Skip to main content
Every command supports --output-format to control how results are rendered to stdout:
FormatFlagWhen to use
Pretty--output-format pretty (default)Reading output in a terminal. Aligned key-value pairs with color and nested indentation.
JSON--output-format jsonScripting, piping into jq, saving to a file. Passes through original field order and numeric precision when the response is already JSON.
YAML--output-format yamlHuman-diffable configuration-like output.
Table--output-format tableArray responses where each item has the same shape — e.g. llm models list.
TOON--output-format toonToken-Oriented Object Notation. Compact, line-oriented, 30–60% fewer tokens than JSON — well-suited for piping into LLM prompts.
# Human-readable default
mka1 llm models list

# Machine-readable JSON
mka1 llm models list --output-format json

# Compact TOON for passing into another LLM
mka1 llm responses get --response-id resp_123 --output-format toon

​
Filter and transform with jq

Use --jq to filter or reshape responses inline with a jq expression. This always outputs JSON and overrides --output-format: 
# Extract a single field
mka1 llm models list --jq '.data[].id'

# Filter an array
mka1 llm responses list --jq '.data[] | select(.status == "completed")'

# Reshape into a new object
mka1 llm files list --jq '.data[] | {id, name: .filename, bytes}'
Because --jq prints plain JSON, it composes well with shells and with other CLI calls: 
mka1 llm files upload --file ./doc.pdf --purpose assistants --jq '.id' \
  | xargs -I {} mka1 llm vector-stores create-file --vector-store-id vs_123 --file-id {}

​
Pagination and streaming

Some operations return many items or server-sent events. Output is written incrementally as each item arrives:
FormatStreaming behavior
jsonOne compact JSON object per line (NDJSON)
yamlYAML documents separated by ---
toonOne TOON-encoded object per block, separated by blank lines
pretty (default)Pretty-printed items separated by blank lines

​
Paginate through a collection

Pass --all on list commands to follow the next cursor automatically and emit every page as it arrives: 
mka1 llm responses list --all --output-format json > responses.ndjson

​
Stream server-sent events

Streaming operations (for example, llm responses create --stream) emit one event per line: 
mka1 llm responses create \
  --model auto \
  --input '"Write three release-notes bullets."' \
  --stream \
  --output-format json

# Only keep text deltas
mka1 llm responses create \
  --model auto \
  --input '"Write three release-notes bullets."' \
  --stream \
  --jq 'select(.type == "response.output_text.delta") | .delta'
Press Ctrl+C to stop consuming a stream early.

​
Color

--color controls terminal colors for the pretty format:
ValueBehavior
auto (default)Colorize when stdout is a TTY, plain text otherwise.
alwaysAlways colorize.
neverNever colorize.
The CLI also respects the widely used NO_COLOR and FORCE_COLOR environment variables.

​
Include response headers

Use --include-headers when you need the HTTP response headers alongside the body — for example, to read a request ID for support or inspect a rate-limit header: 
mka1 llm models list --include-headers

​
Write binary responses to a file

Commands that return binary data (such as llm speech speak) support --output-file to write the body to a path without polluting stdout: 
mka1 llm speech speak \
  --text 'Hello, welcome to our service.' \
  --output-file ./welcome.wav
Use --output-b64 instead when you need to embed the binary payload in a JSON or text pipeline.

​
Exit codes

Exit codeMeaning
0Success. Response written to stdout.
1Error. Details written to stderr.
This makes if-style control flow straightforward in scripts: 
if ! mka1 llm responses get --response-id resp_123 > out.json 2> err.log; then
  echo "Request failed — see err.log"
  exit 1
fi
See debug and inspect for --dry-run and --debug, which give you more visibility into why a command failed.