Lyceum is the agentic AI tutor U.S. universities and high schools deploy at scale. It plans multi-step study paths, calls real tools, retrieves from your textbooks, and stays with each student across semesters — orchestrating Claude Opus 4.7, GPT-5, and Gemini 2.5 Pro under the hood. 12.4 B tokens / month. SOC 2 · FERPA · HECVAT cleared.
We sat down with deans, professors, TAs, and student-success offices at 30+ institutions and walked through three moments that most often break a semester. Each pattern below comes from real TA tickets and student help-desk logs.
The slides didn't go that deep, the textbook doesn't show the steps, no one's awake on Slack. 71% of undergraduate study frustration happens between 11 PM and 2 AM — exactly when faculty support is hardest to staff.
Faculty teaching multiple sections have under 9 seconds of one-on-one time per student per week. Whether you learn ends up depending on how brave you are, where you sit, and whether you got called on.
The week before finals, an average TA fields 380+ inquiries — 64% of which are variations on the same handful of high-frequency questions. Their time should go to deeper guidance, not retyping.
Lyceum isn't a chat box thrown at students. It's a full teaching stack that plugs into your existing LMS, sits inside your course context, and lets faculty review and shape its behavior.
SSO / SAML 2.0 / LDAP into your existing campus portal. Students sign in with their school credentials — no new account. Admins scope access by department, course, or year.
Sync syllabi, lecture decks, problem sets, video transcripts, assigned readings. Lyceum answers within your course materials — not from the open internet.
Students ask questions inside your LMS or a standalone portal. Lyceum walks them through with Socratic-style guidance — never just dumping the answer.
Instructors see class weak points, individual progress, and high-frequency questions — with auto-generated prompts for next week's lecture and individualized practice sets.
Lyceum isn't a chat box wrapped around an LLM. It's a full agentic system that plans multi-step study paths, calls verified tools, retrieves from your textbooks, and stays with each student across semesters — exactly the way a great human tutor would.
Given "I'm stuck on series convergence," the Agent doesn't just answer. It diagnoses prerequisite gaps, builds a 4-step path through the right concepts, and schedules a follow-up.
The Agent calls Python (sympy / numpy) to verify derivations, retrieves from your uploaded textbooks via vector search, runs Wolfram for symbolic math, and queries the curriculum graph — every call is logged and auditable.
The Agent remembers Maya struggled with LIATE in week 7, that she prefers visual derivations, that she always reviews on Sunday nights — and weaves all of it into how it teaches her in week 14.
Overnight, the Agent runs jobs without prompting — generates next week's practice set, drafts a class digest for Dr. Hartman, and pages the dean of students about a struggling cohort it identified at 02:14 AM.
plan → retrieve → reason → tool-call → verify → respond → wait → diagnose → re-plan → ... The student sees a teacher. Internally, an Agent is running.The Agent picks the right brain for each request — deep reasoning for proofs, fast clarifications for follow-ups, long-context for full-textbook ingestion. Multi-vendor by design — never a single point of failure for your students.
| Aa Model | ⚙️ Provider | ⊞ Routed for | ⌗ Share | ⚡ P50 |
|---|---|---|---|---|
🧠Claude Opus 4.7 |
Anthropic | 28% | 2.1 s | |
⚡Claude Sonnet 4.6 |
Anthropic | 34% | 820 ms | |
🌬️Claude Haiku 4.5 |
Anthropic | 14% | 320 ms | |
💻GPT-5 |
OpenAI | 11% | 1.4 s | |
📚Gemini 2.5 Pro |
9% | 1.8 s | ||
🦙Llama 3.3 70B |
Self-hosted · Meta | 4% | 240 ms |
We build the knowledge graph alongside each partner's department chairs — concept definitions, canonical problem types, common misconceptions. Every sub-topic is mapped to academic level, recommended textbook, and a corresponding practice set.
| Aa Subject | ⊞ Core modules | ⌗ Modules | ▤ Levels |
|---|---|---|---|
📐Mathematics |
Core
Advanced
AP / Test
|
52 modules | 9–14 |
⚛️Physics |
Core
Advanced
AP / Test
|
44 modules | 9–14 |
🧪Chemistry |
Core
AP / Test
|
38 modules | 9–14 |
🌱Biology |
Core
AP / Test
|
36 modules | 9–13 |
💻Computer Science |
Core
Advanced
AP / Test
|
48 modules | 9–16 |
📈Economics |
Core
Advanced
AP / Test
|
32 modules | 11–16 |
📖English & Literature |
Core
Advanced
AP / Test
|
30 modules | 9–14 |
🏛️History & Social Studies |
Core
Advanced
AP / Test
|
28 modules | 9–14 |
🤔Philosophy & Logic |
Core
Skills
|
22 modules | 11–16 |
🌍World Languages |
Languages
AP / Test
ESL
|
40 modules | 9–14 |
Lyceum maintains a living learning profile for every student — built from their question history, error patterns, mastery of prerequisite concepts, and preferred way of learning. Next session, it picks up exactly where they left off.
Every question maps to nodes in the subject graph, surfacing prerequisite gaps and patching them before they snowball into the next topic.
