About MsgLens
MsgLens is a local email parser and CLI for `.msg`, `.eml`, and `winmail.dat` files, for command-line users and AI assistants. It extracts message content and attachments without uploading files, and its MCP server lets assistants read and convert files on the machine.
Key features
- 01Parse `.msg`, `.eml`, and `winmail.dat`
- 02Extract headers, bodies, and attachments
- 03Decode compressed RTF and TNEF
- 04Convert messages to `.eml` or `.txt`
- 05Read email files through MCP
Use cases
- 01Inspect Outlook messages from the command line
- 02Save attachments from email files
- 03Convert a folder of `.msg` files
- 04Let an AI assistant read local email
MCP support
Runs an MCP server over stdio with tools to read email, save attachments, and convert `.msg` or `winmail.dat` files. Files remain local and the server makes no network requests.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
3Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| path | string | Yes | Absolute path to a .msg, .eml or winmail.dat file. |
| include_headers | boolean | No | Also return the raw internet headers. |
Details
From the project READMEmsglens-cli
Read Outlook .msg, .eml and winmail.dat files from the command line,
from Node, or from an AI assistant over MCP. Headers, the message body and
attachments, with no Outlook and nothing uploaded.
These are the same parsers that run in the browser at msglens.app.
npx msglens-cli read message.msg
Why another .msg parser
Most real Outlook mail does not store its body as HTML or plain text. In a
sample of 20 business .msg files, 17 kept the body only in
PidTagRtfCompressed: compressed RTF with the original HTML wrapped inside it
(MS-OXRTFCP and MS-OXRTFEX). A parser that reads PidTagHtml and
PidTagBody and stops returns an empty body for those messages. This one
decompresses the RTF and un-wraps the HTML, falling back through
PidTagHtml, PidTagBodyHtml, compressed RTF and PidTagBody in that order.
It also:
- detects the format from the bytes, so a renamed file still opens
- unpacks
winmail.dat(TNEF), including the attachments trapped inside it - decodes non-Latin bodies with the message's declared code page
- reads saved appointments and contacts, not only mail
CLI
Install it with npm install -g msglens-cli and the command is msglens;
or run it without installing as npx msglens-cli.
msglens read <file...> [--json] print headers and the plain-text body
msglens attachments <file...> [-o <dir>] save every attachment
msglens convert <file...> --to eml|txt [-o <dir>]
msglens mcp MCP server on stdio (see below)
read --jsonprints the full parsed message (one object for one file, an array for several). Attachment bytes are left out; names, types and sizes stay.attachmentsandconvertnever overwrite: a secondinvoice.pdfis saved asinvoice (1).pdf. Attachment names are reduced to a plain file name, so a message cannot write outside the output folder.- Several files at once work, and a file that fails is reported without stopping the rest. The exit code is 1 if any file failed.
Convert a folder of .msg files to .eml:
npx msglens-cli convert ./mail/*.msg --to eml -o ./eml
MCP server for AI assistants
msglens mcp runs a Model Context Protocol
server on stdio, so Claude, Codex, Cursor and other assistants can open email
files on your machine themselves. It has three tools:
read_email: subject, sender, recipients, date, plain-text body and the attachment list (optionally the raw headers)save_attachments: writes every attachment into a folderconvert_email: .msg or winmail.dat to .eml or .txt
The files stay on your machine; the server makes no network requests.
Claude Code
claude mcp add msglens -- npx -y msglens-cli mcp
Codex (~/.codex/config.toml)
[mcp_servers.msglens]
command = "npx"
args = ["-y", "msglens-cli", "mcp"]
Claude Desktop, Cursor and other JSON-configured clients
{
"mcpServers": {
"msglens": { "command": "npx", "args": ["-y", "msglens-cli", "mcp"] }
}
}
Then ask, for example, "summarise ~/Downloads/invoice-query.msg and save its attachments to ~/Desktop/invoice".
Message text is written by whoever sent the email, so the tools tell the assistant to treat it as data and not to follow instructions inside it.
Library
import { parseFile, parse, toEml, toText } from "msglens-cli";
const email = await parseFile("message.msg");
email.subject; // string
email.from; // { name, address } | null
email.date; // Date | null
email.bodyText; // plain text, always filled
email.body; // HTML when email.bodyKind === "html"
email.attachments; // [{ fileName, mimeType, size, content: Uint8Array, isEmbeddedMessage, ... }]
const again = await parse(bytes, "message.msg"); // from a Uint8Array or ArrayBuffer
const eml = await toEml(email); // RFC 822 text
email.body is the message's HTML exactly as the sender wrote it. It is
untrusted: sanitise it (for example with DOMPurify) before rendering it
anywhere.
An attachment with isEmbeddedMessage: true is a message attached to a
message; pass its content to parse() to read it.
Requires Node 20 or later.
Want to just look at a file?
msglens.app opens .msg, .eml and winmail.dat
in the browser, parsed on your device, with attachments and PDF export.
Licence
MIT. Microsoft and Outlook are trademarks of Microsoft Corporation; this project is not affiliated with Microsoft.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.