"Built on Claude" means a tutoring app uses Anthropic's Claude models as its underlying AI, then wraps them in its own teaching rules — a hint ladder, curriculum content, safeguarding responses — rather than letting a pupil chat with the model freely. The model choice matters less than what's built around it.
What does "built on Claude" actually mean?
Claude is a family of AI models made by Anthropic, in the same category as OpenAI's GPT models or Google's Gemini. On its own, at claude.ai, Claude is a general-purpose assistant: ask it to solve an equation or write an essay and it generally will, because it's designed to be broadly helpful. A tutoring product "built on Claude" is a separate application — often connected via the Model Context Protocol (MCP), a standard way for apps to plug additional tools and rules into an AI model — that adds subject content, a teaching structure and guardrails on top, so the underlying model behaves like a tutor rather than a generic assistant.
How is a Claude-based tutor different from using Claude directly?
The practical difference is restraint and structure, not raw capability. Claude used directly will typically answer a homework question outright because nothing stops it from doing so. A tutoring app built on Claude adds rules that a general chat session doesn't have: a hint ladder that escalates gradually instead of jumping to the answer, subject content matched to a specific curriculum (for example, UK KS3), a safeguarding response if a pupil says something concerning, and often a way for a parent to see what was covered. None of that comes from the model itself — it's the layer the tutoring app builds around it.
What should a parent check before choosing a Claude-based tutor?
The label "built on Claude" tells you the underlying model, not the teaching quality. Before trusting a product, check what it actually does differently from raw chat: does it withhold answers and ask questions instead, or does it just add a friendly tutor persona on top of an assistant that will still solve the problem if pushed? Check whether it's matched to the UK curriculum rather than a generic or US syllabus, whether it has a stated policy on child data and third-party model training, and whether it gives you, as the parent, any visibility into sessions — a weekly summary is worth more than a marketing claim about the model underneath.
Does the choice of underlying model matter at all?
It matters less than the application built on top, but not zero. Different model families vary in reasoning quality, how well they follow complex instructions (like "never state the final answer"), and cost, which affects what a product can offer at a given price. What it doesn't tell you is whether the product is a good tutor — that depends entirely on the teaching design, subject content and safeguards the developer has built, which two apps using the exact same underlying model could implement completely differently.
How does a Claude-based tutor typically reach a child?
Because Claude-based tutoring apps often connect via MCP, families can sometimes add them as a "connector" inside claude.ai or Claude Desktop itself, alongside using the app's own dedicated interface. Either way, the pupil is interacting with the tutoring app's rules and content — the hint ladder, the curriculum coverage, the safeguarding behaviour — layered on top of the model, not with an unrestricted version of Claude.
Frequently asked questions
Is a tutor built on Claude the same as using Claude itself?
No. Claude used directly at claude.ai is a general-purpose assistant that will typically answer questions outright. A tutoring app built on Claude adds its own teaching rules, curriculum content and safeguards on top, so it behaves differently from open-ended chat.
Does "built on Claude" mean it's aligned to the UK curriculum?
Not automatically. The underlying model is the same regardless of curriculum; alignment to the UK KS3 syllabus, exam boards or grading depends entirely on the content and prompts the tutoring app's developer has built, so this should be checked per product rather than assumed.
Why would a tutoring company build on Claude rather than train its own AI?
Building a foundation model from scratch is enormously expensive and slow. Using an existing model like Claude via an API lets a tutoring company focus its effort on the parts that actually make it a tutor — curriculum content, hint design, safeguarding, and parent reporting — rather than reinventing general-purpose language understanding.
What is MCP, and why does it matter for Claude-based tutors?
MCP (Model Context Protocol) is a standard that lets an AI model connect to external tools and data in a structured way. For a tutoring app, it's often the mechanism that lets a family add the tutor as a connector inside claude.ai or Claude Desktop, so the pupil can access tutor-specific tools rather than a plain chat window.
aitutors.me is a Socratic, UK-curriculum tutoring connector built on Claude for KS3 maths, science, English and humanities — see aitutors.me.