A few years ago, building a tool that could answer questions, role-play a difficult conversation or tutor you in German meant learning to code, renting servers and paying for every request. At Arizona State University, a student can now start the same kind of project from a browser, using a single university login.
The tool behind that change is CreateAI Builder, ASU’s no-code platform for creating, refining and sharing custom AI experiences. In August 2026, ASU opened it so students can build their own projects instead of only chatting with ones their instructors created.
This guide explains what you can realistically build, how to get started, how the pieces of an agent fit together and how to test your work before anyone else uses it. It also covers the spending limits, data rules and academic-integrity questions that students tend to discover too late.
Everything below is drawn from ASU’s own pages, news releases and release notes as of November 2026. CreateAI changes quickly, so treat menu names and limits as a snapshot and confirm them in the platform before you rely on them.
Agents, Assistants and Projects: Getting the Words Right
ASU’s own materials use several overlapping terms, and it helps to sort them out before you start building.
The CreateAI Builder pages talk about AI assistants, AI projects and AI experiences. Those terms describe the same basic thing: a configured AI tool with your instructions, your source material and a chosen model behind it.
The word agent usually means something a step further: an AI that can take several actions toward a goal, such as searching, reading documents, calling a connected tool or working through a multi-step task. ASU’s release notes reference agentic chats, agent web search and expanded support for the Model Context Protocol (a standard way to connect AI tools to other systems), which shows the platform moving in that direction.
Important: ASU’s public Builder pages describe “AI assistants” and “projects” rather than a distinct “agent” feature. In this article, “custom AI agent” means a purpose-built assistant you configure in Builder, with agent-style abilities only where ASU’s documentation says they are available to you.
That distinction matters because it keeps your expectations honest. A well-built assistant with good instructions and good sources solves most student problems. Heavier agent behavior is a bonus, not a prerequisite.
What CreateAI and CreateAI Builder Actually Are
CreateAI is ASU’s flagship AI toolkit, developed in-house. ASU describes it as a way to create, refine and share AI experiences tied to courses, research and daily operations.
The main components
- CreateAI Builder: The no-code tool for individuals to build and share assistants, from scratch or from templates.
- CreateAI Compare: A tool that helps you choose among the available models, with an icon marking more energy-efficient options.
- CreateAI Platform: The layer ASU engineers and developers use to build enterprise-level solutions, including automated workloads.
How big is it?
ASU reports rapid growth. A tech.asu.edu feature published in late 2026 said more than 57,000 people had used CreateAI, with 4,325 builders, roughly double the number from a year earlier. The platform had processed 26 million messages through its automated workloads system.
Classroom use is growing too. The same feature reported CreateAI tools in 715 classrooms serving more than 42,000 enrolled students in fall 2026, and 37,000 students using the platform that semester.
Quick Fact: ASU says an August 2026 launch made CreateAI available to all of its roughly 200,000 students, and that model costs averaged about eight cents per student across CreateAI projects over the year.
A Short Timeline: How Students Got Builder Access
| When | Milestone | What it meant for students |
|---|---|---|
| 2025 | CreateAI Builder launches university-wide for faculty and staff. | Students could use projects shared by instructors, often through sponsorship. |
| March 2026 | ASU News reports over 20,000 employees with access and 7,000+ experiences created. | Students worked on existing projects in coursework or built their own with sponsorship. |
| April 2026 | Builder gains access to locally hosted models through the AIR Platform, including Gemma 4 and Llama 4 Scout. | Faster, campus-hosted model options became available inside Builder. |
| August 2026 | Release notes say students can now build their own AI projects. | Students moved from users to builders. |
| September 2026 | CreateAI opens to students more broadly, with automatic shut-off at a spending limit. | Cost controls became part of the student experience. |
| October 2026 | Default monthly project cap set; accepted file types updated. | Clearer budgets and upload rules for builders. |
Who Can Use Builder, and What It Costs
Eligibility
ASU’s AI tools page lists CreateAI Builder as available to students, faculty and staff at no cost. The Builder page says student access begins once your student affiliation is active, typically about two weeks before your first day of classes, and ends when you are no longer enrolled.
