Feature 04 — MCP & API
Your AI agents get deliverability as a native tool.
SimplyWarmup runs as a Model Context Protocol (MCP) server. That means your AI agents — whether they're Claude, GPT-4, or custom — can query inbox health, halt campaigns, and provision inboxes directly from within the agent runtime.
{
"mcpServers": {
"simplywarmup": {
"command": "npx",
"args": ["-y", "@simplywarmup/mcp-server"],
"env": {
"SIMPLYWARMUP_API_KEY":
"sw_live_*******************"
}
}
}
}What the MCP server exposes.
These tools are available directly to any MCP-compatible agent. No webhook wiring, no polling infrastructure.
Available MCP Tools
Tools your AI agents can invoke natively via the MCP protocol.
| Tool Name | Description | Agent Use Case |
|---|---|---|
| get_inbox_health | Returns numerical health score (0–100) and status for a connected inbox | Check before campaign send |
| list_inboxes | Returns all connected inboxes for the authenticated tenant with current health and pacing stage | Fleet health monitoring |
| pause_inbox | Pauses warmup activity on a specific inbox without disconnecting it | Agent-triggered risk halt |
| resume_inbox | Resumes warmup activity on a paused inbox | Auto-resume after cooldown |
| get_warmup_config | Returns current pacing configuration, target volume, and reply rate for an inbox | Inspect warmup parameters |
| update_warmup_config | Updates pacing targets and reply rate for an inbox | Dynamic pacing adjustment |
The Safety Interlock pattern.
The most common use case for the MCP integration is an AI sales agent querying inbox health before executing a large outbound campaign. If reputation has dipped — perhaps due to a provider issue or a temporary blacklist — the agent halts, triggers additional warmup, and retries.
This pattern protects domains from permanent blacklisting during high-volume send events. In a fully autonomous AI outbound stack, it's the circuit breaker that prevents a cascade failure from destroying your entire sending infrastructure.
// Before each campaign send cycle
async function runCampaignSafely(inboxId: string) {
const health = await mcp.get_inbox_health({ inboxId });
if (health.score < 90) {
// Reputation at risk — do not send
await mcp.pause_inbox({ inboxId });
await notifyTeam(`Score: ${health.score} — paused`);
return;
}
// Health confirmed — proceed with campaign
await sendCampaignEmails(inboxId);
}REST API for non-MCP workflows.
All MCP tools are also available as authenticated REST endpoints for teams not yet using an MCP-compatible agent runtime.
FAQ
Common questions about MCP integration
What is MCP and why does it matter for email infrastructure?
MCP (Model Context Protocol) is an open standard that lets AI language models communicate with external tools and services in a structured, authenticated way. For email infrastructure, this means an AI agent running outbound campaigns can directly query inbox health, pause campaigns when reputation drops, and resume when health recovers — all without human intervention. SimplyWarmup exposes a native MCP server so AI agents have first-class access to deliverability data.
Can AI agents safely run autonomous cold email campaigns?
Yes — when they have access to real-time inbox health data. The risk in autonomous email sending is not the AI itself but the absence of circuit-breakers. If an agent continues sending from an at-risk inbox, it can irreversibly burn a domain. SimplyWarmup's MCP tools expose health scores and reputation status so agents can implement pause and resume logic before damage occurs. The get_inbox_health check should be a prerequisite in any autonomous outbound workflow.
Is it safe to give an AI agent write access to warmup controls?
SimplyWarmup's API uses bearer token authentication over HTTPS. All write operations — pause, resume, config updates — are scoped per inbox and logged. The MCP server only controls warmup state and surfaces health metrics; it never has access to the contents of your cold email campaigns. We recommend read-only tokens for monitoring agents and write-scoped tokens only for automation agents that implement explicit pause/resume logic.
Does SimplyWarmup support the REST API independently of MCP?
Yes. Every capability available through the MCP server is also available through the REST API. If your automation stack doesn't support MCP yet, you can call the JSON REST endpoints directly with bearer token authentication. The same endpoints power the dashboard, so every feature you see in the UI is accessible programmatically.
MCP server endpoint: sse+mcp://api.simplywarmup.com/mcp — Compatible with Claude, GPT-4o, Gemini, and any MCP-compliant AI runtime. Requires Bearer token auth on the Authorization header.