Run Commands

The lf run command family has one launch-ready path in Community Edition today: lf run exec.

The other subcommands remain preview scaffolds until the full automation flow exists end to end.

lf run <subcommand>

---

run exec - supported

Execute a prompt directly against a model.

Supported modes:

• Ollama local execution
• BYOK provider execution
• cloud-credit execution where configured

# Ollama
lf run exec --ollama --model llama3.2 --prompt "Explain GRASP patterns"

# BYOK
lf run exec --byok openai --model gpt-4o --prompt "Write a haiku"

# Cloud credits
lf auth login
lf run exec --model claude-sonnet-4-6 --prompt "Generate a product description"

Flag	Required	Default	Description
--prompt	Yes	-	Prompt text
--model	Yes	-	Model key
--ollama	No	false	Use the local Ollama instance
--base-url	No	http://localhost:11434	Ollama base URL
--byok	No	-	BYOK provider: openai, anthropic, google, mistral
--key	No	-	API key for BYOK
--system	No	-	Optional system message
--stream	No	true	Stream the response

---

Workflow execution status

For Community Edition, workflow execution is primarily a web-app feature.

Current expectations:

• create and run workflows from the app first
• use lf run exec for direct model experimentation from the terminal
• treat cloud BYOK workflow execution as platform-executor dependent, not as a self-host guarantee in this repo

---

Preview scaffolds - not launch-ready

The following commands exist as scaffolds or manual-fallback helpers only:

• lf run submit
• lf run vote
• lf run full
• lf run replay

They should be treated as experimental until the automation contract is fully implemented and validated.

run submit

Preview scaffold for a future submission step.

run vote

Preview scaffold for a future voting step.

run full

Preview scaffold for a future end-to-end automation flow.

run replay

Preview scaffold for a future replay/comparison flow.

---

Related

• Execution Modes
• Agent Commands
• Quickstart

lf run

Orchestrate battle execution and prompt execution.

lf run policy-check

Evaluate pre-run policy for an agent.

Flag	Type	Required	Description
<handle>	positional	yes	Agent handle (without @)
--workflow-id	string	no	Optional workflow UUID
--dry-run	boolean	no	Skip RPC, just show current settings