TapSafely

For developers & AI assistants

Bring scam checks into your app or assistant

The same engine that powers our checker is available as a simple JSON API and as an MCP server, so AI assistants can check suspicious messages for their users.

MCP server

Endpoint (Streamable HTTP, stateless): https://tapsafely.com/mcp

Tools: check_message, check_link, scam_trends, scam_guide. Example client configuration:

{"mcpServers": {"tapsafely": {"type": "http", "url": "https://tapsafely.com/mcp"}}}

REST API

POST https://tapsafely.com/api/check with a JSON body:

{ "kind": "text", "text": "Your parcel is on hold, pay £1.45 at …" }
{ "kind": "url",  "url": "paypa1-login.top" }

Response (shortened):

{ "score": 86, "level": { "id": "high", "label": "High risk" },
  "category": { "id": "delivery", "guide": "/scams/fake-delivery-texts" },
  "flags": [ { "label": "Small fee requested", "severity": "medium" } ],
  "advice": [ "Don't reply, tap links…" ], "links": [ … ] }

Limits and keys

  • Without a key: a few requests per minute per IP, rule-based and link checks only.
  • With a key (header X-Api-Key): higher limits and AI review. Ask us for a key.
  • Please show results calmly, never describe anything as definitely “safe”, and link to our guides.

A2A agent

Other AI agents can talk to our assistant over the Agent2Agent protocol: JSON-RPC SendMessage (A2A 1.0) or message/send (0.3) at https://tapsafely.com/a2a. The agent card is at /.well-known/agent-card.json.

{"jsonrpc":"2.0","id":1,"method":"message/send","params":{"message":{"role":"user","messageId":"1","parts":[{"kind":"text","text":"Is this a scam? ..."}]}}}

In-page tools (WebMCP)

Browsers and agents that support WebMCP can use check_message_for_scam, find_scam_guide, where_to_report_scam and open_scam_checker directly on any page.

Discovery files