AegisMission Control for Mac

Your agents. Your models. One desk.

Grok, ChatGPT, Claude, Kimi — your subscriptions or your keys — running in parallel worktrees on your Mac, with every session in view.

Confirm your email. We’ll use it for the Mac beta. Privacy.

Mac Built in Rust Model-agnostic
Mission Control 3 running · 0 awaiting ok · 1 done · $2.15 in flight

The watch post

Every session in view.

Most agents lock you to one vendor’s meter and one terminal. Aegis is the watch post over all of them.

  • Sign in with the subscriptions you already have — xAI, ChatGPT Plus/Pro, Claude subscription (Pro or Max), Kimi — or bring any API key
  • Parallel sessions in isolated worktrees — see what each one did, take it or drop it
  • Sessions, progress, and cost in one desktop
  • Voice Mode (Mirra) — speak to steer, approve, and follow along
  • A native Rust app around a real agent runtime, on your machine

One conversation, a versatile bench

One conversation. Many models.

Even in a regular conversation, one lead can delegate to supporting agents using different models and providers. Choose the model that suits each job.

  • Mix connected models and providers within a single conversation
  • Give planning, building, and review to different supporting agents
  • Follow the handoffs while the lead coordinates their work
  • Use Teams mode below when you want a separate layer of model teammates comparing perspectives
Athena’s owl on an olive branch, overseeing a lead and four supporting agents in one conversation

Athena’s owl sees in the dark

Team mode. Different minds, one job.

Choose two model teammates to examine the same prompt. They compare perspectives; a separate lead reviews their findings and coordinates the work.

  • Pair different models — such as Astra and Fable — for two perspectives on the same problem
  • Keep a distinct lead to weigh the findings, resolve disagreements, and bring the answer together
  • Supporting agents can explore, plan, or review alongside the team
  • Watch the handoffs and steer the work. You keep the final say.
Athena’s owl overlooking two model teammates and a separate lead
Conversation 01

Checkout
performance

Fable 5.1Lead
Conversation 02

API
investigation

ChatGPT-6 AstraLead
Checkout asks“Is the cart endpoint cached?”
API investigation replies“No. Every update hits the database.”

Across the workspace

Separate chats. Connected thinking.

One conversation can ask another for an answer. Keep each thread focused and pass along the finding that moves the work forward.

  • Send a question from one conversation to another
  • Bring the reply back to the conversation that asked
  • Connect two leads, or reach an agent working in another chat

Illustrated example. Each conversation keeps its own context.

A small Greek temple with three equal columns, Athena’s owl perched on the pediment before a rising sun
On your MacQwen 3.8In the cloudFable 5.1

Three ways to work

Local. Cloud. Or both.

Your Mac, your connected cloud models, or a team that brings them together. Choose where the work happens.

  1. Local. Run Qwen 3.8 on your Mac through Ollama or LM Studio.
  2. Cloud. Connect the models you use with supported subscriptions or API keys.
  3. Hybrid. Let a local model explore and a cloud teammate review in the same workflow.

Good to know

  • Subscription sign-in goes through each provider’s own login. Your credentials stay on your Mac.
  • Named for the aegis — the shield Zeus lent Athena. Built on Grok Build.

Just want launch news?

One email when Aegis ships. Separate list from beta testers. The owl keeps watch until then.

Confirm your email. One message when Aegis launches. Privacy.