AE Agent

Connecting an AI provider

AE Agent works with the AI provider of your choice. There are three ways to connect: bring your own API key, use an existing subscription through the official CLI, or run a model locally on your machine. Mix and match — switch between them anytime from the model picker.

Three ways to connect

a. API key

Paste your Anthropic, OpenAI, or Google API key and pay per use, billed directly by the provider. The simplest option — works instantly and supports every model the provider offers.

b. Subscription (official CLI)

Connect the official command-line tool — Claude Code, OpenAI Codex, or Gemini CLI — and use your existing Pro / Max / Plus plan. No per-token cost: usage counts against your subscription instead of a metered API bill.

c. Local AI

Run a model entirely on your own machine with Ollama or LM Studio. Free, private, and works offline — nothing leaves your computer.

Option A — API key

Open Settings, pick your provider, and paste your key. See Setup & API Keys for where to find each provider's key. Keys are stored locally and pay-as-you-go.

Option B — Subscription via the CLI

If you already pay for Claude Pro/Max, ChatGPT Plus, or Gemini, you can use that plan inside AE Agent through the provider's official CLI — with no extra per-token charges.

Install the CLI

Install the official command-line tool for your provider:

npm install -g @anthropic-ai/claude-code

npm install -g @openai/codex

npm install -g @google/gemini-cli

The desktop app is not the CLI. You must install the actual command-line tool via npm — having the ChatGPT or Claude desktop app installed is not enough.

Run the CLI once and complete the browser OAuth sign-in. This links the CLI to your subscription. You only do this once — the session is remembered.

Select it in AE Agent

Open the model picker in the panel. AE Agent detects the signed-in CLI automatically and lists its models under Subscription. Pick one and start.

The Gemini subscription is noticeably slower for multi-step Agent tasks. It works well for Chat mode and short operations, but for long Agent runs prefer an API key or a Claude/OpenAI subscription.

Option C — Local AI

Run a model locally with Ollama or LM Studio for free, private, offline use. See the dedicated Local models guide for setup.

The unified model picker

Whatever you connect, every available model shows up in one picker, grouped by source so you always know what you're paying (or not paying) for:

Group	Source	Cost
Subscription	Connected CLI (Claude Code, Codex, Gemini CLI)	Counts against your plan
API key	Anthropic / OpenAI / Google key	Pay per token
Local	Ollama / LM Studio on your machine	Free

You can keep an API key and a subscription connected at the same time and switch per task — e.g. a subscription for everyday work and an API key as a fallback when you hit a plan limit.

Chat vs Agent Local models