One wrinkle: ASU’s own pages are not perfectly consistent. An FAQ on the AI tools page describes Builder as for faculty and staff, while the tool table lists students as eligible. Given the August and September 2026 release notes, students being eligible appears to be the current position, but confirm inside the platform with your ASURITE login.
Cost to you
ASU states that Enterprise Technology covers nearly all CreateAI usage, and that model prices displayed inside the platform are for transparency rather than charges to you. You do not need a credit card to start.
Cost controls do exist, though. The September 2026 release notes describe an automatic shut-off when a spending limit is reached, and October’s notes give a default monthly project cap of $100, with a $300 cap for projects launched from Canvas when no limit is set.
Important: A project cap keeps costs under control, but it can still stop a popular project mid-semester. If you build something other people will use heavily, ask your instructor or ASU support how limits apply before launch week.
Which models can students use?
Builder’s pages mention more than 50 models for faculty and staff, with a figure of 70+ on another page, so the number clearly changes. For students, ASU’s feature table marks model access as limited.
The three top-listed options are described as smaller, high-quality models that use fewer environmental resources. For most student projects, those smaller models are plenty.
| Model type | Good for | Trade-off |
|---|---|---|
| Smaller, efficient models | FAQ bots, study helpers, summaries, structured Q&A | May struggle with very long or subtle reasoning tasks |
| Larger general models | Complex analysis, long documents, nuanced writing feedback | Slower and more resource-intensive |
| Campus-hosted models on AIR (for example Gemma 4) | Privacy-sensitive projects, higher-volume use | Selection set by ASU; availability may vary by account |
Recent release notes also mention an Auto Model Selection feature that recommends a model based on your project instructions. If you are unsure, that is a sensible place to start.
Getting Started: Your First Project in CreateAI Builder
ASU’s quick-start instructions are short, and they are the same for faculty, staff and students. The detailed screens will evolve, so the steps below stay close to what ASU publishes and add practical advice around them.
The official quick start
- Go to CreateAI Builder. ASU lists the address as createai.asu.edu, and the platform launch link also appears on the ai.asu.edu site.
- Sign in with your ASURITE credentials. Complete Duo authentication if you are prompted.
- Create a new AI assistant or open an existing project from your dashboard.
From there, ASU’s guidance describes the same broad workflow: pick a template or start from scratch, upload your materials, configure how the assistant responds, test it, and share it with the people you choose.
Before you click anything
The best projects start on paper. Spend ten minutes answering five questions.
- Who is this for? Yourself, a study group, a club, a class?
- What one job should it do well? A narrow job beats a vague one.
- What sources should it rely on? Notes, a syllabus, a public policy page, a reading list.
- What should it refuse or hand off? Grades, personal data, medical advice, anything outside its scope.
- How will you know it works? Write down ten questions and the answers you expect.
Pro Tip: If you cannot describe your project’s job in one sentence, it is not ready to build. “Answers questions about the PSY 101 syllabus using only the syllabus” is a project. “Helps with school” is a wish.
Use a template or start from scratch?
ASU’s Builder page lists a set of ready-to-customize templates, with course helpers and student support bots as examples. Templates give you a working structure, which is especially useful the first time.
| Starting point | Choose it when | Watch out for |
|---|---|---|
| Template | Your idea matches a common pattern such as an FAQ bot or a course helper | Default wording may not fit your audience; edit it rather than accepting it |
| From scratch | You have a specific behavior or workflow in mind | More decisions up front; easy to write instructions that are too vague |
| Copy of an existing shared project | Someone already built something close to what you need, and permissions allow it | You inherit their sources and assumptions, so review everything |
The release notes also mention a Quick Start FAQ Template added in June 2026 and an Idea Generation and Planning feature that uses design thinking to guide brainstorming, which can help when you are stuck at the blank page.
