DOT: Digital Chief of Staff

DOT: Digital Chief of Staff

DOT: Digital Chief of Staff (referred to throughout this page as “DOT”) is a personal AI agent whose purpose is to automate the everyday administrative tasks that fill up your week, by creating documents, calendar entries, and emails on your behalf. It combines language-model reasoning with a library of skills that connect to the apps and tools you already use every day, through an open protocol layer called Model Context Protocol (MCP).

Where a conventional chatbot can only converse, this agent acts. Ask it once to handle a recurring obligation — sending a weekly status email to your team, drafting a meeting invite for a client and adding the event to your calendar, putting together a proposal document with the latest project numbers, reminding you to book your physiotherapy appointment, confirming a hotel reservation — and the agent carries out the multi-step workflow end-to-end. It reads the necessary context from your connected services, drafts whatever artefacts are needed (an email, a calendar entry, a Drive document), and either delivers them directly or queues them for your review.

Purpose

DOT: Digital Chief of Staff is an AI agent whose purpose is to automate the everyday administrative tasks that fill up your week, by integrating with the apps and tools you use every day.

The everyday work DOT is designed to take off your plate falls into a few familiar categories. Most weeks, a typical user will lose between three and ten hours to tasks that are necessary but mechanical — drafting the same kind of email to the same kind of recipient, scheduling the same kind of meeting, generating the same kind of document, chasing the same kind of confirmation, summarising the same kind of data into the same kind of weekly report. None of that work is hard, but it consumes time, attention, and willpower, and the cost of it compounds across a month and a year. DOT exists to give that time back.

Concretely, DOT automates everyday tasks across these categories:

DOT does all of this by sitting between you and the services your work and personal life already live in — Gmail, Google Calendar, Google Drive, Strava, your health and fitness apps, and other tools you add over time. You describe a recurring task once: what kind of email or document or calendar entry should be produced, who should receive it, and what should trigger it (a fixed date, a recurring schedule, an inbound message, or an explicit instruction). The agent owns the execution from then on. When the trigger fires, DOT reads the necessary context from your connected services, drafts the artefact the task requires, and either delivers it on your behalf or queues it for your one-tap approval.

The integration layer that makes this possible is the Model Context Protocol (MCP) — an open protocol that lets a language model securely call defined operations exposed by an external service. The protocol is what gives DOT the bridge it needs to read your calendar, draft your email, and create your documents. It is the difference between an AI that can describe the right action and an AI that can perform it. The protocol is also why the hosting subdomain for this site is mcp.hyani.net — DOT was built on, and around, MCP.

DOT began as a personal experiment to find out how much weekly admin work could be safely delegated to a language model with the right tool access. It quickly became clear that the limit is not the model — current language models are easily capable of drafting a competent status update, a coherent proposal, or a sensible meeting agenda. The limit is integration. A chatbot can write a perfect message but cannot send it; it can describe an ideal calendar entry but cannot create it; it can describe the right document but cannot put it in Drive. The MCP integration layer is what closes that gap, and is the reason DOT can act rather than merely converse.

DOT is designed for individual, personal use — one human delegating their own routine work to one agent, operating against the human's own accounts on connected services. It is not a multi-tenant SaaS, not a data-aggregation product, and not a marketing tool. The relationship is strictly between you, the agent acting on your behalf, and the services you have connected to it.

Design principles

What the agent does

The agent's core job is to take repetitive work off your plate by producing the actual artefacts that the work requires:

About the MCP integration layer

MCP (Model Context Protocol) is the integration layer that connects DOT to third-party applications and systems. It is an open protocol that lets a language model securely call defined operations exposed by an external service — the technical bridge that turns a passive chatbot into an agent that can read your calendar, draft your email, and create your documents.

For each external service DOT supports, a small dedicated MCP server exposes a defined set of operations. A Google Workspace MCP server, for example, exposes operations such as composing and sending Gmail messages, listing and creating Calendar events, and creating or editing Drive documents. DOT invokes those operations using the OAuth credentials you have granted to that specific service, and only those credentials. Adding support for a new service means adding a new MCP server — the core DOT agent does not change, and the new server has no access to any other service's data or credentials.

This design has two consequences worth understanding:

What it integrates with

Purpose for which DOT requests access to your data

To perform the automation tasks described on this page, DOT needs to read and act on data inside the services you choose to connect. Each access request has a single, narrow purpose tied to a specific user-facing feature. This section sets out, for each category of data, the exact purpose DOT has in requesting it — so you can see, and Google can verify, that the access requested is limited to what is required to deliver the user-facing features.

