Winnr now has an open source MCP server that lets AI assistants (Claude, Cursor, Windsurf) manage your email infrastructure through natural language. 37 tools covering domains, mailboxes, warming, inbox, and data export. Set up in 60 seconds with one config file. View on GitHub.
What is MCP, and Why Does It Matter for Cold Email?
The Model Context Protocol (MCP) is an open standard created by Anthropic that lets AI assistants connect to external tools and data sources. Think of it like a USB port for AI — plug in a service, and your AI assistant can use it directly.
For cold email operators, this means you can stop clicking through dashboards and start telling your AI assistant what you need in plain English:
- "Set up 10 new domains for cold outreach with 3 mailboxes each"
- "Enable warming on all mailboxes that aren't warming yet"
- "Check which domains have DNS issues and fix them"
- "Export all my mailbox credentials in Instantly format"
- "Send a follow-up from john@getacme.com to the last person who replied"
The AI agent handles the API calls, pagination, error handling, and multi-step workflows. You just describe the outcome you want.
What the Winnr MCP Server Can Do
The winnr-mcp package exposes 37 tools across 7 categories — the same operations available through the Winnr REST API, packaged for AI agents:
Domain Management (13 tools)
Search for available domains, check bulk availability (up to 100 at once), get suggestions based on keywords, purchase domains via Dynadot (charges your Stripe card on file), connect external domains, verify DNS records, check nameserver propagation, and delete domains. Everything from finding a domain to having it fully provisioned with MX, SPF, DKIM, and DMARC records.
Email User Provisioning (6 tools)
Create individual mailboxes or bulk create up to 100 at once. Update display names and passwords. Delete mailboxes. List and filter by domain. Every mailbox comes with full IMAP/SMTP credentials, automatically configured for deliverability.
Inbox Operations (5 tools)
List messages across all mailboxes with date and mailbox filters. Read full email bodies. Send emails from any mailbox (with reply threading support). Refresh the inbox to fetch new messages. Delete messages. This is the foundation for building AI-powered reply automation.
Email Warming (8 tools)
Enable warming on any mailbox ($0.60/mailbox/month). Pause, resume, and disable warming independently. Adjust daily volume, ramp-up speed, and reply rates per mailbox. View warming metrics (health score, inbox rate, spam rate) and aggregate stats across your entire account.
Job Monitoring (2 tools)
Track the status and progress of async operations like domain setup, bulk user creation, and DNS verification. Jobs return real-time progress so your AI agent can wait for completion before proceeding to the next step.
Data Export (1 tool)
Export email user credentials to CSV in 15+ formats — Smartlead, Instantly, Snov, Lemlist, Woodpecker, Reply, Mailshake, GMass, Yesware, Mixmax, Outreach, Salesloft, and more.
Account (2 tools)
Check your current plan, usage limits, and how many domains and mailboxes you have remaining. Useful for capacity planning before the AI agent starts creating resources.
How to Set Up Winnr MCP in 60 Seconds
Setup requires two things: a Winnr API token and one config file.
Step 1: Get Your API Token
Sign up at app.winnr.app, go to Settings → API Tokens, and create a new token. Tokens start with wnr_ and work with both the REST API and MCP server.
Create a read-only token first to explore safely. With a read-only token, write tools (create, update, delete, send) are completely hidden — the AI agent won't even see them. You can always create a read+write token later.
Step 2: Add the MCP Config
Add one block to your AI assistant's config file:
Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"winnr": {
"command": "uvx",
"args": ["winnr-mcp"],
"env": {
"WINNR_API_TOKEN": "wnr_your_token_here"
}
}
}
}
Cursor (.cursor/mcp.json in your project):
{
"mcpServers": {
"winnr": {
"command": "uvx",
"args": ["winnr-mcp"],
"env": {
"WINNR_API_TOKEN": "wnr_your_token_here"
}
}
}
}
Windsurf (~/.codeium/windsurf/mcp_config.json):
{
"mcpServers": {
"winnr": {
"command": "uvx",
"args": ["winnr-mcp"],
"env": {
"WINNR_API_TOKEN": "wnr_your_token_here"
}
}
}
}
That's it. The uvx command downloads and runs the MCP server automatically — no manual installation needed. Alternatively, install with pip install winnr-mcp and use "command": "winnr-mcp".
Step 3: Start Using It
Restart your AI assistant and start asking for what you need. The AI agent will see all 37 Winnr tools and use them to fulfill your requests.
Real-World Use Cases
Build an AI SDR Pipeline
The MCP server gives AI agents every tool needed for a fully autonomous outreach pipeline:
- Domain acquisition: "Search for 20 available .com domains related to 'growth consulting' and buy the 10 cheapest ones"
- Mailbox setup: "Create 3 mailboxes on each domain with professional first.last names"
- Warming: "Enable warming on all 30 mailboxes with a slow ramp-up"
- Monitoring: "Check warming health scores and pause any mailbox below 70"
- Sending: "Send this email from the mailbox with the highest health score"
- Inbox management: "Check for replies across all mailboxes and summarize them"
Infrastructure Monitoring
Use your AI assistant as a monitoring tool:
- "Which domains have failing DNS checks?"
