Cadence by Stratium is built on Claude for reasoning and the Model Context Protocol for connectors. Stratium provides the executive operating layer. The structure. The routines. The work items. The briefing. The focus surfacing. Claude reads, classifies and prioritises. MCP plugs the whole thing into the systems an executive already uses.
That is the architecture. The rest is what it means.
Key facts
- Stratium is the executive operating layer. Claude is the reasoning engine. MCP is the connector standard.
- Cadence does not pretend to be an AI company. It is built on the best one.
- MCP is an open standard now adopted across ChatGPT, Gemini, Microsoft Copilot, Cursor and VS Code. Your data does not stay in Stratium because it never needed to leave your own systems.
- IT buyers approve Cadence in a single review because the architecture is honest, the data residency is explicit, and the protocol is open.
Why this matters
Most productivity software sells the appearance of AI. The features carry the right buzzwords, the demos generate confident output, and the buyer is left to work out what the model actually does, what it does not, and where the data goes.
That is not a serious posture for a tool an executive runs their day on. Senior leaders do not have time to reverse-engineer marketing copy to find out whether their inbox is being read, whether their commitments are being stored, or whether the model behind the badge is one they would trust with their calendar.
We decided early at Stratium that we would not play that game. The product names the engine. The engine is named clearly enough that the IT buyer can look it up, the security team can read the safety posture, and the executive can understand what is happening on their behalf without a translator.
The three roles
The product has three layers. Each one has a single job.
Stratium is the operating layer. The structure of an executive's day. The plan lifecycle from Draft to Closed. The work item taxonomy. The focus blocks. The strategic drivers that pull the annual pillars all the way down to today's first action. The weekly review. The full product in any browser, and the native apps on Mac, iPhone, iPad and Apple Watch. None of this is AI. It is product. It is the discipline we believe an executive day needs, expressed in software.
Claude is the reasoning engine. When Cadence reads your inbox at 05:30 and decides which of last night's messages imply a commitment you need to remember, Claude makes that call. When yesterday's transcript is processed into a list of actions with the right owner, the right project and the right priority, Claude does the reading. When the morning briefing is composed from your calendar, your open work items, the things you said you would do last week and the things you have not yet answered, Claude assembles the language. Stratium tells Claude what to do, in what order, and what success looks like. The execution is the model.
MCP is the connector standard. The Model Context Protocol is how Claude reads your calendar, your inbox, your sent items, your transcripts and your CRM, without those systems exposing themselves to a hundred bespoke integrations. MCP runs in both directions. Claude can read what it needs to reason about. It can write the result back. Cadence uses MCP because the standard is open, the industry has adopted it, and it keeps the data where it belongs.
Three layers, three jobs. Stratium decides what should happen. Claude works out what is happening. MCP carries the conversation between Claude and the systems the executive already trusts.
What this looks like, named
Theory is cheap. The architecture lands when you see what it does. Cadence ships a set of named routines, each one a Claude Skill running on a schedule through MCP. Four examples worth naming.
The daily briefing routine. Runs overnight. Pulls your calendar for the day, your open work items, yesterday's transcripts, the inbox traffic since you last checked, and the sent items from the past week. Builds the briefing. Lands on your Mac before 06:00. The executive opens the laptop to a structured view of the day, not a pile of unread messages and an empty plan.
The inbox routine. Reads inbox and sent items together. The sent items matter as much as the inbox, because the sent items show what you said you would do. The routine surfaces commitments you made last week that have not been closed, replies you owe people, and email threads that have moved into "needs a decision" without being flagged. The executive sees the open commitments. The EA sees the queue.
The transcript intake routine. Connects to Fireflies, Otter, Microsoft Teams transcripts and the rest. Reads the transcript when it lands. Extracts the actions with owner, due date and the strategic driver they sit under. Drops the proposed actions into the right project. The executive confirms before anything is created. Nothing slips into the plan without a human in the loop.
The focus surfacing routine. Watches the calendar for the start of a focus block. When the block begins, the Mac app surfaces the right work item with one tap to Defer or Complete. macOS Focus Filter silences the rest of the system. The block becomes defensible at the operating-system level, not just on the calendar.
Four routines. All Claude Skills under the hood. All running through MCP into the systems the executive already uses. None of them require the executive to remember to open an app.
Why we picked Claude
We considered three reasoning engines: Claude, ChatGPT and Gemini. All three are capable. All three support MCP. The decision was not about which one is best at any single benchmark. It was about which one fitted the shape of the work.
Three things tipped Claude.
Skills. Cadence is not a chat product. The executive does not open a window and type a question. The product runs routines. The daily briefing is a routine. The inbox cross-reference is a routine. Transcript intake is a routine. Focus surfacing is a routine. Claude's Skills system, which lets us write a procedural instruction set the model executes the same way every time, is the natural shape for routine work. The same pattern does not exist in the same form in the other two platforms today.
Scheduled tasks. A briefing that arrives at 06:00 because the executive set it to is a different product from a chatbot that answers when prompted. Claude's Cowork scheduled task pattern runs the routine at the right time without the executive lifting a finger. The discipline lives in the schedule. The reasoning lives in the model. Nothing depends on whether the user remembered to open an app.
Safety posture. Executive data is sensitive. Calendar entries reveal strategy. Inbox content reveals deal status. Sent items reveal commitments and judgement. We wanted to build on a model where the company behind it had published its safety thinking, defended it under pressure, and enforced operational discipline on its API customers. Anthropic does this. We picked Claude in part because the company behind it is the one we want our security team to read about.
