Reference: flags, modes, and exit codes

Global flags

Available on any command:

Flag	Description
`--json`	Structured JSON output on stdout. Implies `--no-input`.
`--agent`	Agent mode: implies `--json`, `--no-input`, `--compact`, `NO_COLOR=1`, `JELOU_NO_SPINNERS=1`.
`--compact`	JSON without whitespace (saves ~30-40% of tokens).
`--human`	Forces human-readable output even if stdout is redirected (overrides auto-JSON).
`--no-input`	Disables interactive prompts; destructive operations fail hard instead of asking.
`--profile <name>`	Uses a specific auth profile for that command.
`--describe`	Emits the command schema as JSON, without executing it.

Output modes for agents

Auto-JSON on redirect: if stdout is not a terminal (e.g. jelou whoami | jq), the CLI switches to JSON automatically. Force it to human-readable with --human or JELOU_FORCE_HUMAN_OUTPUT=1. Response shape:

// Success
{ "ok": true, "data": { } }
// Error
{ "ok": false, "error": { "code": "...", "message": "...", "details": {} } }

Error codes: AUTH_ERROR, FORBIDDEN, NOT_FOUND, VALIDATION_ERROR, API_ERROR, MISSING_FUNCTIONS_PROJECT, MISSING_WORKFLOW_PROJECT, MISSING_AUTH, INPUT_ERROR, UNKNOWN_ERROR, NETWORK_ERROR, TRANSIENT_ERROR, MISSING_LOCKFILE. The error object can also include a retryable boolean field indicating whether it’s worth retrying the command as-is, with nothing changed:

{ "ok": false, "error": { "code": "NOT_FOUND", "message": "Not found.", "retryable": false } }

The --json responses of action commands include a breadcrumbs array with the suggested next commands. Streaming logs emit NDJSON (one JSON object per line).

Mode flags are output-only, never security. Collision preservation, the push rejection on unresolved incoming, schema validation, and dirty-state checks always run, with or without --agent. Destructive operations require --yes in JSON/agent mode — and that flag is for CI, not a shortcut to skip confirmations.

Introspection with `--describe`

--describe walks the command tree and emits its signature as JSON, without executing anything. It’s read-only — use it to plan.

jelou --describe                       # root command: lists top-level subcommands
jelou project --describe               # project subcommands
jelou databases create --describe      # arguments and options of a command

It returns { command, description, arguments, options, examples, subcommands }. Global flags are filtered out so you only see the command-specific surface.

Exit codes

The CLI uses granular exit codes so an agent can branch without parsing stderr:

Code	Meaning	Typical cause
0	Success	—
1	Generic / unknown	Build failures, uncategorized errors
2	Input / validation	Invalid flag, missing argument, schema mismatch
3	Not found	Unknown id, slug, or name
4	Auth / forbidden	Missing or expired token, insufficient scope
5	DB not ready	Database still provisioning
6	Transient / API	5xx, network timeout — safe to retry
7	State conflict	Lockfile drift, unresolved incoming, partial push
8	Secret detected (reserved)	The secret-scan helper exists but isn’t wired into production paths yet
9	Lock busy	Another `jelou pull`/`push` in progress
10	Local schema mismatch	The `qa.db` schema is incompatible (`QA_SCHEMA_MISMATCH`) — upgrade `jelou`

Codes 7-9 are specific to jelou pull/push/incoming (8 is reserved and not yet emitted on production paths). Code 10 shows up when the local toolchain has fallen behind the schema the CLI expects. In --json mode, the same code appears in error.code with a hint field that suggests the next command.

Environment variables

Variable	Description
`JELOU_TOKEN`	Authentication token (highest priority; no need for `jelou login`)
`JELOU_PROFILE`	Default profile
`JELOU_NO_INPUT`	`1` for non-interactive mode
`JELOU_FORCE_HUMAN_OUTPUT`	`1` to force human-readable output even if stdout is redirected
`NO_COLOR`	`1` to disable ANSI colors
`JELOU_NO_SPINNERS`	`1` to disable progress spinners (implied by `--agent`)
`CI`	Auto-detected for non-interactive mode

Maintenance commands

jelou update --agent     # is there a new version? (read-only)
jelou doctor --json      # health check (api, auth, profile, lockfile, project, authoring, skills, version)
jelou changelog          # release notes
jelou feedback           # send a report (always asks first)

To update the CLI:

npm install -g @jelou/cli@latest
jelou --version
jelou agent install --global --all-targets --yes --no-input   # refreshes the skills

To query company analytics (the catalog, one metric, or a whole dashboard), use jelou metrics — it has its own page in Metrics.

Getting Started

Developer

Jelou CLI

Outbound Messaging

Management

Marketplace

Channels

Configuration

Billing

Reference: flags, modes, and exit codes

Global flags

Output modes for agents

Introspection with `--describe`

Exit codes

Environment variables

Maintenance commands

​Global flags

​Output modes for agents

​Introspection with --describe

​Exit codes

​Environment variables

​Maintenance commands

Global flags

Output modes for agents

Introspection with `--describe`

Exit codes

Environment variables

Maintenance commands