TRAX MCP Server

AI assistants have become remarkably good at reasoning, writing and analysing. What they lack is access to your world. Ask about your learners' progress, and it can only tell you it doesn't know them. TRAX LRS 4.0 changes that: your Learning Record Store now speaks MCP, the language AI agents use to connect to real systems.

MCP in two minutes

Think of MCP — the Model Context Protocol — as a universal plug between AI assistants and applications. An application publishes a set of tools ("query the statements", "record a statement"…). An assistant plugged into it discovers these tools and uses them on its own, whenever your request needs them.

You don't write code or learn a query language. You just ask:

"Which learners completed the onboarding course this month, and how did they score?"

The assistant finds the right tools, queries your LRS, and answers in plain words. MCP has become the industry standard for connecting AI to real systems: it is supported by Anthropic's Claude, OpenAI's ChatGPT, Google's Gemini, Mistral's Le Chat and Microsoft Copilot, among a fast-growing list of AI agents — and from now on, by TRAX LRS.

Plugged in, in a few minutes

In the TRAX LRS web application, you create an MCP access key, then decide which stores it can reach and what it can do there: read the data, record statements, or both. You paste the key into your assistant's configuration, and that's it. One key can reach several stores; each one keeps its own permissions.

The documentation walks you through the setup, step by step.

Three things you can do now

1. Generate realistic xAPI data in seconds

Designing an xAPI profile or testing a reporting chain needs data — consistent, realistic, and compliant. Writing it by hand is tedious. Ask instead:

"Record a typical cmi5 session in the test store: initialized, passed with 85%, completed, terminated."

The assistant writes the statements and TRAX validates every one of them against xAPI, exactly as it would for a real learning content.

Why it matters: test scenarios that used to take hours take a sentence. You see immediately whether your statements are valid, and you can seed a whole learning context — a class, a course, a semester — to try out your analytics before the first real learner arrives.

2. Give your AI tutor a memory of record

AI coaches and tutors are appearing in every learning platform. Their conversations, their exercises, their assessments usually vanish in the chat history. With TRAX, the agent can record what happened as xAPI statements: the situation it set up, the learner's answers, the skills it observed.

Why it matters: the AI-driven part of the learning experience joins the rest of your learning data, in a standard format, in your LRS. It becomes auditable, comparable over time, and usable by every tool you already rely on — not locked in a chat window.

3. Ask your LRS anything

xAPI data is rich, but it is rarely easy to read: identifiers, verbs, nested contexts. Connected to TRAX, an assistant reads the statements for you — those of your learning contents, and those your AI agents recorded — and turns them into answers.

"Summarise the learning journey of Jane over the last quarter: what she did, when, and how her results evolved."

Why it matters: no dashboard to build in advance, no query to write. Managers, trainers and designers get answers in their own words — a narrative, a comparison, a trend — and the assistant can go further and help you understand what your data actually contains before you decide how to analyse it at scale.

Safe by design

Opening data to AI deserves caution, and TRAX was designed with it:

• You decide the scope. A key only reaches the stores you attach to it, with the permissions you chose.
• Everything is traceable. Each statement recorded by an agent is tagged with its MCP origin, and you know which key wrote it. Revoking a key cuts its access instantly, while keeping the history for audit.
• Nothing gets erased. Agents can record statements — which are never modified nor deleted — but they can't overwrite your learning content's data.
• No admin powers. Stores, accounts, permissions and settings are out of reach: an agent can never grant itself more than you gave it.

An LRS that works with your AI

xAPI was designed to capture learning wherever it happens. With MCP, TRAX LRS makes that data available where more and more work happens: in conversations with AI agents. Create a key, connect your assistant, and start asking.