Sydney, Australia
Albireo builds infrastructure for AI-generated education. Documents, slides and photographs go in. Structured, cited, reusable context comes out — and from that, courses and assessments.
// One document in. Reusable context out.
const job = await achird.ingest({
source_id,
profile: 'knowledge-source',
});
// Every unit carries a page and a bounding box.
const { units } = await achird.context.list({
kind: 'concept',
});
// Same units, any output. Extraction is paid once.
await achird.generate({ recipe: 'course', units });
await achird.generate({ recipe: 'mcq', units });The problem
Every organisation sits on textbooks, past papers, slide decks and manuals. Turning that into teaching material is still mostly manual — and the tools that automate it tend to make the same three mistakes.
Most tools read a document straight into a finished artifact. Want flashcards from the same PDF tomorrow? You upload it again and pay to read it again.
Pipelines that only lift the embedded text layer fail silently on scans and handwriting — returning an empty result that looks like a clean run.
When a lesson makes a claim, someone eventually asks where it came from. Without page-level provenance there is no answer, and no copyright defence.
The engine
Between the raw file and the finished course sits a durable layer of context units — small, self-contained, individually cited. Everything downstream reads from it, so the expensive part happens exactly once.
Model Context Protocol
Achird speaks MCP, so the assistants your team already uses can read the corpus directly — Claude, Claude Code, Cursor, or anything else that speaks the protocol. No new interface to learn, and no copy-pasting between tools.
{
"mcpServers": {
"achird": {
"url": "https://mcp.achird.com",
"headers": {
"Authorization": "Bearer ach_live_…"
}
}
}
}
// Your corpus is now readable by the assistant,
// with a page and bounding box on every claim.Every result carries its provenance. When an assistant answers from your material it can point at the page it came from — which is the difference between a citation and a guess.
Products
Achird is the engine. LearnRadar is what it looks like aimed at one market — and the first customer holding it to a production standard.
The Source → Context → Generation engine, offered as an API. Ingest documents, get back typed and cited context units, then generate courses, lessons and assessments from them — exportable to any LMS.
Eta Cassiopeiae — a gold-and-purple double star
An AI academic coach for NSW primary students preparing for Opportunity Class, Selective High School, Scholarship and NAPLAN. It finds the gaps, teaches the concept, then practises what actually matters.
Contact
Building something that needs to turn source material into teaching material? We are talking to early partners for Achird.
hello@albireosoftware.com.au