Those three, in that order, made the decision straightforward.
What MCP is, and why it matters
The Model Context Protocol is an open standard. Anthropic proposed it in late 2024. By late 2025 it had passed ten thousand active public servers and ninety-seven million monthly SDK downloads. ChatGPT supports it in developer mode. Gemini Enterprise supports it. Microsoft Copilot, Cursor and VS Code all ship MCP clients. It is the closest thing the industry has to a USB-C standard for AI connectors.
For Cadence, MCP does three things.
It keeps your data where you put it. When Cadence reads your Outlook inbox, MCP carries the request to Microsoft Graph and the response back to Claude. The email content does not move into a Stratium database. The message stays in your tenant. The same applies to your calendar, your sent items, your transcripts and your CRM. Cadence stores the operating data it needs to run the plan. The underlying content lives in the systems you already trust.
It avoids lock-in. Because MCP is open, the connector work we have done is portable in principle. If a future version of Cadence runs on a different reasoning engine, the connector layer comes with it. You are not signing up for a closed protocol that only Stratium and Anthropic understand. You are signing up for the protocol the industry has chosen.
It shortens IT due diligence. When the IT buyer asks how Cadence reads the inbox, the answer is one sentence and one link. MCP, through your existing Microsoft 365 OAuth flow, with scopes documented and data residency unchanged. The conversation moves on. Most IT teams sign off the architecture in a single working session because there is nothing hidden in it.
Where your data lives, and where it does not
Your data lives in the systems you already use. Outlook stays in Microsoft 365. Google Calendar stays in Google. Salesforce stays in Salesforce. Transcripts stay in Fireflies or Otter or wherever you store them. Cadence reads through MCP, surfaces the result, and the source content stays in place.
The operating data Cadence does store is small and specific. Your plan. Your work items. Your strategic drivers. Your focus blocks. The metadata Cadence needs to run the rhythm. All of it sits on Microsoft Azure West Europe by default, tenant-isolated, encrypted in transit with TLS 1.2+ and at rest. Sovereign hosting is available for organisations that need it.
The content of your inbox, your calendar and your transcripts is not stored by Cadence. The routine reads at the moment it runs, the insight is extracted, the source stays put. You can confirm this on the Trust page. Any IT review can verify it against the data flow diagram we publish.
The IT due diligence answer
The short version for an IT leader being asked to approve Cadence.
- Cadence runs on Claude through Anthropic's commercial API, under Anthropic's published data protections.
- Connections from Claude to your systems use MCP, an open standard, with OAuth scoped to the minimum the routine requires.
- Cadence operating data sits on Azure West Europe by default, tenant-isolated, encrypted, GDPR compliant. Sovereign hosting available.
- SOC 2 Type II is in flight.
- The source content in your inbox, calendar, transcripts and CRM stays where it is.
Most IT reviews close in a single working session.
Why we did not pretend to be an AI company
Stratium is not an AI company. We are an executive operating layer. We are the discipline of how a senior leader's day should run, expressed in product. The reasoning is done by an engine we did not build, and we are not embarrassed to say so. We picked the best one for the job, we use it well, we name it on every page where the IT buyer needs to know, and we ship the operating layer underneath that justifies the price.
That position holds up three ways.
Commercially, buyers are tired of products that sprinkle "AI" across the marketing and produce mediocre output. Naming the model is now a credibility signal, not a risk. Buyers prefer to know.
Technically, as Claude gets better, Cadence gets better. As MCP adoption broadens, Cadence reaches further. We inherit the curve of the underlying industry and focus our own engineering on the operating layer that is genuinely our problem to solve.
Strategically, the next decade of executive software will be defined by which products honestly distinguish between reasoning and structure. Reasoning is becoming a commodity. A small number of model providers will supply it. Structure is the product. Companies that pretend otherwise will spend the next five years rebuilding the structure layer they should have built first.
Stratium is the structure layer. Claude is the reasoning. MCP is the connection. That is the architecture, and that is what we will keep saying.
Frequently asked questions
Is Cadence built on Claude? Yes. Cadence uses Anthropic's Claude as the reasoning engine, accessed through Anthropic's commercial API. Stratium provides the executive operating layer on top.
What is MCP? The Model Context Protocol. An open standard proposed by Anthropic in 2024, now adopted across ChatGPT, Gemini, Microsoft Copilot, Cursor and VS Code. It is how AI tools connect securely to the systems you already use.
Does Cadence store my data? Cadence stores only the operating data it needs to run the briefing and the plan. The content of your calendar, inbox, sent items and transcripts stays in the systems you already use. Cadence reads through MCP, surfaces the result, and the underlying content does not move.
Can I use Cadence with ChatGPT or Gemini instead of Claude? Not today. The Skills and scheduled task patterns Cadence relies on do not exist in the same form on the other two platforms. The MCP plumbing is shared across the industry. The operating model Cadence is built around is Claude-specific. As that pattern broadens, our position may broaden with it.
Where is Cadence hosted? Microsoft Azure West Europe by default, with tenant isolation, TLS 1.2+ in transit, encryption at rest, and GDPR compliance. Sovereign hosting is available. SOC 2 Type II is in flight.
Who can use Cadence? Senior operating leaders (SVPs, MDs, CTOs, COOs, GMs, C-suite) and the EAs, Chiefs of Staff and team coordinators who support them. Cadence Mac runs on macOS. Cadence Web runs in any modern browser.
For how Cadence runs an executive day in practice, see How it works. For the full security and data architecture, see Trust and security. For the systems Cadence connects to, see Connected stack.