iMessage Conversation Analyzer

Copyright 2020-2025 Caleb Evans
Released under the MIT license

iMessage Conversation Analyzer (ICA) is a fully-typed Python library that will read the contents of an iMessage conversation via the Messages app's database on macOS. You can then gather various metrics of interest on the messages. The library also includes a CLI utility for easy use.

Much of this program was inspired by and built using findings from this blog post by Yorgos Askalidis.

Caveats

• Group chats (three or more people) are not supported at this time
• This program only runs on macOS

Installation

Open a Terminal and run the following:

pip3 install imessage-conversation-analyzer

Usage

The package includes both a Command Line API for simplicity/convenience, as well as a Python API for developers who want maximum flexibility.

Command Line API

To use ICA from the command line, run the ica command from the Terminal. The minimum required arguments are:

1. A path to an analyzer file to run, or the name of a built-in analyzer
2. The first and last name of the contact, via the -c/--contact flag
   1. If the contact has no last name on record, you can just pass the first name

Example

ica message_totals -c 'Jane Fernbrook'

The following outputs a table like:

Metric               Total
Messages             14535
Messages From Me      7289
Messages From Them    7246
Reactions             5050
Reactions From Me     3901
Reactions From Them   1149
Days Messaged          115
Days Missed              0
Days With No Reply       0

Built-in analyzers

ICA includes several built-in analyzers out of the box:

1. message_totals: a summary of message and reaction counts, by person and in total, as well as other insightful metrics
2. attachment_totals: lists count data by attachment type, including number of Spotify links shared, YouTube videos, Apple Music, etc.
3. most_frequent_emojis: count data for the top 10 most frequently used emojis across the entire conversation
4. totals_by_day: a comprehensive breakdown of message totals for every day you and the other person have been messaging in the conversation
5. transcript: a full, unedited transcript of every message, including reactions, between you and the other person (attachment files not included)
6. count_phrases: count the number of case-insensitive occurrences of any arbitrary strings across all messages in a conversation (excluding reactions)

Filtering

There are several built-in flags you can use to filter messages and attachments.

• --from-date: A start date to filter messages by (inclusive); the format must be ISO 8601-compliant, e.g. YYYY-MM-DD or YYYY-MM-DDTHH:MM:SS
• --to-date: An end date to filter messages by (exclusive); the format must be ISO 8601-compliant, e.g. YYYY-MM-DD or YYYY-MM-DDTHH:MM:SS
• --from-person: A reference to the person by whom to filter messages; accepted values can be me, them, or both (the default)

ica message_totals -c 'Jane Fernbrook' --from-date 2024-12-01 --to-date 2025-01-01 --from-person them

Other formats

You can optionally pass the -f/--format flag to output to a specific format like CSV (supported formats include csv, excel/xlsx, and markdown/md).

ica message_totals -c 'Jane Fernbrook' -f csv

ica ./my_custom_analyzer.py -c 'Jane Fernbrook' -f csv

Writing to a file

Finally, there is an optional -o/--output flag if you want to output to a specified file. ICA will do its best to infer the format from the file extension, although you could also pass --format if you have special filename requirements.

ica transcript -c 'Thomas Riverstone' -o ./my_transcript.xlsx

Python API

The Python API is much more powerful, allowing you to integrate ICA into any type of Python project that can run on macOS. All of the built-in analyzers (under the ica/analyzers directory) actually use this API.

Here's a complete example that shows how to retrieve the transcript of an entire iMessage conversation with one other person.

# get_my_transcript.py

import pandas as pd

import ica


# Export a transcript of the entire conversation
def main() -> None:
    # Allow your program to accept all the same CLI arguments as the `ica`
    # command; you can skip calling this if have other means of specifying the
    # contact name and output format; you can also add your own arguments this
    # way (see the count_phrases analyzer for an example of this)
    cli_args = ica.get_cli_parser().parse_args()
    # Retrieve the dataframes corresponding to the processed contents of the
    # database; dataframes include `messages` and `attachments`
    dfs = ica.get_dataframes(
        contact_name=cli_args.contact_name,
        timezone=cli_args.timezone,
        from_date=cli_args.from_date,
        to_date=cli_args.to_date,
        from_person=cli_args.from_person,
    )
    # Send the results to stdout (or to file) in the given format
    ica.output_results(
        pd.DataFrame(
            {
                "timestamp": dfs.messages["datetime"],
                "is_from_me": dfs.messages["is_from_me"],
                "is_reaction": dfs.messages["is_reaction"],
                # U+FFFC is the object replacement character, which appears as
                # the textual message for every attachment
                "message": dfs.messages["text"].replace(
                    r"\ufffc", "(attachment)", regex=True
                ),
            }
        ),
        # The default format (None) corresponds to the pandas default dataframe
        # table format
        format=cli_args.format,
        # When output is None (the default), ICA will print to stdout
        output=cli_args.output,
    )


if __name__ == "__main__":
    main()

You can run the above program using the ica command, or execute it directly like any other Python program.

ica ./get_my_transcript.py -c 'Thomas Riverstone'

python ./get_my_transcript.py -c 'Thomas Riverstone'

python -m get_my_transcript -c 'Thomas Riverstone'

You're not limited to writing a command line program, though! The ica.get_dataframes() function is the only function you will need in any analyzer program. But beyond that, feel free to import other modules, send your results to other processes, or whatever you need to do!

You can also import any built-in analyzer (for your own post-processing) via the ica.analyzers namespace.

import ica.analyzers.message_totals as message_totals

Errors and exceptions

• BaseAnalyzerException: the base exception class for all library-related errors and exceptions
• ContactNotFoundError: raised if the specified contact was not found
• ConversationNotFoundError: raised if the specified conversation was not found
• FormatNotSupportedError: raised if the specified format is not supported by the library

Using a specific timezone

By default, all dates and times are in the local timezone of the system on which ICA is run. If you'd like to change this, you can pass the --timezone / -t option to the CLI with an IANA timezone name.

ica totals_by_day -c 'Daniel Brightingale' -t UTC

ica totals_by_day -c 'Daniel Brightingale' -t America/New_York

The equivalent option for the Python API is the timezone parameter to ica.get_dataframes:

dfs = ica.get_dataframes(contact_name=my_contact_name, timezone='UTC')

Developer Setup

The following instructions are written for developers who want to run the package locally, or write their own analyzers.

1. Set up virtualenv

pip3 install virtualenv

virtualenv --python=python3 .virtualenv
source .virtualenv/bin/activate

2. Install project depdendencies

pip install -r requirements.txt