Skip to main content
ADE runs Claude Code, Codex, Cursor, Factory Droid, and OpenCode — every chat, CTO task, and worker routes through one of these five agents. Connect at least one before starting agent work. Each can use your own API key or, where supported, an existing subscription sign-in.

​
The five agents

AgentConnect withTypical use
Claude CodeClaude subscription or Anthropic API keyGeneral coding, planning, review, worker tasks.
CodexOpenAI sign-in or API keyOpenAI coding workflows and CLI-backed sessions.
CursorCursor API key (CURSOR_API_KEY)Cursor agent sessions through the Cursor SDK.
Factory DroidFactory sign-in or API keyDroid agent sessions, including AGI mode.
OpenCodeAPI key, OpenRouter, or local endpointBroadest model reach; local or provider-backed.

​
The models you can reach through ADE’s agents

These agents — plus OpenCode and OpenRouter routing — give you access to models from every major lab. Pick whichever fits the task.
Anthropic ClaudeOpenAIGoogle GeminiDeepSeekQwenMistralMeta LlamaGrokOllamaPerplexityCohere

​
Connect a provider

1

Open Settings → AI Connections

Choose the agent you want to connect.
2

Add credentials or sign in

Paste an API key, sign in with a subscription, configure a local endpoint, or authenticate the provider CLI.
3

Verify

ADE probes the provider CLI and lists its available models. A green status means it’s ready to start a session.
4

Choose defaults

Set your default chat model and any role-specific overrides for CTO, workers, or specific lanes.
Per-agent key steps are below.
Sign in with a Claude subscription or paste an Anthropic API key. ADE runs Claude through the bundled Claude Agent SDK, so the claude CLI does not need to be installed separately. See Claude Code for lane context and permission modes.
Sign in with your OpenAI account or provide an API key. Codex runs as a CLI-backed session and reaches OpenAI models.
Provide a Cursor API key (or set CURSOR_API_KEY in the environment ADE launches from). ADE recognizes the key as valid auth and discovers Cursor’s available models at startup. See Cursor.
Sign in with Factory or provide an API key. Droid runs as an agent provider and supports AGI mode for larger autonomous tasks.
Toggle OpenCode on; ADE downloads and manages the binary and local server. OpenCode is what powers API-backed and local-model chats — back it with an Anthropic or OpenAI key, an OpenRouter key for any model, or a local OpenAI-compatible endpoint such as LM Studio or Ollama. See OpenCode & Windsurf.

​
Permission presets

Providers expose different native permission models. ADE maps them into a small set of practical choices.
Preset styleUse it when
Plan / read-onlyYou want the agent to inspect and propose first.
Default / editYou trust the lane and want normal coding flow.
Full accessYou are in a trusted environment and want minimal prompts.
CustomYou want the provider’s own config to decide.
Set stricter modes for unfamiliar repositories. The full mapping lives in Permissions.

​
Cost and usage

Running several agents in parallel can spend quickly. Use cheaper models for routine follow-up and stronger models for planning, architecture, and hard debugging. The top-bar Usage popup shows live provider quota windows and lets you set a budget cap; Settings → Stats is the retrospective view of token, cost, and GitHub activity.
Per-provider usage and quota windows in ADE
Provider calls go directly from your machine to the provider or local endpoint you configure. Keep keys out of committed files — ADE stores them in local encrypted settings.

Permissions

Lane scope, provider modes, and how secrets are handled.

Settings

Every Settings section and a first-pass order.