A native SwiftUI app for talking to your models, managing knowledge, and keeping your sessions in one place. Point it at a local backend or your own Eldric cluster — and let it update itself, safely, over a cryptographically signed channel.
The macOS app is a native client — it runs on your machine, stores your data on your machine, and connects only to the backends you choose. Chat, retrieval, sessions, and model management all live in one window. Nothing phones home; nothing leaves your network unless you point it at an external provider.
It pairs with the same platform the web chat and CLI use, so a conversation you start on your Mac can continue against your shared cluster — or stay entirely on your laptop with a local model.
Streaming conversations with markdown, code blocks, and inline tool results. Switch models mid-session and keep the thread.
Talk to local runtimes like Ollama and vLLM, or cloud providers — OpenAI, Anthropic, Groq, xAI, Together, and more — each configured with your own keys.
Native function calling where the backend supports it, with an XML fallback for everything else. The app picks the right mode per model automatically.
Build knowledge bases, import documents, and let answers draw on your own material. Preview what gets retrieved before it reaches the model.
Every conversation is saved locally and searchable. Export a thread, branch from it, or pick up where you left off.
Connect to a remote Eldric cluster with authentication and shared sessions, or run fully self-contained against a local model.
The retrieval manager lets you keep separate knowledge bases for separate projects. Drop in documents and folders across common formats, then test a query and see exactly which chunks would be pulled in. When you're satisfied, that context is injected into the conversation automatically.
A prompt library sits alongside it — reusable templates with variables you fill in at send time, organised by category, and multi-step chains for tasks that need more than one pass. Everything is stored on your device.
Beyond the conversation, the app gives you the tools to compare, configure, and prepare models for your workflow.
Run the same prompt against two models side by side and weigh quality, speed, and style before you commit.
Spin up configured models from starting points for chat, code, reasoning, and vision — with system prompts and parameters set.
Per-model behaviour — tool-calling mode, parameters, system prompt — saved and applied consistently across sessions.
The app updates itself in place, with every release cryptographically signed and verified before it installs — so you always know an update came from us and arrived intact. Choose the stable channel for tested releases, or the beta channel to try what's next early. You stay in control of when an update applies.
Download the macOS package from the downloads page (served from repo.eldric.ai), open it, and drag Eldric into Applications. macOS 13 or later. It's free to use against a local backend — no account required. From there, the app keeps itself up to date over the signed update channel.