How do I let an AI agent sendgrid send email?

Connect SendGrid to CybrLink, then expose the `sendgrid_send_email` tool to any MCP-capable agent (Claude, GPT, or a custom framework). The agent calls it with typed arguments; CybrLink injects the SendGrid credential server-side and runs one scoped API call. Send a transactional email via the connected SendGrid account. Use only after explicit user approval. Provide at least one of `text` or `html`. SendGrid returns HTTP 202 with an empty body on success.

Does the AI model see my SendGrid credentials when using SENDGRID_SEND_EMAIL?

No. CybrLink stores the credential in a token vault and injects it on the server for each call, so the model can take this action in SendGrid without the OAuth token or API key ever entering its context.

SendGrid · Marketing

`SENDGRID_SEND_EMAIL`

Send a transactional email via the connected SendGrid account. Use only after explicit user approval. Provide at least one of `text` or `html`. SendGrid returns HTTP 202 with an empty body on success.

Provider: SendGrid
Access: Write
Category: Marketing
Version: v1

Input schema

{
  "type": "object",
  "properties": {
    "to": {
      "type": "string",
      "description": "Recipient email address."
    },
    "from": {
      "type": "string",
      "description": "Verified sender email address."
    },
    "subject": {
      "type": "string",
      "description": "Email subject line."
    },
    "text": {
      "type": "string",
      "description": "Plain-text body (provide text or html)."
    },
    "html": {
      "type": "string",
      "description": "HTML body (provide text or html)."
    }
  },
  "required": [
    "to",
    "from",
    "subject"
  ],
  "additionalProperties": false
}

How to call it

Expose this tool to any MCP-capable agent (Claude, GPT, or your own) through CybrLink's MCP server. The agent calls sendgrid_send_email with arguments matching the schema above.

CybrLink validates the input, injects the connected SendGrid credentials server-side, and runs a single scoped provider API call through the shared proxy — the same guarded path as the generic call_api tool. The raw secret never reaches the model, and the response is trimmed to just the fields the agent needs.

Write action — gate behind explicit user approval and write scope.

Frequently asked

How do I let an AI agent sendgrid send email?: Connect SendGrid to CybrLink, then expose the `sendgrid_send_email` tool to any MCP-capable agent (Claude, GPT, or a custom framework). The agent calls it with typed arguments; CybrLink injects the SendGrid credential server-side and runs one scoped API call. Send a transactional email via the connected SendGrid account. Use only after explicit user approval. Provide at least one of `text` or `html`. SendGrid returns HTTP 202 with an empty body on success.
Does the AI model see my SendGrid credentials when using SENDGRID_SEND_EMAIL?: No. CybrLink stores the credential in a token vault and injects it on the server for each call, so the model can take this action in SendGrid without the OAuth token or API key ever entering its context.

Connect SendGrid All tools