Concrete analogy, formal proof, hands-on derivation, visual diagram — Lyceum learns which one makes this student nod, and leans into it next time.
Solving too easily? Step up. Stuck twice? Step back to a more foundational practice problem. No student gets stuck against the same wall until they give up.
Backed by a forgetting-curve model tuned to each student's recall data, Lyceum surfaces a quick review problem right before they're about to lose a concept.
Lyceum's pedagogy was designed alongside classroom instructors: an AI that behaves like a restrained, question-asking teaching assistant — one that leaves the moment of "oh, I get it" for the student to find on their own.
When a student is stuck, Lyceum's first response is a clarifying question — not the next step. A typical problem walk-through cycles through 3–5 rounds of AI prompt ↔ student response before the student reconstructs the path themselves.
All STEM solutions are presented stepwise: each step has a motivation, an expression, and a sentence on why it works. Students can collapse any step and try to re-derive it before checking against ours.
When Lyceum detects a homework- or exam-pattern question, it slows down, leans on guidance, and writes the entire session into the student's learning record. Faculty can pull up the full transcript when grading.
Every concept explanation cites the page in your uploaded materials — not Wikipedia, not a different version of the textbook. The instructor stays the source of truth; Lyceum is a patient tutor pointing students back to the right page.
If a question falls outside the course materials, or Lyceum's confidence drops below threshold, it explicitly says so and routes the student to a human TA — instead of confidently making something up.
Every Monday morning, instructors get a digest: the three concepts the class is most stuck on, the students who need a personal check-in, and the topics worth revisiting in lecture — generated from real session data, with the instructor making the final call.
LMS, identity, procurement, legal, IT security — Lyceum is the only AI tutor you can stand up at an R1 university without rewriting your vendor management workflow. Single-tenant deployment available. Student data can stay on your network.
Built on the LTI 1.3 standard. Once configured, Lyceum auto-syncs courses, sections, instructors, students, and grade-passback. Roster updates at term boundaries are zero-touch.
Read integration docs →Public cloud, hybrid, and on-premise deployments. Audit logs, content moderation, and data export aligned with U.S. K-12 and higher-ed compliance frameworks.
Aggregate results from 12 partner institutions across the 2024–2025 academic year. Each metric compares students with Lyceum access to a matched control cohort.
Students with Lyceum logged meaningfully more effective study time outside scheduled class — most of it in the gaps between classes.
In historically weed-out courses (intro calculus, organic chemistry, circuit analysis), failure rates dropped 27% — biggest gains in the bottom quartile.
Graduate TAs reported a 42% drop in repeated low-complexity questions — freeing them up for deeper one-on-one work and grading.
In Spring 2025, the math department rolled Lyceum out to four sections of Calculus II — about 600 undergraduates — as a complement to traditional office hours, the evening study center, and TA shifts. Excerpts from the end-of-semester review below.
The biggest concern going in was simple: would students stop showing up to office hours once they had an AI? The end-of-semester data was the opposite. Office-hour attendance went up — and the questions students brought in were noticeably deeper and more research-oriented.
"The AI took the routine questions. What was left in office hours was exactly the kind of conversation we got into this profession for." — the project lead, in the department's term-end review.
This is the feedback we care about most: not a polished talking point, but each role describing — in their own words — the specific thing Lyceum is solving for them.
"My biggest worry going in was data privacy. Their FERPA documentation was more thorough than the policy our own office wrote. Private-cloud was running in under a week."
"I've taught Calculus for eleven years. This is the first time I felt like I knew more about how my class was thinking before I walked into lecture than after. The Monday digest writes my opening five minutes."
"Used to be I was answering messages until midnight in finals week. Now Lyceum picks up the routine ones and I get to spend that hour with the students who are stuck on something interesting."
"What surprised me most is that it doesn't just dump the answer. One time it asked me, 'why can't you just substitute here?' and I sat with that for a full minute. That was the first time something other than a person had actually taught me."
"About 60% of our students don't have anyone at home who can help with calculus or AP Bio. With Lyceum, they have someone still up at 9 PM willing to help them through it. That's a bigger deal here than any test-score chart."
"What we value most isn't test-score lift — it's that Lyceum can flag a struggling student at 2 AM, so our advisors can reach out by 9 the next morning, before things spiral."
Lyceum is sold as a flat institutional license. Every plan includes unlimited students within scope, the full subject library, all six frontier models, faculty insights, and our security packet. Pricing scales with deployment depth — not headcount.
These are the eight questions provosts, IT directors, deans, and student-affairs offices ask us most often. Have something else? Talk to a solutions architect →
A 30-minute demo is enough to tell whether Lyceum fits your campus. We'll walk through three of the most common student help-requests at your institution — and show you exactly how Lyceum handles them.
We'll walk through three of the most common student help-requests at your institution and show you exactly how Lyceum handles them.
Lyceum sign-in is via your institution's SSO (SAML / OIDC). Your IT administrator provisions access through Canvas / Blackboard / your campus IdP. Need help getting set up?
Thanks, there — we've received your request. A solutions architect will email you within 4 business hours with a couple of times to meet.