Solutions · AI Agents
AI agents
that text.
Ship AI agents that send, receive, and reason over messages — with approvals, memory, and auditability built in. senderZ ships the only MCP server in the messaging market, so Claude Code and any MCP-aware agent can send iMessage and SMS with a single tool call.
How it works
One MCP install.
Then your agent can text.
The agent reads context, drafts the reply, optionally waits for approval. Compliance runs server-side either way.
01
Install the MCP server
One command: claude mcp add senderz -- npx @senderz/mcp --api-key sz_live_xxx. Or call the REST API from any agent framework.
02
Agent reasons over the conversation
The agent reads message history via list_messages, understands context, and decides what to send next.
03
Human approves (optional)
For sensitive workflows, the agent drafts a message and surfaces it for human review. The human edits if needed and sends.
04
senderZ delivers
iMessage if available, RCS or SMS fallback. Compliance checks run server-side — the agent cannot skip them.
claude mcp add senderz -- npx @senderz/mcp --api-key sz_live_xxxUse cases
Agents that earn the right to send.
Support agent
An AI agent monitors inbound messages, drafts replies using knowledge base context, and surfaces them to a human for approval. Handles tier-1 questions autonomously; escalates complex issues to a human agent with full thread context.
Scheduling agent
Customers text "I need to reschedule" and the AI agent checks calendar availability, proposes new times, and confirms the booking — all via iMessage. Human approval optional for established customers.
Sales follow-up agent
After a demo or trial signup, an AI agent sends personalized follow-up messages based on the prospect's product usage. Human-in-the-loop approval keeps the tone authentic — the agent drafts, the rep sends.
What you get
MCP, approvals, memory, audit.
Official MCP server
@senderz/mcp on npm — the only Model Context Protocol server in the messaging market. Claude Code, Cursor, or any MCP-aware agent can send messages with a single tool call.
Human-in-the-loop approvals
Configure approval gates so the agent drafts the message but a human reviews and sends. Critical for support, sales, and compliance-sensitive workflows.
Conversation memory
Agents access the full message history via the list_messages tool. Context carries across sessions — the agent knows what was said last week without re-asking.
Cost control
Set daily send caps per agent. If an agent hits the cap, further sends are blocked and a notification fires. No runaway messaging bills.
Audit trail
Every agent-initiated message is tagged with the agent ID and approval chain. Audit logs show who (or what) sent each message and whether a human approved it.
Multi-agent routing
Planned — Q4 2026Different agents can handle different conversation types. A support agent handles inbound questions; a scheduling agent handles appointment requests; a sales agent handles lead follow-up. Route by keyword, sender, or time.
FAQ
Frequently asked questions
Which AI models work with senderZ?
Any model that supports the Model Context Protocol — Claude (all versions), and any MCP-aware agent. For non-MCP models (GPT, Gemini, Llama), call the senderZ REST API directly from your agent code. The API is model-agnostic.
Can an agent send messages without human approval?
Yes, if you configure it that way. For high-trust workflows (order confirmations, OTP codes), set the agent to auto-send. For sensitive workflows (support, sales outreach), enable human-in-the-loop approval. You control the policy per agent.
How does conversation memory work?
The agent calls the list_messages MCP tool to retrieve conversation history. senderZ stores the full thread — the agent reads it at the start of each session. There is no separate memory store to configure; the message history IS the memory.
What about abuse prevention?
Three layers: daily send caps per agent, senderZ's built-in compliance suite (STOP keywords, quiet hours, warming limits), and the human-in-the-loop gate for sensitive workflows. An agent cannot bypass compliance — the rules run server-side before any message is delivered.
Can I build custom agents that use senderZ?
Yes. Use the MCP server for MCP-aware frameworks, or call the REST API from any agent framework (LangChain, CrewAI, AutoGen, custom). The @senderz/sdk TypeScript package wraps the API with full types.
Ship your first AI agent.
14-day free trial. MCP server included on all plans.