Gmail — read access. Purpose: to locate and read the specific email thread you have referred to (for example, “the email from procurement last Tuesday”) so the agent can ground its drafts and summaries in the actual conversation. Without this access, the agent can describe what an email reply might look like but cannot tailor it to the message you have asked it to work with.

Gmail — compose and send access. Purpose: to create the email drafts you have asked the agent to write, save them to your Gmail drafts folder for review, and (only for workflows where you have explicitly opted in) send them on your behalf. Without this access, the agent can write a message body but cannot produce a real Gmail draft or send a message you have approved.

Google Calendar — read access. Purpose: to find available time, look up events you have referenced (“the call with Sarah next Tuesday”), and check for conflicts before scheduling something new. Without this access, the agent cannot reliably propose a meeting time or pull context for a meeting prep brief.

Google Calendar — write access. Purpose: to create, update, or remove calendar events at your request, including attaching invitees, location, video-conference link, agenda, and description. Without this access, the agent can describe a calendar entry but cannot put it on your calendar.

Google Drive — read access (file metadata and content for files the user explicitly references). Purpose: to open and reference documents you have asked the agent to work with — a brief, a template, a prior proposal, a meeting notes document — so the agent can use your existing materials as input when drafting new artefacts. The application does not request broad Drive scopes; it requests only the access required to read files you have specifically pointed it to or that are shared with the workflow you set up.

Google Drive — create and edit access. Purpose: to write out the documents the agent has drafted on your behalf (proposals, meeting invites, agendas, summaries, weekly digests, training reviews) and place them in your Drive so you can find, edit, and share them. Without this access, the agent can describe a document but cannot deliver one.

Strava — activity read access. Purpose: to pull your recent training data when you have asked the agent to summarise a week, draft an email update to a coach, generate a recovery review, or plan workouts around your activity history. Without this access, the agent has no source data for training-related workflows.

Health and fitness platforms — metric read access. Purpose: to pull workout, sleep, recovery, and body-metric data when you have asked the agent to surface trends, generate a weekly health review, or flag concerning patterns. Read-only; the agent does not write health data back to these platforms.

Scope minimisation as a rule. The same principle applies across every integration: the agent requests the narrowest scope that lets it perform the operations you have explicitly enabled, and nothing wider. Workflows that only need calendar read do not trigger a calendar write request. Workflows that only need to draft email (never autosend) can be configured to request only the compose scope. Any future integration is wired the same way — one MCP server per service, with declared operations and minimum scopes.

What DOT does not request. DOT does not request access it does not need for a declared user-facing feature. It does not ask for full-Drive enumeration when a single file reference is sufficient. It does not ask for the entire Gmail mailbox when a single thread reference is sufficient. It does not ask for “profile” or “contacts” scopes that do not contribute to a workflow the user set up.

How it works

You describe what you want in plain language. The assistant chooses the matching skill, calls the underlying MCP server, and reports back with the result — or asks a clarifying question if needed. Each connector requests only the OAuth scopes it needs to fulfil the task, and you can revoke access at any time from the originating platform. For Google Workspace integrations, that's the Third-party apps with account access section of your Google Account settings.

Example tasks

Single instructions that produce concrete artefacts:

Scheduled, recurring workflows

The agent is designed for tasks that repeat. You describe the pattern once and the agent carries it out on schedule, producing the artefacts each time:

How your data is used

Data retrieved from connected services through DOT: Digital Chief of Staff (via the MCP integration layer) is used only to fulfil the request you made. It is not transferred to third parties, not used to train language models, not used to serve advertising, and not sold. DOT: Digital Chief of Staff's use and transfer to any other application of information received from Google APIs adheres to Google's API Services User Data Policy, including the Limited Use requirements. See the Privacy Policy for details on what is logged, retained, and how to exercise your rights, and the Terms & Conditions for the rules governing use of this service.

Who it's for

Individuals who want a single conversational interface for the work and personal services they already use, without context-switching between a dozen apps. DOT is a personal-use desktop tool; access is granted per user account through the standard OAuth consent flow of each connected service.

Contact

For questions about the application, this page, or to request information about data associated with your account, write to: roryhw [at] gmail [dot] com.