Skip to content

CLI Reference

dumpcat [OPTIONS] [PATH]

PATH defaults to . (current directory).

Flags

Flag Short Type Default Description
--output PATH -o string stdout Write output to a file
--output-append -A flag off Append to output file instead of overwriting
--depth INT -d integer unlimited Max directory depth
--include EXT -i string (repeatable) all files Include only these extensions
--exclude PATTERN -e string (repeatable) none Exclude glob patterns
--gitignore flag on Respect .gitignore rules
--no-gitignore flag off Ignore .gitignore rules
--tree-only flag off Show only the file tree, no contents
--no-tree flag off Show only file contents, no tree
--prepend TEXT -p string none Text or @filepath to prepend
--append TEXT -a string none Text or @filepath to append
--max-size SIZE string 1mb Skip files larger than this
--stats -s flag off Show file count, lines, and estimated tokens
--format FORMAT -f choice markdown Output format: markdown, plain, or json
--config PATH string auto Path to config file
--profile NAME string default Named profile from config
--follow-symlinks flag off Follow symbolic links
--hidden flag off Include dotfiles and dotdirs
--line-numbers -n flag off Add line numbers to file contents
--version -V Show version and exit
--llm PROVIDER string none Send output to an LLM (ollama, vllm, lmstudio, or a URL)
--target NAME -t string none Named LLM target from config
--model NAME -m string none LLM model name (e.g. llama3, gpt-4o)
--system-prompt TEXT string none System prompt for the LLM
--set KEY=VALUE string (repeatable) none LLM parameter (e.g. temperature=0.7)
--api-key KEY string none API key for the endpoint

Subcommands

dumpcat init

Create a starter LLM profiles config at ~/.dumpcat/dumpcat_profiles.toml:

dumpcat init

Use --force to overwrite an existing config:

dumpcat init --force

Flag details

--output / -o and --output-append / -A

Write output to a file. By default, -o overwrites the file. Use -A to append instead.

# Overwrite (default)
dumpcat -o dump.md

# Append to an existing file
dumpcat -o dump.md -A

# Accumulate multiple dumps
dumpcat src/ -i .py -o all-code.md -A
dumpcat tests/ -i .py -o all-code.md -A

--include / -i

Filter by file extension. Repeatable. Only files matching at least one extension are included.

# Only Python and JavaScript files
dumpcat -i .py -i .js

Note

Files with no extension are excluded when --include is active.

--exclude / -e

Exclude files or directories matching glob patterns. Repeatable. Matches against file names, path components, and full relative paths.

# Exclude test files and cache directories
dumpcat -e tests -e "*.pyc" -e "__pycache__"

--prepend / --append

Add text before or after the main output. Useful for adding LLM prompts.

  • Plain string: used as-is
  • @filepath: reads the file's contents
# Inline prompt
dumpcat -p "Review this code for security issues"

# From a file
dumpcat -p @prompts/review.md -a @prompts/footer.md

--max-size

Skip files larger than the given size. Accepts human-readable sizes:

  • 500b — 500 bytes
  • 50kb — 50 kilobytes
  • 1mb — 1 megabyte (default)
  • 1gb — 1 gigabyte
dumpcat --max-size 500kb

--tree-only / --no-tree

These flags are mutually exclusive.

# Just the tree structure
dumpcat --tree-only

# Just the file contents
dumpcat --no-tree

--config / --profile

See Configuration for details on .dumpcat.toml and profiles.

# Use a specific config file and profile
dumpcat --config myconfig.toml --profile python

--llm

Send the dump output to a local or remote LLM via an OpenAI-compatible chat completions API. The value can be a built-in provider name or a full URL:

  • ollama — http://localhost:11434/v1/chat/completions
  • vllm — http://localhost:8000/v1/chat/completions
  • lmstudio — http://localhost:1234/v1/chat/completions
  • Any http:// or https:// URL
# Use a built-in provider
dumpcat -i .py --llm ollama -m llama3

# Use a custom URL
dumpcat --llm http://localhost:9000/v1/chat/completions -m my-model

When --llm is used without a value, Dumpcat looks for a default_target in the LLM config file (~/.dumpcat/dumpcat_profiles.toml). See Configuration for details.

The LLM response replaces the normal dump output on stdout. Use -o to write the response to a file instead.

--target / -t

Use a named target from the LLM config file:

dumpcat --llm -t mylocal

--model / -m

Specify the model name. Required unless the target config includes a model:

dumpcat --llm ollama -m llama3

--system-prompt

Set a system prompt for the LLM request:

dumpcat --llm ollama -m llama3 --system-prompt "You are a code reviewer. Be concise."

--set

Pass extra parameters to the LLM API. Repeatable. Values are auto-coerced to the correct type:

dumpcat --llm ollama -m llama3 --set temperature=0.5 --set max_tokens=4096

--api-key

Provide an API key for authenticated endpoints:

dumpcat --llm https://api.openai.com/v1/chat/completions -m gpt-4o --api-key sk-...

For reusable setups, store the API key env var name in a target config instead. See Configuration.