Moodle AI Agent
A Moodle AI agent is an AI tutor that lives inside your Moodle course and answers student questions using only the course material, citing the unit and page it came from. Students ask on the page where they already study, at any hour, with no new platform and no new login. E-comienza builds and deploys these agents. One runs in production today at Estudios Tributarios.
What a Moodle AI agent is
A Moodle AI agent is a language model agent connected to one specific course and restricted to that course material. It appears inside Moodle as part of the course, takes a question in plain language, finds the relevant passage in the curriculum, and answers with a reference to where that passage lives.
It is not a general chatbot pointed at a school. A general chatbot will answer anything, including things your program never taught and things that are simply wrong. A course agent is built the other way around: the curriculum is the only source it is allowed to draw on, so the answer a student reads is the answer the course gives.
The practical effect is a tutor that is available when the teacher is not. Students study at night, on weekends and the day before an exam. That is when the doubt appears and that is when nobody is there.
It lives inside the course
The agent is integrated natively into Moodle. The student opens the course, sees the agent where the course content is, and asks. There is nothing to download, no separate account to create and no second password to remember.
That matters for the people who run the LMS as much as for students. A campus that already administers Moodle does not need another platform, another vendor login or another support queue. The agent is part of the course you already maintain.
It also keeps the student in context. A question asked while reading Unit 4 is answered against Unit 4, in the same tab, without the student losing their place.
It answers only from the course material
The agent uses retrieval-augmented generation, usually shortened to RAG. Before answering, it searches an index built from your curriculum, pulls the passages that match the question, and writes the answer from those passages. The model supplies the language. The course supplies the content.
This is what keeps the answer aligned with what you actually teach. If your program defines a term one way and the wider internet defines it another way, the agent uses your definition, because your unit is the source it retrieved.
It also makes the agent auditable. Every answer traces back to a passage in material you wrote and approved, which is the review a program director needs before signing off on an AI tutor.
It cites the unit and the page
Every answer names its source. At Estudios Tributarios a student asks what the requirements are for VAT exemption. The agent answers with the three points set out in Unit 4 and states the source: Unit 4, page 23.
The citation does two jobs. It lets the student verify the answer and read the full passage in context, which is the behaviour you want from someone studying. And it lets a teacher check the agent quickly. If the cited page says what the agent said, the answer is sound.
A cited answer also teaches students where to look next time. Over a term, that is a different habit than getting an unattributed paragraph from a general chatbot.
It says when something is not in the material
When a question falls outside the curriculum, the agent says so. It does not improvise an answer to fill the silence. A student asking what will be on next week exam is told that this is not in the course material and pointed back to the teacher.
This is the single behaviour that decides whether an AI tutor can be approved. The risk with a general model is not that it is unhelpful. It is that it is confidently wrong, and a student believes it because it appeared inside the official course.
Refusing to answer outside the source is a design decision, not a limitation we apologise for. The agent covers the course completely and stops there.
Available 24/7, with no new login
The agent is available at any hour. Students in different time zones, adults studying after work and anyone preparing the night before get the same response as someone asking at midday.
Nobody installs anything. Students use the Moodle account they already have. Administrators keep a single platform. Teachers do not moderate a second inbox.
The teaching team keeps the questions that need a person. The repeated questions, the ones that come back with every cohort and every forum thread, are handled where they are asked.
How a deployment works
There are three steps. First, we connect the course. We get access to the Moodle course and agree on exactly which material is in scope, so the boundary of what the agent may answer is explicit from the start.
Second, we index the material. The curriculum is processed into a searchable index, keeping the structure of units and pages, which is what makes a citation like Unit 4, page 23 possible rather than a vague reference to the course.
Third, the agent goes live inside Moodle. It appears in the course, answers from the index, cites its sources, and declines questions outside the material. From there we tune behaviour against real student questions.
Estudios Tributarios in production
Estudios Tributarios, at estudiostributarios.com, runs an E-comienza Moodle agent today. It tutors students 24/7 using exclusively the course curriculum, built on RAG over that curriculum.
The exchange we show publicly is a real pattern from the product. The student asks about VAT exemption requirements. The agent gives the three requirements from Unit 4, cites Unit 4, page 23, and offers to explain the next concept.
If you are comparing vendors for a Moodle AI project, ask each of them to show you one running. That is the question that separates a demo from a deployment.
Frequently asked questions
What is a Moodle AI agent?
A Moodle AI agent is an AI tutor that runs inside a Moodle course and answers student questions using only that course material. It appears on the course page, replies in plain language at any hour, and cites the unit and page each answer came from so students and teachers can verify it.
Can AI answer student questions using only my course material?
Yes. A course agent uses retrieval-augmented generation, or RAG, to search an index built from your curriculum and write the answer from the passages it finds. The curriculum is the only source it draws on, so answers match what your program teaches rather than what a general model assumes.
What happens if a student asks something the course does not cover?
The agent says the question is not in the course material and points the student back to the teacher. It does not improvise an answer. That behaviour is deliberate, because a confident wrong answer shown inside an official course is more damaging than no answer at all.
Do students need a new app or a new login?
No. The agent is integrated natively into Moodle, so students use the account they already have and ask on the same page where they study. There is nothing to install, no second password, and no separate platform for administrators to run or support.
How does the agent cite its sources?
Each answer names the unit and page it came from. In the Estudios Tributarios course, an answer about VAT exemption requirements cites Unit 4, page 23. The citation lets the student read the full passage in context and lets a teacher verify the agent quickly against the original material.
How is a Moodle AI agent deployed?
Deployment has three steps. E-comienza connects the Moodle course and agrees which material is in scope, indexes that material while keeping its unit and page structure, and then makes the agent live inside the course. Behaviour is tuned afterwards against real student questions.
Is the Moodle agent available outside class hours?
Yes. The agent answers 24 hours a day, seven days a week. Students studying at night, on weekends or across time zones get the same cited answers from the course material as anyone asking during the day, without waiting for a teacher to come online.
