MCP server

Give Claude, Cursor or any MCP client tools to create monitors and check in.

Vivere is also an MCP server. Connect it and an agent can create its own monitor, then call check_in after each piece of scheduled work; when the check-ins stop, you are alerted like for any other monitor. The endpoint is https://vivere.dev/mcp (Streamable HTTP, stateless) and it authenticates with an API key as a bearer token.

Tools

Tool	What it does
`list_monitors`	All monitors with status and last ping.
`get_monitor`	One monitor, including its ping URL.
`create_monitor`	New heartbeat (name, period or cron, grace) or HTTP check (url, interval).
`check_in`	Report `success`, `start` or `fail` for a monitor, with an optional message kept as the run log.
`pause_monitor`, `resume_monitor`	What they say.

Claude Code

claude mcp add --transport http vivere https://vivere.dev/mcp --header "Authorization: Bearer vv_…"

Then, in a scheduled task's prompt: "When you have finished, call the Vivere check_in tool for monitor <id> with a one-line summary."

Claude Desktop, Cursor, other clients

{
  "mcpServers": {
    "vivere": {
      "url": "https://vivere.dev/mcp",
      "headers": { "Authorization": "Bearer vv_…" }
    }
  }
}

Raw JSON-RPC

curl -X POST https://vivere.dev/mcp \
  -H "Authorization: Bearer vv_…" -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"check_in","arguments":{"monitor_id":"<id>","status":"success","message":"processed 14 tickets"}}}'

Notifications (requests without an id) are acknowledged with 202. There is no server-initiated stream, so GET /mcp returns 405; clients that require one fall back to plain request/response, which is all these tools need.

The API key grants full access to the account. Give an agent its own key from API keys so it can be revoked on its own.