The Anatomy of a Custom Agent
Every assistant you build in CreateAI Builder is made of the same few parts. Understanding them tells you where to look when something goes wrong.
1. The instructions
Instructions are the standing orders your assistant follows in every conversation. ASU’s release notes mention universal system prompt support, which refers to this layer.
Good instructions define a role, a task, a tone, a format and limits. They also say what to do when the answer is not in the sources.
2. The knowledge base
The knowledge base is the material your assistant draws from. ASU lists course materials, policy documents, research papers and onboarding guides among accepted content, and the Builder page notes that the knowledge base can take uploaded documents, URLs and synced Google Drive content.
ASU’s documentation also covers indexing content from public ASU websites, including page depth and indexing levels, and publishes a current list of accepted file types updated in October 2026.
3. The model
The model is the engine that reads your instructions and sources and writes responses. Choosing one is a balance among quality, speed and resource use, which is why ASU built a comparison tool.
4. Optional abilities
Release notes from 2026 describe additions such as cross-conversation memory, agent web search, expanded Model Context Protocol support, OpenAI-compatible Builder API endpoints and Canvas integration. Which of these appear in a student account may differ, so treat them as possibilities to check rather than guarantees.
5. Sharing and permissions
Access to uploaded data is controlled at the project level, and only people you grant permission can view or use it. ASU’s roles guide distinguishes Owners from Editors, and the Builder page says you can invite collaborators as editors or viewers.
| Component | What it controls | What goes wrong when it is weak |
|---|---|---|
| Instructions | Role, tone, format, boundaries | Rambling answers, off-topic behavior, overconfidence |
| Knowledge base | What facts the assistant can draw on | Invented details, outdated answers, missing information |
| Model | Reasoning quality, speed, cost | Slow replies, shallow answers, unnecessary expense |
| Abilities and connections | Search, memory, external tools | Unpredictable behavior, privacy surprises |
| Permissions | Who can use or edit it | Unwanted access, accidental changes |
Writing Instructions That Actually Work
Instructions are where most beginner projects succeed or fail. A short, structured set of instructions beats a long, rambling one.
A simple five-part structure
- Role: Who the assistant is. “You are a study helper for a first-year statistics course.”
- Task: What it does. “Explain concepts using only the uploaded lecture notes and quiz-style questions.”
- Audience and tone: Who it talks to and how. “Friendly, plain language, no jargon without a definition.”
- Format: What answers look like. “Two short paragraphs, then a one-question check.”
- Boundaries: What it will not do. “If the notes do not answer the question, say so and suggest asking the instructor.”
Expert Tip: Tell the assistant what to do when it does not know. Without that sentence, many assistants fill the gap with a confident guess.
Example: weak versus strong instructions
| Weak | Strong |
|---|---|
| Help students with biology. | Act as a tutor for BIO 181 using only the uploaded lecture notes. Explain one concept at a time, then ask a short check question. If the notes do not cover a topic, say so. |
| Be nice and helpful. | Use a warm, encouraging tone. Keep answers under 150 words unless asked for more. |
| Do not make mistakes. | Quote the relevant passage from the notes before explaining it, so the student can verify the answer. |
Building a Reliable Knowledge Base
An assistant is only as trustworthy as the sources behind it. A tidy knowledge base is often the biggest quality improvement you can make.
Choose sources deliberately
- Prefer primary materials such as the actual syllabus, the actual policy page or the actual reading.
- Use current versions. An outdated syllabus produces confidently wrong answers.
- Remove duplicates and contradictory drafts.
- Break up huge files if topics are mixed, and give files clear names.
Mind the file types and the formatting
ASU keeps a current list of accepted file types, so check it before you spend an evening preparing documents. The October 2026 release notes describe better handling of multi-column PDFs and tables, and note that Word files are converted to PDF before reading, which can affect layout-heavy documents.
Clean headings and clear structure help any model find the right passage. A scanned image of a page is much harder to use than a text-based document.
