Connect an OpenAI Agent

This tutorial shows the beta-safe way to work with an OpenAI-backed integration in Community Edition: use lf run exec for direct execution, and treat the agent surface as preview metadata rather than a full automation contract.

Prerequisites

CLI built and configured (lenserfight init, lenserfight doctor)
Authenticated (lenserfight auth login)
An OpenAI API key (BYOK)

Step 1: Use direct BYOK execution first

bash

lenserfight run exec --byok openai --model gpt-4o --prompt "Explain workflow retries simply"

This is the launch-ready path for OpenAI use in Community Edition.

Step 2: Optionally register preview metadata

bash

lenserfight agent connect \
  --name "GPT-4o Agent" \
  --type openai-agents \
  --config '{"model": "gpt-4o", "temperature": 0.7}'

Use this only when you want a managed record in the UI or CLI. Do not treat it as a stable adapter-runtime contract.

Supported adapter types

Type	Use case
`openai-agents`	OpenAI Agents SDK — best for GPT-4o, GPT-5
`langchain`	LangChain chains and agents
`crewai`	CrewAI multi-agent frameworks
`mcp`	Model Context Protocol servers
`ollama`	Local models via Ollama
`http`	Any HTTP endpoint
`custom`	Custom integration

BYOK model

Community Edition favors explicit BYOK usage:

lf run exec --byok openai ... for direct execution
local or UI-managed metadata when you want a preview agent record
workflow UI for workflow composition and run monitoring

What you learned

how to use OpenAI through the supported run exec path
how to register preview metadata if you need an agent record
why Community Edition does not yet promise autonomous adapter flows

Connect an OpenAI Agent ​

Prerequisites ​

Step 1: Use direct BYOK execution first ​

Step 2: Optionally register preview metadata ​

Supported adapter types ​

BYOK model ​

What you learned ​

Related ​