Give AI clients native access to your email.
InboxParse ships a production-ready MCP server. Plug it into Claude Desktop, Cursor, or any MCP-compatible AI client and let your agent read, search, and summarise real email threads - out of the box.
{
"mcpServers": {
"inboxparse": {
"command": "npx",
"args": ["-y", "@inboxparse/mcp"],
"env": {
"INBOXPARSE_API_KEY": "ip_..."
}
}
}
}
// That's it. Claude can now call:
// - inboxparse.list_threads({ limit: 10, format: "markdown" })
// - inboxparse.get_thread({ id: "th_abc123" })
// - inboxparse.search({ q: "invoice overdue", mode: "semantic" })
// - inboxparse.get_labels({ thread_id: "th_abc123" })Drop-in npm package
One npx command. No server to deploy, no auth code to write. The MCP binary runs locally and proxies to the InboxParse API.
OAuth 2.1 support
Authenticate via OAuth 2.1 instead of raw API keys - ideal for multi-user deployments and enterprise environments.
10 scopes, zero boilerplate
Read threads, search semantically, fetch labels, list mailboxes - all exposed as typed MCP tool calls your agent can invoke directly.
Works with any MCP client
Claude Desktop, Cursor, claude.ai Projects, and any other client that speaks the Model Context Protocol.
Scoped permissions
Grant only the scopes your agent needs. Read-only, search-only, or full access - configured per API key.
Markdown-optimised output
Every email thread is returned as clean Markdown - token-efficient and ready to drop straight into your LLM context window.
Use InboxParse inside a Vercel AI SDK agent
Copy-and-paste ready. No boilerplate.
import { createMcpClient } from "@modelcontextprotocol/sdk/client/mcp.js"
import { generateText } from "ai"
import { anthropic } from "@ai-sdk/anthropic"
// Connect to InboxParse MCP server
const mcp = await createMcpClient({ command: "npx", args: ["-y", "@inboxparse/mcp"] })
const tools = await mcp.tools()
// Let the agent decide which emails to read
const { text } = await generateText({
model: anthropic("claude-3-5-sonnet-20241022"),
tools,
maxSteps: 5,
prompt: "Summarise my 5 most recent unread emails and flag any that need urgent action.",
})
console.log(text)
await mcp.close()Frequently asked questions
What is MCP and how does it work with InboxParse?+
MCP (Model Context Protocol) is a standard for connecting AI clients to external data sources. InboxParse ships a native MCP server that exposes your email data as typed tool calls - AI clients like Claude Desktop or Cursor can read, search, and label emails directly.
Which AI clients support InboxParse MCP?+
Any MCP-compatible client works, including Claude Desktop, Cursor, claude.ai Projects, and custom agents built with the Vercel AI SDK or Anthropic SDK.
Do I need to deploy a server?+
No. The MCP server runs locally via npx. One command in your MCP client config and it connects to the InboxParse API. No infrastructure to manage.
Explore more use cases
The fastest way to give Claude a mailbox.
One config line. No servers. No parsers. Just your AI agent reading real email.