DELIRIUM — AI MORE THAN THOUGHT

Your AI tools.
Connected by choice.

Send a task through supported AI connections and receive the answer in Delirium. Choose which connections Auto may use, with fallback after confirmed availability failures. Each app’s entry explains its available options.

OpenAI

ChatGPT · Codex

Anthropic

Claude · Claude Code

Google

Gemini · AI Studio · NotebookLM

xAI

Grok

Perplexity

Search · API · Comet

Microsoft

Copilot · Microsoft 365

GitHub

Copilot · VS Code

Superhuman / Grammarly

Writing tools

Replit

Agent · Coding workspace

Character.AI

Character chat

Qwen

Browser chat · Optional API connection

DeepSeek

Browser chat · Optional API connection

Local models

Ollama · LM Studio

Catalog entries include launch and setup links. They do not all support integrated chat or model control. Listed companies are not sponsors or endorsers.

Search this page9 topics

Three ways to reach your AI.

KNOW WHAT CONNECTS

Choose the kind of access that fits your work.

AI CHAT API CONNECTIONS

Send from Delirium

Use supported Anthropic, Google, xAI, Perplexity, Qwen and DeepSeek API connections in AI chat & tasks. Configure your own keys, allowed models and preferences.

Auto can choose among eligible connections when enabled. A configured key is not proof of provider account access.

SUPPORTED NATIVE TOOLS

Bring native replies into Delirium

Codex, Claude Code and Gemini CLI can receive a task and return an answer through their installed-client interfaces. Delirium starts a separate conversation; your existing chat windows and histories stay separate.

Native tools keep their own sign-ins, permissions and billing. Delirium does not control every external app.

OFFICIAL LAUNCH & SETUP

Find the right starting point

Open official websites and download or setup pages from the app directory. Useful even when a tool has no direct connection.

Launch-only entries do not send your task or change the model inside another service.

Apps & connections in the combined Delirium Desktop app — Apps & connections · Interface preview

Put your hardware
to work.

LOCAL MODEL CONNECTIONS

Connect a compatible local model through Ollama or LM Studio. Select it in Studio or explicitly allow local AI in Auto.

Set up your local server

Install Ollama or LM Studio, download a compatible model in that app, start its local server, then discover and select it in Delirium’s AI controls.

Understand what runs locally

Direct local-model requests use a loopback connection without cloud fallback. Auto: apps & models can fall back only to connections you have enabled; allow only local connections when you want local-only routing. Your server’s configuration controls whether its own execution is local. Compatible hardware and downloaded model files are required.

Locally processed requests have no cloud API fee; hardware and electricity still have a cost. Model origins and licenses vary. Cloud provider subscriptions and usage are separate from Delirium.

Start your 7-day free trial ↗Read the setup guide ↗

Replies in Delirium chat

AI CHAT & TASKS

Auto: apps & models can send tasks through installed Codex, Claude Code and Gemini CLI connections and return their answers. Press Enter to send, use Shift + Enter for a new line, and copy the latest answer from chat. Optional API connections include Claude, Gemini, Grok, Perplexity, Qwen and DeepSeek; local models use configured compatible servers. Other app entries provide launch and setup options, not automatic control of their chat windows.

Set up connected apps →

Routing tools

CHOOSE YOUR CONNECTIONS

Save routing preferences for different kinds of work, compare eligible models before sending, and choose a fallback order for your allowed connections.

Use local-only mode in Auto: apps & models to restrict that selection to configured local models. Direct cloud selections and other AI apps have their own settings.

Explore routing tools →

Task status & progress

Live connection status

Status updates automatically when an observed connection, app, route, model or task changes. It shows the latest change time and distinguishes a requested model from one confirmed by the connected app.

An enabled connection alone is not proof of a working chat. Closed, disconnected and unobserved sessions are reported clearly. This status does not test every provider or identify an unrelated open chat.

Qwen and DeepSeek

Open their web chats from Delirium, or use your own API account to bring answers into Delirium chat. API keys, billing and model access are separate from app subscriptions. Qwen requires the official endpoint for your workspace.

Free-first AI routing

Automatic routing prefers a suitable configured local model or a Gemini API project you have confirmed has billing disabled for simple questions. Suitability is estimated from the task and model settings.

If no suitable free route completes the request, Delirium explains why it could not send and keeps your draft. Paid fallback is a saved preference in Routing tools, so you do not have to choose free or paid for every message. Browser handoffs do not send your question or return the answer automatically. Provider free access and quotas may vary.

More demanding tasks keep normal model routing. Manually selected providers keep their chosen route. Subscriptions, promotional credits and unknown prices are not treated as free.

Chat controls

DESKTOP 1.3.1 / EVERYDAY CONTROLS

Chat from the keyboard

Enter sends your message or task. Shift + Enter adds a line. Copy the latest answer and keep a newer draft if an earlier send fails.

Understand the route

Simple chat prefers suitable configured free connections. Save optional paid fallback once in Routing tools. Local diagnostics record route decisions, changes and mismatches without question text or credentials.

Media in its own group

Movies, TV, Music, Games, Social video and Library sit under Media in the sidebar. The page stays open for browsing and playback, with shared clock and desktop controls.

Auto API recommendations and connected-app recommendations use different eligible connections. Requested models and backend-confirmed models are recorded separately; some clients do not confirm a model. Provider access, quotas and charges remain separate.

New agent connections

Agents adds Muse Code developer-preview support and optional OpenAI API background chat, research, image generation and GPT-Live voice. Muse Code requires an installed signed-in client; OpenAI API features require your own key and saved usage opt-in.

ChatGPT Dots setup opens its official guide and exports a local Delirium MCP connection configuration. Dots stay in ChatGPT’s cloud and need their supported account and computer/app connections. Delirium does not provide an embedded Dots API.

Agent setup and requirements →

Connection-specific assistant features

Native Codex, Claude Code and Gemini CLI connections retain their own account, model access, tools and permission controls. Auto considers your permitted connections. Muse uses its supported installed client. Explicit OpenAI API connections add supported research, image input, image generation, transcription and voice; availability and charges depend on that account.

Health checks report only information exposed by the connection. No provider-owned avatar, personal messaging inbox or desktop action is claimed merely because its app is installed.

Assistant capabilities and limits