Skip to main content
Trackly’s Model Context Protocol (MCP) integration lets an AI assistant propose SMS messages, show them to a person for approval, and execute the approved send. The assistant can also check sending lists, inspect delivery and replies, and pause its API key. Use this integration when you want a team member’s assistant to prepare a customer update or a reviewed message batch. You remain responsible for recipient consent, sender setup, and reviewing the proposed content.

​
From a request to a reviewed send

  1. Identify the account. The assistant checks the account, sandbox mode, and key policy.
  2. Check the message. It selects an eligible sending list and runs preflight.
  3. Hold the draft. Trackly stores the recipients, rendered message bodies, and link destinations for review. Nothing sends at this step.
  4. Approve in Trackly. A person with send permission reviews the full content in Settings → Send Approvals.
  5. Execute and read back. The assistant executes the approved hold, then checks the stored result. Live delivery and replies are read separately from message records.
Opening an approval link or accepting an AI client’s confirmation does not approve the send in Trackly. Trackly checks the stored decision before execution.

​
Choose your starting point

Connect your AI assistant

Connect through your client’s sign-in flow where enabled, or configure a private API key.

Send with human approval

Follow a complete tool workflow and recover safely from an uncertain result.

New customer setup

Create your account, verify ownership in Trackly, and prepare the first approved sandbox message.

SMS v2 API reference

Build your own integration with the same identity, preflight, and pending-send endpoints.

​
Controls that stay with your account

Create a separate key for your assistant and start in sandbox mode. An owner can configure allowed sending lists, a daily live-send cap, and Require send approval in Settings → API Keys. A paused key cannot authenticate or resume itself; an owner reviews and resumes it in Trackly. The MCP tools expose held SMS sends, not direct or raw sending. Enable confirmation on the key as well, so that using the same credential outside MCP cannot bypass the required SMS approval flow. Confirmation-required keys also refuse direct email, OTP, and schedule submissions; those operations do not have an MCP approval workflow. These controls complement your consent and compliance practices. They do not establish recipient consent or guarantee delivery. A preflight result is advisory, and a successful live execution means acceptance for processing rather than delivery to a handset.

​
Client support and availability

Trackly provides MCP over Streamable HTTP or local stdio, a portable agent skill, and the SMS v2 API for custom agents. This is intended for any compatible harness. Credential storage, OAuth support, and approval-link handling differ between clients; the connection guide explains the required checks. An assistant can discover Trackly through search, a supplied documentation link, an installed skill, or a configured MCP connection. Registry and harness catalog listings are additional distribution channels. None guarantees that every assistant will automatically recommend Trackly when asked about SMS. Installing a skill also does not authorize an account or install an MCP connection. Hosted OAuth account connection requires a verified deployment and explicit rollout. The proposed hosted address is https://mcp.tracklysms.com/mcp; use it only after Trackly support confirms availability. Repository examples and prepared registry descriptors do not establish a live endpoint, accepted catalog listing, or completed end-to-end client test. This integration connects an external assistant to SMS tools. Trackly’s built-in AI Agents & CRM feature is a separate product for managing conversational agents.