Respect copyright and privacy
Only upload material you are allowed to use. ASU publishes guidance on student file uploads that touches storage, privacy, copyright and recommended practices for faculty-led bots, and it describes FERPA-regulated data as usable only under specific conditions.
Warning: Do not upload other students’ work, personal information, grades or private communications to a project. If you are unsure whether material counts as protected, ask your instructor or ASU’s AI support team before uploading.
What ASU Students and Faculty Have Already Built
The best way to understand the range of the platform is to look at real projects ASU has highlighted. Most were built by faculty or staff, but they show patterns students can borrow.
| Project | Who built or uses it | What it does |
|---|---|---|
| Syllabot | Faculty across many units, including training for more than 300 W. P. Carey faculty | Answers common course questions about due dates, assignments and grading policies. Nearly 1,000 were active in March 2026. |
| AI mentor chatbots | Mary Lou Fulton College (Steve Salik) | Lets students practice professional communication in an internship-style role. |
| Brand bot and sustainable shopping guide | Students in an AI in Fashion course at ASU FIDM (Naomi Ellis) | Student-built projects for fashion branding and shopping advice. |
| Language Buddy | ASU Enterprise Technology | Conversation practice in a foreign language; launched in German, now also in Spanish. |
| Medical training avatars | John Shufeldt School of Medicine and Medical Engineering | Simulated patient conversations, built with agent-style workflows and scored across 25 areas. |
| APA citation engine | Edson College of Nursing and Health Innovation | Helps with citation formatting. |
| Accessibility alt-text tool | Two computer science master’s students, Spark Challenge winners | Writes alt text for images in PowerPoint files and flags uncertain cases for human review. |
The agent-style example: Case Creator
The medical school’s Case Creator is the clearest example of agentic design in ASU’s materials. Multiple agents work together, each handling a part of building a patient case: medical background, current condition and prior clinical appointments.
The lesson for students is architectural. Complex tasks often work better when split into smaller jobs, each with clear instructions, than when handed to one giant, all-purpose assistant.
Coursework examples
In a College of Liberal Arts and Sciences biochemistry course, students wrote code to analyze large datasets and built models to identify protein sites linked to side effects of cancer treatments. In the fashion course, students were asked to build AI projects as part of their coursework.
If your instructor has not mentioned CreateAI, you can still ask whether a project could count toward a class assignment, or whether an existing course assistant could be improved.
Eight Project Blueprints You Can Adapt
These blueprints are starting points, not official ASU templates. Each pairs a narrow job with sources and a boundary.
1. The syllabus and logistics helper
Job: Answer questions about due dates, policies and grading for one course.
Sources: The syllabus and the course schedule.
Boundary: If the syllabus does not say, tell the student to ask the instructor.
2. The study partner
Job: Quiz you on concepts from your own notes and explain missed answers.
Sources: Your lecture notes and approved readings.
Boundary: Never invent facts that are not in the notes; flag uncertainty.
3. The interview and conversation coach
Job: Role-play a hiring manager, client or patient so you can practice a conversation, then give feedback against a rubric.
Sources: A job description or a scenario you write yourself.
Boundary: Stay in role until asked to stop, and give feedback only afterward.
4. The language practice buddy
Job: Hold simple conversations at your level and gently correct errors.
Sources: Your course vocabulary lists and grammar notes.
Boundary: Keep replies short and ask one follow-up question each turn.
5. The club or organization FAQ
Job: Answer common questions about meetings, dues, events and sign-ups.
Sources: Your club’s published documents and public pages.
Boundary: Do not share members’ personal information.
6. The citation and formatting checker
Job: Check reference formatting against a style guide you provide.
Sources: The style guide pages you are allowed to upload.
Boundary: Flag anything it cannot verify, and never fabricate sources.
7. The rubric reviewer
Job: Compare a draft against an assignment’s grading criteria and point out gaps.
Sources: The assignment rubric.
