Letting agents send email: MCP for cold outreach

Cold outreach is a workflow, and workflows are exactly what language-model agents are good at. The missing piece has always been hands — a way for the agent to actually buy the domain, create the mailbox, load the list, and press send. That's what MCP gives it.

One endpoint, one key

sendzy exposes an MCP server at a single URL. Every public API is available as a tool, scoped to your organization, authenticated with the same API key the REST surface uses:

claude mcp add sendzy \
  --transport http \
  --url https://app.sendzy.dev/mcp \
  --header "X-Api-Key: sk_yourkey.yoursecret"

What an agent can actually do

• list-domains / buy-domain — check what you have, register what you need.
• create-mailbox — stand up sending identities on any domain.
• upload-contacts — load a prospect list straight from a CSV.
• create-campaign — build a multi-step sequence with delays between steps.
• send-campaign — push due sends; call it on a schedule and the sequence runs itself.
• list-mails / list-threads — read what came back, so the agent can draft replies.
• email-verifier — sanity-check an address before it's added to a list.

Why this is the interesting part

A human doing outreach spends most of their time on ceremony: formatting lists, picking senders, scheduling follow-ups, checking inboxes. An agent does all of it in the time it takes to describe the task. The human moves up to the work that actually needs judgment — the offer, the list quality, the reply.

Guardrails matter more here

An agent that can send email is an agent that can embarrass you. sendzy keeps the blast radius contained: every request is scoped to your organization, plan-level volume caps apply no matter how enthusiastically the agent calls the API, and round-robin spreading keeps any single mailbox from taking the heat. Give the agent the keys — the cage comes built in.

The full tool list and setup guide is in the docs. Bring your favorite agent.