Calibrated AI you can audit

Run your AI fleet as one

Cortex is the backend. The extension is where you govern it. Together they turn a set of independent AI practices into a fleet you can trust and control.

Cortex connects your practices into one shared, coordinated mesh. The extension is where a human sees what the fleet is doing and approves the actions it proposes. Neither is much use without the other, so they ship as one product, for a single developer through a whole team.

🧠

Cortex

early access

The shared-brain backend. It holds what your practices learn, routes the actions they want to take through your approval, and keeps the agreed plans that span more than one session.

🧠

Knowledge that compounds

What each practice learns (the findings, decisions, and dead-ends that matter) becomes available across your org's practices when they opt in to share it, grounded and on demand. A practice in one project can be handed a lesson learned in another, an hour earlier, with no one re-explaining.

🛂

Action that gates through you

When one practice proposes acting on another's work, the request routes to a human for a yes or no before anything runs. The gate is built into how practices coordinate, so your fleet cannot quietly do something you did not approve.

📌

Shared plans that persist

When work spans several practices and several sessions, the agreed plan is recorded with its participants and current state, so coordination survives long after the conversation that started it.

Each organization's data lives in its own store; sharing across organizations is opt-in and gated. Cortex runs as a hosted service today, with self-host and org-managed deployments for teams that need to run their own.

Connecting to Cortex: the local daemon plus the Cortex cloud connection, configured in the extension

Your AI tools and the extension connect to Cortex, the backend that links your fleet. It speaks MCP, so any MCP-capable tool can plug straight in.

🔭

The extension

beta

The control plane in your browser. It is where the abstract becomes operable: you see what every project and practice is doing, and you make the call on what they are allowed to do next. Watching is half of it. Governing is the other half.

✅

Review and decide

Proposals from any practice, in any project, land in one place. Accept, decline, or send back for a change. This is where the gate actually happens.

📊

See the whole fleet

One cross-project view of what is open, what each practice is working on, and what needs attention, without opening eight terminals to find out.

📥

Capture from anywhere

Pull useful context out of your AI conversations (Claude, ChatGPT, Gemini, and more) and into the practice that should keep it. A low-friction way in, useful even before Cortex.

The extension's review inbox: a cross-AI proposal awaiting accept or decline

A short walkthrough of a real approval, from proposal to sign-off to the action landing, is on the way.

Ask for a live demo →
🛡️

Stays safe as it scales

beta

As a fleet grows, you cannot watch every step yourself. A governance layer reasons about the risky, cross-boundary actions and routes them for a human decision, and it sweeps for anything that quietly slipped. It reasons and recommends. It does not act on the high-stakes work on its own. The human stays the gate by default.

🗂️

A shared record of who you work with

early

Your organization's shared, calibrated knowledge of its contacts, companies, and engagements, navigable from a terminal dashboard. The knowledge is shared across the org by agreement, so the whole team works from the same record. Your own working notes stay yours unless you choose to share them.

Your knowledge stays yours

Sharing is something you opt into, not the default. Each organization's data is isolated and encrypted at rest, with self-host federation in development for teams that need to run their own. Knowledge is shared by agreement, between parties who both said yes.

We hold ourselves to strong isolation discipline. We do not claim formal compliance certifications we have not earned.

See it on a real fleet

The fastest way to understand the product is to watch it run. Book a walkthrough and we will show you what it would take on yours.