CLI Reference¶

Kortex includes a terminal CLI for inspecting providers, models, routing decisions, and checkpoint history.

Installation¶

The CLI is installed automatically with the package:

pip install kortex-ai

Run with:

kortex <command>
# or
python -m kortex.dashboard.cli <command>

Demo Mode¶

When no API keys are detected, the CLI runs in demo mode with example models. This lets you explore all commands without any configuration.

Commands¶

status¶

Show system status — providers, models by tier, registered agents, and state store.

$ kortex status
Kortex Status
[demo mode - no API keys detected, showing example models]

Providers (2):
  openai: OK
  anthropic: OK

Models (4):
  fast: 1
  balanced: 2
  powerful: 1

Agents: 3
  researcher: Gathers and analyzes information
  writer: Drafts content and documentation
  reviewer: Reviews and provides feedback

State Store: InMemoryStateStore

models¶

List all available models across all providers, sorted by tier and cost.

$ kortex models
PROVIDER   MODEL                     TIER      INPUT/1K  OUTPUT/1K  LATENCY  CAPABILITIES
---------  ------------------------  --------  --------  ---------  -------  -------------------------
openai     gpt-4o-mini               fast      $0.0001   $0.0006    200ms    reasoning, content_generation
anthropic  claude-sonnet-4-20250514  balanced  $0.0030   $0.01      800ms    reasoning, code_generation, content_generation
openai     gpt-4o                    balanced  $0.0050   $0.01      600ms    reasoning, code_generation, content_generation
anthropic  claude-opus-4-20250514    powerful  $0.01     $0.07      2.0s     reasoning, code_generation, content_generation, analysis

Local models (zero cost) are tagged with [local].

dry-run¶

Run a routing dry-run to see what models would be selected and at what cost.

$ kortex dry-run --task "Analyze data" --complexity complex --pipeline researcher,writer,reviewer

Options:

Flag	Default	Description
`--task`	(required)	Task description
`--complexity`	`moderate`	`simple`, `moderate`, or `complex`
`--pipeline`	all agents	Comma-separated agent IDs

Output includes routing decisions (model, reasoning, cost, latency, fallback), total cost estimate, and handoff chain with checkpoint IDs.

history¶

Show checkpoint history from the state store.

$ kortex history
$ kortex history --agent researcher
$ kortex history --task <task-id>
$ kortex history --last 5

Options:

Flag	Description
`--agent`	Filter by agent ID
`--task`	Filter by task ID
`--last N`	Show only the N most recent checkpoints

config¶

Show current configuration — state store, providers, routing strategy, environment variables.

$ kortex config
Kortex Configuration

State Store: InMemoryStateStore
Providers: openai, anthropic
Routing Strategy: HeuristicRoutingStrategy
Log Level: INFO

Environment:
  KORTEX_LOG_LEVEL=DEBUG
  KORTEX_STATE_BACKEND=sqlite

Environment Variables¶

Variable	Description
`KORTEX_STATE_BACKEND`	State store backend: `memory`, `redis`, or `sqlite`
`KORTEX_LOG_LEVEL`	Log level for structlog output
`OPENAI_API_KEY`	Enables OpenAI provider auto-discovery
`ANTHROPIC_API_KEY`	Enables Anthropic provider auto-discovery
`OPENROUTER_API_KEY`	Enables OpenRouter provider auto-discovery
`NO_COLOR`	Disables ANSI color output when set