MCP-native for Cursor, Claude Code, and Windsurf

For AI Agents

Give your agent an inbox.

One MCP server, 32 tools, every model. Drop the snippet below into Claude Desktop, Cursor, or any MCP-aware agent and it can send, schedule, and manage email in natural language.

MCP Tools

Agents Supported

~90s

To First Send

Start with MCP Read the docs

MCP Server

Connect Posthawk MCP to
any AI agent in minutes.

Instead of switching between your AI assistant and the Posthawk dashboard, use natural language to send, schedule, query logs, manage contacts, and inspect deliverability — all from inside the agent.

Claude

Cursor

ChatGPT

Copilot

Windsurf

Lovable

Gemini

Tool catalog

32 tools an agent can call,
grouped by resource.

Every tool maps 1:1 to a Posthawk API endpoint. The MCP server exposes them with JSON schemas so the agent picks the right one without prompting tricks.

Emails

send-emailSend transactional email
get-emailFetch a sent email by job ID

Scheduled

list-scheduledList pending scheduled sends
get-scheduledInspect one scheduled send
cancel-scheduledCancel before it fires
reschedule-emailChange the send-at time
send-nowPromote a scheduled send to immediate

Contacts

list-contactsPage through audience contacts
get-contactRead a single contact record
create-contactAdd a contact with tags + metadata
update-contactEdit fields, tags, or metadata
delete-contactHard-delete a contact

Domains

list-domainsList sending domains in the workspace
get-domainInspect DKIM, SPF, DMARC, status
verify-domainTrigger a verification re-check

Newsletters

list-newslettersList all newsletters
get-newsletterRead newsletter settings
create-newsletterSpin up a new newsletter
update-newsletterEdit settings, branding, footer
list-subscribersPage through confirmed subscribers
add-subscriberAdd a subscriber (with DOI if required)
unsubscribe-subscriberSoft-unsubscribe a member
delete-subscriberHard-delete a subscriber
list-issuesList drafted and sent issues
get-issueRead one issue’s content + status
create-issueCompose a new issue
update-issueEdit subject, body, schedule
send-issueSend a finished issue

Templates

render-templateServer-render a template to HTML

Webhooks

list-webhooksList configured webhook endpoints
create-webhookSubscribe to email events
delete-webhookRemove a webhook endpoint

Total: 32 tools · All registered via the official MCP SDK · source on GitHub →

Integration paths

Everything an agent needs,
however it's built

Whether your agent calls tools over MCP, imports an SDK, shells out to a CLI, or hits the raw API — Posthawk meets it where it is.

MCP Server

Connect the official Posthawk MCP server to any compatible agent — Claude, Claude Code, Cursor, Windsurf, or VS Code — and send emails, manage contacts, schedule sends, and check domains through natural language. 32 tools, OAuth 2.1.

npx posthawk-mcpView MCP server

Official SDKs

First-class libraries for the languages agents write most. Type-safe clients with the same API surface — Emails, Scheduled, Contacts, and more — so generated code just works.

npm i posthawkBrowse SDKs

CLI

A single binary an agent can shell out to — send, preview, and tail logs from any terminal or CI step. One-shot inline sends, no SDK import, exit codes you can branch on.

npm i -g posthawkExplore the CLI

REST API

A clean, predictable REST API for direct tool use — send email, manage contacts, create campaigns, render templates, and configure webhooks. Stable endpoints, bearer auth.

api.posthawk.devAPI reference

LLM-ready docs

A machine-readable llms.txt and docs written to be parsed by models, not just humans. Point an agent at it and it learns the auth flow, deliverability rules, and every endpoint.

posthawk.dev/llms.txtRead llms.txt

Self-hosted ready

Set POSTHAWK_BASE_URL and the same MCP server, SDKs, and CLI point at your own instance. Agents operate your infrastructure — your data never leaves your network.

POSTHAWK_BASE_URL=…Self-hosting guide

Example Conversation

Natural language

Describe it.
The agent sends it.

No HTTP plumbing, no SDK boilerplate in the prompt. The agent reads the tool schema, picks the right call, and fills the arguments — you just say what you want.

Getting started

From zero to sending
in minutes

Connect once

One command wires Posthawk into your agent. Claude Code uses OAuth — no API key to paste. Other clients drop in an npx config with a key.

Authorize the workspace

OAuth opens your browser, you pick a workspace and click Authorize. An API key is created and scoped automatically.

Let the agent send

"Email the new signups a welcome from hello@acme.com." The agent picks the right tool, fills the args, and executes — no dashboard.

OAuth 2.1 + scoped keys

PKCE and dynamic client registration for secure remote access. Every agent acts inside one authorized workspace.

Your infra, optional

Run the same MCP server, SDKs, and CLI against a self-hosted instance. Agents never touch a third party.

One API surface

MCP, SDKs, CLI, and REST all map to the same endpoints — what the agent learns once works everywhere.

FAQ

Common questions before
you connect an agent.

No. The MCP server runs locally on the user’s machine (stdio transport) or on a host you control (HTTP transport with OAuth 2.1). The agent only sees the tool schemas and the responses to its tool calls. Your Posthawk API key never crosses to the model provider.