CLI Examples¶

Creating Mimeograms¶

Basic File Bundling¶

Bundle specific files into a mimeogram:

mimeogram create src/*.py

Bundle multiple types of files, each with its content type detected:

mimeogram create src/*.py README.rst

Recurse into subdirectories (subject to Git ignore rules):

mimeogram create --recurse-directories=True src/ tests/

By default, the mimeogram will be copied to the clipboard. To print to stdout instead:

mimeogram create --clipboard=False src/*.py

Adding Context¶

Add a message to provide context to the LLM:

mimeogram create --edit-message src/*.py

For LLM interfaces without project support, include format instructions:

mimeogram create --prepend-prompt src/*.py

Token Counting¶

Count the total number of tokens in the mimeogram, useful for estimating LLM context window usage:

mimeogram create --count-tokens=True src/*.py

Specify a tokenizer to use for counting (defaults to “tiktoken”):

mimeogram create --count-tokens=True --tokenizer=tiktoken src/*.py

Use a specific tokenizer variant, if supported by the tokenizer:

mimeogram create --count-tokens=True --tokenizer=tiktoken \
    --tokenizer-variant=o200k_base src/*.py

The token count will be logged to the console after creating the mimeogram. This feature helps you manage token usage when working with LLMs that have strict context limits.

Applying Mimeograms¶

Basic File Unpacking¶

Apply changes from clipboard (default):

mimeogram apply

Apply changes from standard input:

mimeogram apply --clipboard=False

Apply changes from file:

mimeogram apply --clipboard=False changes.mimeogram

Apply changes relative to a different base directory than the current working directory:

mimeogram apply --base-directory /path/to/project

Apply changes in non-interactive mode:

mimeogram apply --review-mode=silent /path/to/project

Interactive Review¶

By default, when running on a terminal, you will be presented with interactive review mode. For each file, you’ll see a menu like this:

src/example.py [2.5K]
Action? (a)pply, (d)iff, (e)dit, (i)gnore, (s)elect hunks, (v)iew >

Available actions:

a: Apply the proposed content as-is.
d: Show differences between current and proposed content.
e: Edit the proposed content.
i: Skip this file.
s: Interactively select which proposed changes to apply.
v: View the proposed content.

For protected paths, you’ll see a modified menu:

~/.config/sensitive.conf [1.2K] [PROTECTED]
Action? (d)iff, (i)gnore, (p)ermit changes, (v)iew >

The p option lets you override protection for that specific file.

Using the Hunk Selector¶

When using the s (select hunks) option, you’ll review each change block:

@@ -1,5 +1,7 @@
 def example():
-    return 42
+    """Example function with docstring."""
+    return 42

Apply this change? (y)es, (n)o, (v)iew >

This lets you cherry-pick specific changes within each file.

Setting Project Instructions¶

For LLM interfaces that support project-level instructions (like Claude.ai or ChatGPT), you can setup the mimeogram format prompt once and reuse it thereafter.

Copy instructions to clipboard (default behavior):

mimeogram provide-prompt

Print to stdout instead:

mimeogram provide-prompt --clipboard=False

Then paste these into your project instructions. All subsequent chats will understand mimeograms without needing to include the format instructions in each message.