CLI reference

Every redcrown subcommand and flag.

Environment variables

Keys a local redcrown eval reads from your shell. Each key goes only to its own provider, as the authentication for that provider's calls. RedCrown receives no key from a local eval.

Variable	Provider
`OPENAI_API_KEY`	OpenAI
`ANTHROPIC_API_KEY`	Anthropic
`OPENROUTER_API_KEY`	OpenRouter
`GROQ_API_KEY`	Groq
`HUGGINGFACE_API_KEY`	Hugging Face
`DEEPGRAM_API_KEY`	Deepgram
`GEMINI_API_KEY`	Google AI Studio
`NVIDIA_API_KEY`	NVIDIA NIM
`MISTRAL_API_KEY`	Mistral
`COHERE_API_KEY`	Cohere
`REDCROWN_CUSTOM_HTTP`	A custom OpenAI-compatible endpoint, as JSON (base_url, auth_header_name, auth_value, body_style, model)

`redcrown harness report`

Status counts and per-check pass rates for one run

Flag	Description	Default
`run_id`
`--runs`	the journal directory (one subdir per run)
`--rulings`	rulings file (JSON, or YAML with PyYAML)
`--policy`	check the run against this written policy file (JSON)
`--baseline`	the baseline run for a no_regression rule. Needs --policy
`--format`	print the policy result as text or JSON. Needs --policy	`text`

`redcrown harness compare`

Per-check diff of two runs; exit 1 on any regression

Flag	Description	Default
`base_run`
`head_run`
`--runs`
`--rulings`
`--format`		`text`
`--allow-corpus-drift`	compare the shared case ids even when the corpus versions differ	`False`

`redcrown harness export`

Write an upload payload for `redcrown push`

Flag	Description	Default
`run_id`
`--runs`
`--out`
`--name`	the experiment name shown on the record
`--label`	the system under test, e.g. one-button@<commit>
`--rulings`
`--quality-bar`		`0.0`
`--task-summary`

`redcrown push`

Push a local eval's results JSON, or a `redcrown harness export` payload, to RedCrown as a cloud run

Flag	Description	Default
`path`	results JSON from `redcrown eval --report-json` or `redcrown harness export --out`
`--api-url`		`https://api.redcrown.ai`
`--app-url`		`https://app.redcrown.ai`
`--org`	workspace (org) id; default personal org
`--experiment`	append the run to this standing experiment id
`--no-receipts`	strip per-item receipts/references before upload (confidential outputs never leave this machine)	`False`
`--proof-link`	mint and print a shareable proof link for the pushed run	`False`
`--publish-anyway`	mint the proof link even when the report's own caveats say it is not fit to publish (the override is recorded)	`False`

`redcrown score`

Score outputs produced elsewhere with a deterministic metric. Input: a scoring request JSON (experiment_id, items with references, candidates with a declared configuration and their outputs), or --items-csv plus --outputs-jsonl. Output: the ranked table on stdout; --evidence-json writes the private raw evidence that `redcrown check` reads (it can contain your data, keep it private); --report-json writes the upload format for `redcrown push`. Side effects: writes those local files only. Data destination: local file only, no network call. Authentication: none. Failure: exit 1 on an unreadable or invalid request, or f1 against references that are not JSON objects; nothing is written then. The CSV plus JSONL form carries no declared configuration, so its evidence cannot be checked against a baseline.

Flag	Description	Default
`path`	ScoreOutputsRequest JSON; alternatively use CSV + JSONL flags
`--items-csv`	CSV with item_id (or id), input, and reference columns
`--outputs-jsonl`	JSONL with candidate_id, item_id, and output/error fields
`--name`	comparison name for CSV + JSONL input
`--metric`	deterministic metric for CSV + JSONL input	`similarity`
`--quality-bar`	minimum acceptable quality for CSV + JSONL input	`0.0`
`--objective`	ranking objective for CSV + JSONL input	`quality`
`--report-json`	write upload-schema results for `redcrown push`
`--evidence-json`	write private raw report evidence for offline `redcrown check`

`redcrown check`

Apply a policy to a candidate evidence file, and optionally a baseline evidence file for the same feature, offline. Input: raw evidence from --evidence-json (not the --report-json upload format), --candidate-id, --baseline with --baseline-candidate-id, and --policy (JSON only). Output: pass, fail or indeterminate per rule and overall, with counts and excluded item ids, as text, markdown or json on stdout. Side effects: none; no file is written. Data destination: nothing leaves the machine. Authentication: none. Exit 0 pass; exit 2 fail or indeterminate (an indeterminate required rule is never a pass); exit 1 command, file or policy error (unreadable file, candidate id not in the file, YAML policy, a regression rule with no baseline). Both files must share experiment_id, items, references and scorer, or the comparison is refused with workload_mismatch, content_hash_mismatch, evaluator_hash_mismatch or platform_changed. Cost and latency rules parse but are always indeterminate.

