About Aegis

Aegis is a local-first AI workspace: a desktop app that keeps your data on your machine, with an optional account for the parts that need to live in the cloud.

Why it exists

Most AI assistants ask you to trust a server you can't see with your inbox, your codebase, and your documents. Aegis takes the opposite bet: run the model, the memory, and the tool-calling loop on your own machine, and only reach outward when you explicitly connect something. You get an agent that can plan and execute multi-step work across the tools you already use, without handing your data to a third party to get there.

How to use Aegis

Three steps, all of it running on your own machine.

Download a model

Open the LLM panel and pick a category — Coding, Reasoning, Vision, or a handful of others — or search for something specific. Every model is a plain GGUF file that downloads once and runs entirely on your machine. Pick a Vision model and Aegis automatically fetches its paired vision-tower file too, so it's ready to actually look at images the moment it's downloaded.

The Aegis LLM panel, showing the Vision category with downloadable models that can see images

Chat, or let it plan and execute

Ask questions and get answers in Chat Mode — attach a document or a photo and it's read into the conversation automatically. Switch to Agent Mode when you need Aegis to plan a multi-step task and execute it across your connected tools, reviewing the plan before anything runs.

The Aegis chat view, mid-conversation with a locally running model

Extend it from the Marketplace

Install automation tools and pre-built skills the chat agent draws on automatically. Nothing downloads or activates until you install it here — browse what's available and add only what you need.

The Aegis Marketplace, showing installable automation tools and skills

Local-first by default

Aegis runs as a desktop app. Your conversations, files, and credentials stay on your machine unless you explicitly connect a hosted service.

One account, one place

A single Aegis account follows you across devices and unlocks hosted connectors and paid features as you need them — nothing more.

Built in the open

We'd rather ship something small and honest than something bloated. The desktop app and this website are built by the same small team.

What it can do

Hybrid retrieval

Upload PDFs, Word docs, slides, and images. Aegis extracts, chunks, and embeds them, then answers questions using dense semantic search plus keyword matching, re-ranked before it responds.

Vision-capable models

Download a vision model from the LLM panel and attach a photo or screenshot in chat — Aegis sees the actual image, not just OCR'd text, in both Chat and Agent Mode.

Entity memory

The agent notices important people, projects, repos, and decisions as you talk, and asks before remembering them — so useful context persists into future conversations instead of resetting every session.

A local model hub

Browse and download open models from inside the app, organized by what they're good at — coding, reasoning, vision, and more — and run them fully offline. No API key, no per-token bill, no round trip to someone else's server.