EML, MSG & MBOX to JSON - Emails & Attachments for RAG

Turn email archives into clean JSON. Give it .eml, .msg or .mbox files (by URL, from a key-value store, or pasted) and get one row per message with headers, plain text, a thread ID, the attachment list (files saved to storage) and, if you want, RAG-ready chunks. For e-discovery, mailbox search with an LLM and CRM migrations.

Runs offline in the Actor, no mail server access or login is needed.

Input examples

{ "fileUrls": ["https://example.com/archive/inbox.mbox"] }

{ "keyValueStoreId": "<store id>", "recordKeys": ["export.mbox", "invoice.eml"], "includeChunks": true, "chunkTokens": 400 }

{ "rawContent": "From: a@example.com\nTo: b@example.com\nSubject: Hi\n\nHello", "saveAttachments": false }

Empty input {} parses a small built-in sample mailbox.

Input guide

Field	What it does
fileUrls	Direct links to .eml, .msg, .mbox files. Format is detected from content.
keyValueStoreId, recordKeys	Pick a store with email records; empty keys means every .eml/.msg/.mbox record.
rawContent	Paste a message or whole mbox.
saveAttachments, maxAttachmentMb	Store attachments in the run's key-value store (default on, 10 MB each).
includeHeaders	Add the full headers object.
includeChunks, chunkTokens, overlapTokens	Heading-aware, overlapping chunks for embeddings; each starts with subject, sender and date.
maxMessages, maxFileMb	Safety limits.

Output

Field	Meaning
type	message or error
source, messageIndex, format	File, position inside it, and eml/msg/mbox
messageId, inReplyTo, references	Threading headers
threadId, threadPosition, threadSize	Conversation grouping (built from Message-ID, In-Reply-To and References)
date, from, to, cc, bcc, replyTo, subject	Parsed envelope fields (ISO date, name + address)
text, textPreview, hasHtml, labels	Body as plain text (HTML converted), Gmail labels from X-Gmail-Labels
attachments[]	filename, contentType, size, inline, kvsKey, kvsUrl
headers, chunks[]	All headers; RAG chunks when enabled

Sample row (abridged):

{"type":"message","source":"mail-archive.mbox","format":"mbox","messageId":"<a2@example.com>","threadId":"a2f1c3f0df4a8a73","threadPosition":3,"threadSize":3,"date":"2026-09-29T08:15:00.000Z","from":{"name":"Alice Moreau","address":"alice@example.com"},"subject":"Re: Q3 budget review","attachmentCount":1,"attachments":[{"filename":"q3-budget.txt","size":93,"kvsKey":"att-1a2b3c4d5e-q3-budget.txt"}]}

Use it with

• Google Takeout or Thunderbird .mbox export -> this Actor -> vector database (use the chunks array).
• website-to-rag-incremental covers web pages; this Actor covers the mailbox side of the same RAG pipeline.

FAQ

Are threads reliable? Threads follow the standard headers. Messages with stripped Message-ID/References headers cannot be linked and each become their own thread.

Does it read PST or Outlook OST? No, only EML, MSG and MBOX.

Limits (honest)

• MSG support uses a pure-JavaScript reader; very old or encrypted MSG files may fail and become error rows. Tested on the platform with a sample Outlook MSG with an attachment.
• Reading a key-value store you pick needs the Actor's full-permissions mode.
• Very large mbox files are loaded into memory (default cap 100 MB per file).
• Token counts are estimated (characters / 4).

Price

Pay per event: message-parsed $0.0005 per message row. Example: a 10,000-message mailbox costs $5.00. Error rows are free.