CLI reference
aicr <command> [options]
| Command | Description |
|---|---|
aicr run <dir>... | Run an experiment from one or more project directories |
aicr create <dir>... | Create a new project directory with template config files |
aicr generate summary [run-dir] | Regenerate manifest.csv and summary CSVs from an existing manifest.json |
aicr lock <dir>... | Lock a project to prevent further runs |
aicr version | Print the harness version |
aicr [project-dir] | Shorthand for aicr run — uses current directory if omitted |
aicr --help / -h | Show help |
run
aicr run my-study.project/ # explicit path
aicr run proj-a/ proj-b/ proj-c/ # multiple projects in sequence
aicr # uses current directory if it contains project.json
Output lands in <project-dir>/output/<timestamp>/.
create
Scaffolds a new project directory with template config files:
aicr create my-new-study.project/
generate summary
Regenerates manifest.csv, summary_counts.csv, and summary_pct.csv from an existing manifest.json without re-running the experiment:
aicr generate summary my-study.project/output/20260617-120000/
aicr generate summary # uses current directory
Useful when scoring logic is updated and you want to recalculate results from existing data.
lock
Sets edition.read_only = true in project.json, preventing future runs:
aicr lock my-study.project/
Azure prerequisites
For deployments using AzureModeApi or AzureAgentApiExecutor:
- Run
az loginonce, signed in as an identity with the Cognitive Services User role on the target Azure AI Foundry resource. - Ensure network access for the initial Azure token fetch.
A 401 or 403 on the first call almost always means the role assignment is missing, not a code error.