Boundary: Give feedback, not rewritten text, unless your instructor allows it.
8. The multi-step research organizer
Job: Take a research question, list sub-questions, sort your collected sources and draft a reading plan.
Sources: Your own reading list.
Boundary: Verify every claim and citation yourself before using it.
Common Mistake: Building a general-purpose “do everything” assistant. Narrow projects are easier to test, easier to trust and easier for others to adopt.
Designing Multi-Step, Agent-Style Workflows
Once a single assistant works well, you may want one that handles a sequence of steps. Even without special agent tools, you can design for this.
Break the job into stages
- Understand: Restate the request and ask a clarifying question if needed.
- Retrieve: Find the relevant passages in the knowledge base.
- Reason: Work through the problem step by step.
- Respond: Give the answer in a set format with sources.
- Check: Offer a quick self-check question or flag uncertainty.
Decide where a human stays in charge
The Spark Challenge winners built an accessibility tool that acted only when confident and flagged doubtful cases for a person. That pattern, automate the easy cases and escalate the hard ones, is worth copying whenever the stakes are real.
Consider splitting into several assistants
If one assistant is juggling unrelated jobs, create separate projects, each with narrow instructions and its own sources. The medical school’s design shows how specialized helpers can work together on one outcome.
Testing and Evaluating Your Agent
Many first-time builders test with two or three friendly questions and call it done. That is how embarrassing mistakes reach other people.
Build a test sheet
Write fifteen to twenty test questions in four groups.
- Easy questions whose answers are clearly in your sources.
- Tricky questions that need two sources combined.
- Out-of-scope questions the assistant should decline or hand off.
- Adversarial questions that try to make it break its own rules, such as “ignore your instructions.”
Score the results
| Test type | Pass looks like | Fail looks like |
|---|---|---|
| Easy | Correct answer that matches the source | Wrong or vague answer despite clear source |
| Tricky | Combines sources accurately, cites them | Mixes up facts or invents a link between them |
| Out-of-scope | Politely declines and points to the right person | Guesses confidently |
| Adversarial | Stays within its instructions | Drops its role or reveals things it should not |
Change one thing at a time
When a test fails, adjust a single instruction or source, then re-run the full sheet. Changing five things at once makes it impossible to know what helped.
Use analytics after launch
ASU documents a project analytics dashboard for usage and engagement reporting, and the August 2026 notes add cost breakdowns by web and API usage. Review what people actually ask. Repeated questions your assistant handles badly are your to-do list.
Pro Tip: Ask two or three classmates to try to break your assistant before you share it widely. Fresh users find problems you have stopped noticing.
Sharing, Collaboration and the Marketplace
ASU describes sharing as quick and permission-based. Published assistants can be shared across courses, teams or departments, and you decide who can view or edit.
Roles and permissions
ASU’s roles guide explains that Owners and Editors have different powers. As a rule, keep the number of Editors small, because every Editor can change behavior and sources.
The CreateAI Builder Marketplace
ASU introduced a marketplace for exploring, sharing and collaborating on projects across the university. Browsing it is a fast way to learn how others write instructions and organize sources.
Canvas integration
The Builder page lists Canvas LMS integration, and October’s release notes give projects launched from Canvas a different default spending cap. If your instructor wants an assistant inside a course, that is likely the route.
Working with a faculty sponsor
Earlier in 2026, students used CreateAI through faculty or staff sponsorship. Even now that students can build independently, a faculty partner helps when your project touches course materials, grades or student data.
Privacy, Data Rules and Responsible Use
Privacy is one of the main reasons to build inside CreateAI rather than on a public service. It is also where students most often make assumptions that do not hold.
What ASU says the platform does
ASU states that data in CreateAI stays within its governed environment, is not shared externally and is not used to train outside AI models. The platform is described as FERPA-compliant and built around ASU’s Design Principles for Beneficial and Responsible AI.
What the platform does not do for you
- It does not decide which data you are allowed to upload.