Flag	Description	Default
`path`	raw candidate report JSON
`--candidate-id`
`--baseline`	raw baseline report JSON
`--baseline-candidate-id`
`--policy`	policy JSON (JSON is the only supported format)
`--format`		`text`
`--changed`	Also list the cases whose score moved against the baseline. This prints outputs. Keep it out of shared CI logs.	`False`

`redcrown attest`

Download a signed attestation for a completed run

Flag	Description	Default
`run_id`
`--out`	output file (default <run_id>-attestation.json)
`--api-url`		`https://api.redcrown.ai`

`redcrown build-dataset`

Build an experiment JSON from a public corpus (primock57, mmlu-pro)

Flag	Description	Default
`source`	public corpus source (omit when using --from-csv/--from-jsonl)
`--out`	path to write the experiment JSON
`--limit`	cap the number of (audio, reference) items
`--corpus-dir`	path to an existing corpus checkout (skips cloning)
`--quality-bar`	quality bar a candidate must clear; default = the dataset's published bar (mmlu-pro 0.75), NOT 0.0
`--monthly-minutes`	projected monthly audio minutes for the $/mo column	`20280`
`--no-clone`	do not clone; require --corpus-dir to point at a checkout	`False`
`--ids-file`	path to a newline-separated list of clip ids; builds exactly those items (the pinned benchmark subset). Applied after --limit, so do not combine them
`--from-csv`	build from a CSV with input + reference columns (text tasks)
`--from-jsonl`	build from a JSONL with input + reference fields (text tasks)
`--task`	task kind for --from-csv/--from-jsonl
`--input-col`	input column/field name (default 'input')	`input`
`--reference-col`	reference column/field name (default 'reference')	`reference`

`redcrown eval`

Run an experiment JSON locally and print the report

Flag	Description	Default
`path`	path to an experiment JSON file
`--sample`	run a bundled offline sample with the demo provider (no keys)
`--report-json`	also write the full results (receipts included) as an upload-schema JSON for redcrown push
`--evidence-json`	write the private raw report used by `redcrown check` (may contain customer data; do not publish)
`--audio-base`	base URL for per-item audio (item_audio = base + id + .mp3)
`--concurrency`	number of provider calls to run at once (default 1, sequential; raise it for many-candidate token benchmarks)	`1`
`--repeat`	run each candidate N times and report a run-to-run quality stdev (default: the experiment's repeat_n, or 1). Note: this makes N times the API calls, so N times the eval spend. Tip: set params.seed on a candidate for a provider that supports seeded sampling.
`--judge`	judge panel member for quality_metric judge; repeat for a panel (e.g. --judge openrouter:openai/gpt-5.5)

`redcrown import-results`

Upload a locally-run eval's aggregated results JSON into RedCrown

Flag	Description	Default
`path`	path to a results JSON file (the upload schema)
`--api-url`	RedCrown API base URL	`https://api.redcrown.ai`
`--token`	access token (or set REDCROWN_TOKEN); your RedCrown/Supabase bearer token

Flag	Description	Default
`--api-url`	RedCrown API base URL	`https://api.redcrown.ai`

`redcrown logout`

Delete the stored RedCrown credentials

`redcrown matrix-plan`

Preview a bounded experiment matrix offline (unreleased)

Flag	Description	Default
`path`	private matrix JSON specification
`--output`	explicitly create a new private experiment JSON file; never runs it

`redcrown rejudge`

Re-score an existing report-json's stored outputs with a new judge panel

Flag	Description	Default
`path`	results JSON from `redcrown eval --report-json` (with receipts)
`--judge`	panel member; repeat for a panel
`--keep-existing`	fold the original run's judge into the new panel	`False`
`--out`	path to write the re-judged report JSON
`--concurrency`		`8`

`redcrown verify`

Verify a RedCrown attestation file offline

Flag	Description	Default
`path`
`--key`	a JWKS file to verify against (offline); default: fetch the file's key_url

Generated from source. Last verified 2026-09-03.

CLI reference

Environment variables

redcrown harness report

redcrown harness compare

redcrown harness export

redcrown push

redcrown score

redcrown check

redcrown attest

redcrown build-dataset

redcrown eval

redcrown import-results

redcrown login

redcrown logout

redcrown matrix-plan

redcrown rejudge

redcrown verify