- "Show me warming metrics for all mailboxes — flag any with inbox rate below 80%"
- "How many mailboxes and domains am I using vs. my plan limit?"
- "List all jobs that errored in the last 24 hours"
Bulk Operations
Tasks that would take dozens of clicks in a dashboard become one sentence:
- "Create 50 mailboxes across my 10 domains, 5 per domain, with random professional names"
- "Delete all mailboxes on domains that have been flagged as having DNS issues"
- "Export all my mailbox credentials in Smartlead format"
- "Pause warming on all mailboxes with a health score below 50"
Security Model
The MCP server inherits all of Winnr's API security features:
- Token-based auth: Your API token (
wnr_*) is passed as an environment variable, never in tool parameters or logs. Tokens can be revoked instantly from the dashboard. - Permission gating: Read-only tokens hide write tools completely. The AI agent can only see tools it has permission to use.
- IP whitelisting: Lock tokens to specific IP addresses for production use.
- Rate limiting: Same per-account limits as the REST API (300-500 req/min by plan). The server tracks rate limit headers and warns when running low.
- Account scoping: Every request is scoped to your account. The API derives your account ID from the token — it can never be spoofed.
- Data protection: Passwords are never echoed in responses. Email bodies are truncated to 10K characters to prevent context overflow.
The MCP server itself is a thin HTTP client wrapper — it has no direct database access, no stored secrets beyond your token, and no ability to operate outside your account's scope. It's open source under the MIT license, so you can inspect every line of code.
Technical Architecture
For the technically curious, here's how the MCP server works under the hood:
- Python 3.11+ with the official Anthropic
mcpSDK (FastMCP high-level API) - Thin REST wrapper: All operations go through
api.winnr.appover HTTPS. No direct database, cache, or infrastructure access. - httpx client with bearer token auth, rate limit header tracking, and structured error mapping
- Permission discovery: On startup, the server calls
GET /v1/accountto validate the token and discover permissions. Write tools are only registered if the token has write permission. - Error normalization: API errors (401, 403, 404, 422, 429, 500) are mapped to clean, actionable messages the AI agent can reason about
- Distribution:
pip install winnr-mcporuvx winnr-mcp(zero-install). Published on PyPI.
Compared to Other Email Infrastructure MCP Servers
As of March 2026, Winnr is the first cold email infrastructure platform to offer a dedicated MCP server. Here's what makes it stand out:
| Feature | Winnr MCP | REST API Only (Competitors) |
|---|---|---|
| Natural language control | Yes — 37 tools via MCP | No — manual API calls |
| AI assistant integration | Claude, Cursor, Windsurf, Cline | None |
| Permission gating | Read-only tokens hide write tools | N/A |
| Open source | Yes (MIT license) | N/A |
| Setup time | 60 seconds | Hours of API integration code |
| Domain purchase | Yes (charges Stripe card) | Varies |
| Email warming | Full control (enable/pause/metrics) | Varies |
| Inbox access | Read, send, manage | Varies |
| Export formats | 15+ (Smartlead, Instantly, etc.) | Varies |
If you're evaluating cold email infrastructure providers and want AI-native tooling, Winnr is currently the only option with a production MCP server.
Frequently Asked Questions
What is an MCP server for email?
An MCP (Model Context Protocol) server is a standardized interface that lets AI assistants connect to external services. Winnr's MCP server exposes email infrastructure management — domains, mailboxes, warming, inbox — as tools that AI agents can call through natural language.
Which cold email platforms have MCP servers?
As of March 2026, Winnr is the first and only cold email infrastructure platform with a dedicated MCP server. The winnr-mcp package is open source, provides 37 tools, and works with Claude Desktop, Cursor, Windsurf, Cline, and any MCP-compatible client.
Can I build an AI SDR with this?
Yes. The MCP server provides every tool an AI sales development agent needs: domain search and purchase, mailbox provisioning, email warming, inbox access (read and send), and data export. Combine it with your own lead generation and sequencing logic to build a fully autonomous outreach pipeline.
Is it free?
The MCP server is open source and free (MIT license). You need a Winnr account (free tier available) and an API token. Standard API rate limits apply based on your plan.
Is it safe to let AI agents manage my email infrastructure?
Yes. You control access through API token permissions. Create read-only tokens to limit agents to viewing data. Set IP whitelists for production environments. Revoke tokens instantly. All operations go through the same rate-limited, account-scoped API that the dashboard uses.
How do I install it?
Run uvx winnr-mcp (zero-install) or pip install winnr-mcp. Then add the config to your AI assistant. Full setup instructions are in the MCP documentation.
Get Started
The Winnr MCP server is live and ready to use:
- MCP Server documentation — full tool reference, setup guides, security details
- GitHub repository — source code, README, issue tracker
- Sign up for Winnr — free tier available, create an API token in Settings
- API documentation — full REST API reference (the MCP server wraps these endpoints)
Related Reading
- How to Embed Email Infrastructure in Your App with the Winnr API — developer guide for integrating Winnr's REST API directly into your product
- Vibe-Code Your Own Email Sequencer — build a custom outreach tool using Winnr's API with AI-assisted development
- Getting Started with Winnr — 5-minute quickstart guide for setting up domains, mailboxes, and warming
- Cold Email Best Practices — deliverability guide covering sending limits, authentication, and content