model Context Protocol (MCP) server — a standard way to give an AI assistant safe access to a tool. Connect Claude, the MCP Inspector, or any MCP-compatible agent and let it scan any website for AI agent readiness.

Overview

An MCP server is like a USB port for AI agents: it exposes a small set of named tools that an assistant can call when it needs them. Sourceable's server is public and read-only — it reads public websites and never touches any customer's private data.

Score any site

Ask your agent to scan a URL and get a 0–100 agent-readiness score with fixes.

Works with any MCP client

Claude Desktop, the MCP Inspector, or any agent framework that speaks MCP.

Connection details

Point your MCP client at this endpoint using the Streamable HTTP transport. No API key or login is needed for the public scanner.

Endpointhttps://api.besourceable.com/mcp
TransportStreamable HTTP
AuthNone (public scanner)

Connect from Claude Desktop

Open your claude_desktop_config.json (Settings → Developer → Edit Config) and add the Sourceable server. The mcp-remote helper bridges Claude Desktop to the remote endpoint. Restart Claude after saving.

json
{
  "mcpServers": {
    "sourceable": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.besourceable.com/mcp"]
    }
  }
}

Once restarted, ask Claude something like "Scan stripe.com for AI agent readiness" and it will call the tool automatically.

Connect from MCP Inspector

The MCP Inspector is a visual tool to explore any MCP server. Run it, then connect to the endpoint with transport Streamable HTTP, and click List Tools.

bash
npx @modelcontextprotocol/inspector

Available tools

scan_agent_readiness

Score any public website (0–100) on how ready it is for AI agents, across five pillars, and return concrete fixes.

Inputs

  • urlrequiredThe public website URL to scan.
  • includeWebMcpoptionalAlso run the slower browser-based WebMCP probe (~1–3s extra). Defaults to false.

Returns

A readable summary plus structured data: overall score, maturity level, the five pillar scores, and the top fixes to reach the next level — plus a shareable report link and token you can re-open later with get_scan_report.

get_scan_report

Re-open a finished report using the token returned by a scan (or from a Sourceable report link), and return its scores and fixes.

Inputs

  • reportTokenrequiredThe report token from a scan result, a Sourceable report URL, or an email link.

Returns

The same scores and fixes as a fresh scan, for an already-completed report.

Example walkthrough

Prefer to test without an agent? Call the tool directly with curl. The reply streams back as server-sent events (lines beginning with data:).

bash
curl -sN https://api.besourceable.com/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "scan_agent_readiness",
      "arguments": { "url": "https://example.com" }
    }
  }'

Limits & privacy

  • The scanner reads public websites only. It cannot reach private or internal addresses.
  • Requests are rate-limited per IP address to keep the free service fair.
  • No Sourceable account or private customer data is ever exposed through these tools.

What's coming next

A secure, login-protected tier is on the roadmap — letting your agent read your own brand's AI visibility, monitors, and competitor comparisons straight from your Sourceable dashboard. Want it sooner? Tell us.