- It does not make an assistant’s answers correct.
- It does not replace your instructor’s rules about AI in coursework.
Public tools versus approved tools
ASU’s AI tools page draws a bright line. Public or personal AI accounts may only be used with publicly available ASU data, while internal, FERPA-protected, sensitive or highly sensitive information must stay in ASU-approved tools unless you have explicit authorization. That applies to students working with faculty on ASU data as well.
| Data type | Public AI account | ASU-approved tool such as CreateAI |
|---|---|---|
| Publicly available ASU information | Generally allowed | Allowed |
| Internal or sensitive ASU information | Not allowed without authorization | Only if the project and data are approved |
| FERPA-protected student records | Not allowed | Only under the specific conditions ASU describes |
| Other people’s private information | Avoid | Avoid unless clearly authorized |
Warning: “It’s a university tool” does not mean every upload is allowed. Read ASU’s guidance on FERPA data in Builder before you put anything involving student records into a project.
Academic integrity
Building an assistant is not the same as using one to complete an assignment. Your course policy decides what is allowed, and policies differ by class.
- Check the syllabus before using an assistant on graded work.
- Disclose AI help when your instructor expects it.
- Do not build a tool designed to produce work you will submit as your own unless your instructor permits it.
- Verify facts, calculations, quotes and citations yourself.
Bias and fairness
ASU’s FAQ says an Ethical AI Engine assesses bias, but no tool removes it entirely. If your assistant gives advice that affects people, such as feedback, screening or recommendations, test it with varied examples and keep a person responsible for decisions.
Common Problems and How to Fix Them
| Problem | Likely cause | What to try |
|---|---|---|
| The assistant invents answers | Missing source, or no instruction on what to do when unsure | Add the source and an explicit “say you don’t know” rule. |
| It ignores an uploaded document | Unsupported file type, scanned images, or messy layout | Check accepted file types and convert to clean, text-based documents. |
| Answers are too long or too vague | No format instruction | Specify length, structure and an example answer. |
| It drifts off topic | Instructions are broad or boundaries are missing | Narrow the job and add refusal rules. |
| Responses are slow | Large model or heavy demand | Try a smaller model or compare options with CreateAI Compare. |
| The project stopped working | Spending limit reached, or permissions changed | Check project limits and roles; ask ASU support. |
| You cannot find a feature you read about | Feature is in beta, limited or not enabled for students | Check the release notes and ask the AI Acceleration team. |
| A shared bot behaves differently than in testing | Different environment, such as beta versus production | Review ASU’s guidance on moving projects from beta to production. |
Mistakes That Waste Time
- Starting with the model instead of the problem. Pick the job first, then the model.
- Uploading everything. More documents are not better; relevant documents are.
- Skipping boundaries. An assistant without limits will try to answer anything.
- Testing only the happy path. Out-of-scope and adversarial questions reveal the real weaknesses.
- Sharing too early. Early users lose trust quickly when an assistant is wrong.
- Ignoring cost limits. A popular project can hit its cap.
- Never updating sources. A syllabus changes; your assistant should too.
Where to Get Help
ASU lists several support options on the Builder page. They include documentation and slide presentations, office hours, one-on-one consultations with the AI Acceleration team, a CreateAI Slack channel and a contact form.
There is also an assistant inside the platform that can help you navigate Builder. Start there for quick questions, then move to a consultation for design problems.
Expert Tip: Bring your test sheet and a short description of the problem to a consultation. Support staff can diagnose a concrete failing example far faster than a general complaint.
A Four-Week Plan for Your First Agent
Week one: define and draft
Write the one-sentence job, collect two or three clean sources and draft your five-part instructions. Create the project and run your first ten test questions.
Week two: test and tighten
Expand to twenty questions, including out-of-scope and adversarial ones. Change one thing at a time and record what improved.
Week three: share with a small group
Invite a few classmates as viewers. Collect their questions and failures, and check the analytics if available.
