Everyone’s talking about “AI agents” and “MCP.” Almost no one is telling practitioners what they are, or what to do with them.
AI agents & MCP, for your practice.
Learn what they actually are — in plain language, no code — then apply them to the work you already do: intake, reading the ledger, drafting protocols, and the admin that eats your evenings. Vendor-neutral, GDPR-first, built on the tools you already use.
Start here
What they are, without the jargon.
Two words get thrown around as if they’re the same thing. They’re not. Here’s each one in plain English, with what it looks like on a Tuesday in your practice.
An AI agent
A general AI tool you've given a job, a boundary, and a memory. Instead of answering one question, it runs a small task end-to-end — read this, draft that, flag anything odd — then hands it back to you to approve.
In practice
A client sends 12 days of notes. An agent reads them, names the recurring patterns, and drafts a one-page brief. You edit and sign it.
MCP (Model Context Protocol)
A common plug that lets an AI tool safely read from — and write to — the tools you already use, without copy-paste. Think of it as a standard socket between the AI and your calendar, notes, and forms.
In practice
Your assistant can read a client's intake form from your notes app and add the follow-up to your calendar, because both speak the same plug.
The two together
The agent is the worker; MCP is how it reaches your tools. One without the other is a clever chat window. Both together is quiet leverage that lives inside the software you already pay for.
In practice
One agent, wired via MCP to your intake form, notes, and calendar, turns a new enquiry into a prepped first session — while you sleep.
How we teach it
Judgement is the moat, not the tool.
Method, not platform
Nothing to install.
You learn an architecture and run it inside the general AI tools and apps you already pay for. No portal for clients to log into.
Tool-agnostic
Outlives any vendor.
Agents and MCP connections are configured against a standard, so your setup survives a model change, a price change, or a new owner.
EU-built · GDPR-first
The client's data stays in their tools.
MCP reads where the data already lives. We never see, store, or train on client conversations.
Applied to your practice
Six jobs an agent can quietly take off your desk.
None of these diagnose, prescribe, or hold the relationship. Each one hands you back the hours you currently spend on data archaeology and admin.
01
Client intake, done before session one
An intake agent turns a messy first message into a structured brief: baseline, medications noted, working hypotheses, open questions. You start at judgement, not data-gathering.
02
Reading the ledger
Twelve days of food, sleep, mood and symptom notes, synthesised into named patterns with clinical relevance scored — so between-session review takes five minutes, not fifty.
03
Protocol first drafts
An agent drafts a starting protocol from the brief and the evidence you trust — mechanism plus caveat, never a claim without one. You refine; you never rubber-stamp.
04
The admin that eats your evenings
Follow-up notes, session summaries, resource emails, rebooking nudges — drafted by an agent wired to your calendar and notes, waiting for one click of approval.
05
Out-reading the chatbot in the room
When a client arrives with a confident AI answer, you have your own agent and method to validate, reframe, or correct it on the spot — capable, not threatened.
06
One brief, many tools
MCP lets the same agent reach your form, your notes, and your calendar without you exporting anything. The client's data stays in the tools they and you already own.
Three ways in
Learn it, apply it, or have it built.
Learn
Understand what you're switching on
Start with the plain-language method: what an agent is, what MCP does, where the line sits between the reading layer and your judgement. The Practitioner Course covers it without a single line of code.
See the Practitioner Course →Apply
Build your first agent from a template
Toolkits give you the exact agent briefs, MCP connection maps, and guardrail checklists — written for the tools you already use, so you can stand one up this week.
Browse the toolkits →Delegate
Have the whole stack built for you
Short on time? Our done-for-you builds configure the agents, wire the MCP connections, and hand you SOPs and a QA-tested launch. You approve; we build.
See done-for-you builds →Ethics & boundaries
The agent is the reading layer. You are the practice.
What an agent does NOT do
- Diagnose, prescribe, or treat
- Act without your approval
- Hold the therapeutic relationship
- Send client data anywhere you didn’t authorise
What an agent does extremely well
- Synthesise days of ledger into one brief
- Draft protocols, notes, and follow-ups for review
- Reach your existing tools via MCP, no copy-paste
- Flag interactions and anomalies for your eyes
Start with the free 10 days.
The free 10-day challenge is the same architecture, condensed. Run it on yourself first — then decide whether to learn the full method, grab the toolkits, or have the stack built for you.
App stacks
AI + the apps you already use
Practical stacking guides for practitioners — combining chat assistant, notebook, and agent with the wellness tools your clients already have.