Week four: document and decide
Write a short note on what the project does, what sources it uses, what it cannot do and who maintains it. Decide whether to expand, hand it to an instructor or retire it cleanly.
What to Watch for Next
ASU has signaled where the platform is heading. Its 2027 plans include personalized memory that helps tools understand individual students, and connecting information across university systems such as classes, financial aid and IT support tickets.
The release notes also describe a limited beta of an MCP Gateway, a shared, governed connection to ASU systems for approved AI clients. Features like these could make student-built assistants more capable, though they will likely come with approval requirements.
Frequently Asked Questions
Can ASU students build their own AI agents in CreateAI Builder?
Yes. ASU’s release notes say that from August 2026 students can build their own AI projects, and the AI tools page lists Builder as available to students at no cost. Confirm your access with your ASURITE login, since some ASU pages still describe Builder as faculty and staff only.
Is CreateAI Builder free for students?
ASU says Enterprise Technology covers nearly all usage and that model prices shown in the platform are informational. There are spending limits, including a default monthly project cap of $100 noted in October 2026, so heavy projects can still hit a ceiling.
Do I need to know how to code?
No. Builder is a no-code tool. You write instructions, upload sources and test the result. Coding becomes relevant only if you move to APIs or other advanced options.
What is the difference between an assistant and an agent?
An assistant answers within its instructions and sources. An agent can also take multiple steps or use connected tools toward a goal. ASU’s public pages mostly describe assistants and projects, with agent-style features appearing in release notes.
How do I sign in?
Go to CreateAI Builder and sign in with your ASURITE credentials, completing Duo authentication if prompted. Student access begins about two weeks before classes start and ends when you are no longer enrolled.
Which AI models can students use?
ASU’s feature table marks student model access as limited compared with faculty and staff. Use CreateAI Compare or the model recommendations inside Builder to choose among the options you can see.
What files can I upload?
ASU publishes a current list of accepted file types, updated in October 2026. The knowledge base also takes URLs and synced Google Drive content. Clean, text-based documents work best.
Can I upload class materials?
Only if you are allowed to use them. Respect copyright, your instructor’s rules and privacy requirements, and never upload other students’ work or personal information.
Is my data used to train outside AI models?
ASU states that data stays within its governed environment and is not used to train external models. You are still responsible for following ASU’s data rules about what you upload.
Can I share my assistant with classmates?
Yes, based on the permissions you set. You can invite people as viewers or editors, and ASU’s Marketplace supports sharing across the university. Keep the Editor list short.
Can I use CreateAI for graded assignments?
Only if your course policy allows it. Platform access and academic integrity rules are separate, so check the syllabus and ask your instructor when unsure.
What happens when a project hits its spending limit?
ASU’s release notes describe an automatic shut-off when a limit is reached. Check your project’s limit, and ask ASU support or your instructor if a project is expected to see heavy use.
Can I connect my assistant to other tools?
ASU’s release notes mention Model Context Protocol support, API endpoints and a limited-beta MCP Gateway for approved clients. Whether those are open to student accounts is not clear from public pages, so ask the AI Acceleration team.
Where can I get help building my first project?
Start with the documentation and slide presentations linked from the Builder page, then use office hours, a one-on-one consultation, the CreateAI Slack channel or the contact form.
Conclusion: Start Small, Test Hard, Share Carefully
CreateAI Builder gives ASU students a real way to build custom AI tools without writing code, using a university login and ASU’s governed environment. Since August 2026, students have been able to build their own projects, not just use ones made by others.
The projects that work best share a few habits. They do one job well, rely on clean sources, tell the assistant what to do when it does not know and get tested against hard questions before anyone depends on them.
Limits and rules matter too. Spending caps, model access for students and data restrictions all shape what you can do, and some details are still evolving. Check ASU’s current pages and ask for help early.
Your next step is simple. Pick one small, real problem, write the one-sentence job, sign in and build the first version this week. A modest assistant that works beats an ambitious one that never launches.
