AI Gateway brings multiple AI models into one University of Iowa-managed workspace, powered by Cloudforce NebulaONE. Use it to explore model responses, work with approved documents and create reusable AI agents for research and university work.

Key features of AI Gateway

  • ONEchat: Ask questions, draft content and refine responses through follow-up conversations.
  • Model choice: Select from the models enabled for your account and compare how they approach the same task.
  • Files and reference materials: Work with approved documents using the features available for your selected model.
  • Custom agents: Combine instructions and approved reference materials for repeatable tasks.

Features can vary by model and account configuration. A model available in Gateway does not include every feature of that provider's standalone applications.

Get started

  1. Complete the AI Gateway Program application.
  2. Wait until your request is approved and your access is provisioned. Application approval and access completion are separate steps.
  3. Only after your access is ready, open AI Gateway and sign in with your University of Iowa account when prompted.
  4. Open ONEchat and start a new conversation.
  5. Choose an available model, describe your task and provide relevant context.
  6. Review the response for accuracy. Continue with follow-up questions or ask the model to revise the result.

Who can apply for P3-funded access?

The approved P3 eligibility categories are:

  • Faculty: tenured, tenure-track, clinical-track, instructional-track and visiting.
  • Merit and Professional and Scientific staff with appointments of 50% or greater.
  • Postdoctoral scholars.
  • Graduate students sponsored by a faculty advisor.

University-funded access is provided through the University of Iowa’s FY2026 P3 program, with funding received for the project “AI Discovery Initiative: Expanding Access, Capability, and Innovation." P3 applicants certify that the listed recipients meet these criteria. 

A submitter may certify eligibility for a group request. Include the faculty advisor for sponsored graduate recipients. Appointment-document uploads are not required for the initial self-certification.

The P3 eligibility certification applies to the P3 funding route. Department-funded requests follow the established departmental purchasing and approval process.

How do I request access?

Complete the AI Gateway Program application. Identify each intended recipient, the proposed use and the funding route.

  • P3-funded requests: Describe the specific project, expected benefits, and eligible recipients. P3 supports one funded tool per recipient: AI Gateway or Claude for Education. You may use both if the other tool is funded separately.
  • Department-funded requests: Provide departmental funding information and obtain the applicable purchasing and financial approvals.

Submitting an application does not provide access. Wait for confirmation that your request is approved and your access has been provisioned before signing in.

For help with the application or access process, submit an AI Gateway Licensing Request.

How do funding and monthly allowances work?

Approved P3 funding or departmental funding may cover access. Funding terms and your individual monthly usage allowance are confirmed during request review.

A usage allowance is a spending limit, not a fixed number of tokens or a statement of actual charges. Unused monthly allowance does not carry forward.

Each recipient receives an individual account and allowance. Group applications do not create shared accounts or pooled budgets.

For funding or allowance questions, submit an AI Gateway Question.

What are tokens, and how do they relate to my allowance?

Tokens are units used to process information in an AI model. For text, they can represent words or parts of words. Your allowance is measured in dollars, not a fixed number of tokens.

The amount of work your allowance supports depends on the model, the information processed, and the tools used. Input and output can have different rates. A dollar allowance cannot be converted into one universal number of prompts, pages, or lines of code.

Which models are available, and how should I choose?

Models are grouped by provider and family, with higher version numbers first within each family.

OpenAI

GPT-6

  • GPT-6.1 Sol
  • GPT-6 Sol
  • GPT-6 Luna

GPT-5

  • GPT-5.6 Terra
  • GPT-5.6 Sol
  • GPT-5.6 Luna
  • GPT-5.5
  • GPT-5.4
  • GPT-5.4 Mini
  • GPT-5.3 Codex
  • GPT-5.2
  • GPT-5.1

GPT-4

  • GPT-4.1
  • GPT-4o

Anthropic

Claude Opus

  • Claude Opus 5.5
  • Claude Opus 5
  • Claude Opus 4.8
  • Claude Opus 4.7
  • Claude Opus 4.6
  • Claude Opus 4.5

Claude Sonnet

  • Claude Sonnet 5
  • Claude Sonnet 4.6

Claude Haiku

  • Claude Haiku 4.5

Choosing a model

The model selector in ONEchat shows the models currently enabled for your account. Use that list to check current availability, since ITS may enable or retire models over time.

  1. Identify your task, such as summarizing a document, drafting text, or exploring a research question.
  2. Try a short, representative prompt with an available model.
  3. Review the result for accuracy, completeness, and usefulness. If needed, try the same prompt with another model and compare the responses.
  4. Consider the model's usage cost before expanding to longer documents or repeated tasks.

Models can differ in capabilities, response speed, and cost. A model that works well for one task may not be the best fit for every task.

Access to a provider's model in AI Gateway does not include a separate subscription to that provider's applications or every feature of its standalone products.

How do I get useful responses?

Describe the task, audience, relevant context, and desired result. Include requirements such as length or format. Ask follow-up questions, request explanations, or ask the model to revise its response. For complex work, divide the task into smaller steps and review the results as you go.

For example:

  • Research: "Summarize the methods in the attached paper in five bullets. Identify limitations and flag details the paper does not state."
  • Planning: "Turn these approved project notes into an action list. Separate confirmed decisions from suggestions and identify missing owners."
  • Document review: "Compare these two documents against the criteria below. Cite the supporting passages and explain any uncertainty."

Conversation context can help with follow-up prompts, but do not assume the model remembers information from other conversations.

Vendor reference: Cloudforce ONEchat guide. Vendor knowledge-base access may be required.

How do I work with files and custom agents?

When file features are available for your selected model, add approved reference materials, and explain how you want the model to use them. Ask the model to identify its supporting sources and flag information missing from the files. Review its answer against the originals.

A custom agent combines reusable instructions and approved reference materials for repeatable tasks, such as answering questions about a procedure, or preparing a consistent first draft. Define its purpose, expected output and limits, then test it with representative questions before sharing.

Check permissions before sharing an agent or its source materials. File and agent capabilities can vary by model and account configuration.

Vendor references:

Vendor knowledge-base access may be required. Examples describe the product generally; use the features enabled in the Iowa workspace.

Can I connect other services?

Connections are available subject to the permissions and university requirements for the connected service. The M365 connector is not enabled for this release. Canvas/ICON integration remains under review and should not be treated as an available university integration.

For API use, submit an AI Gateway Question to confirm the appropriate configuration and cost controls before relying on it for a project.

What information can I use?

Review the University of Iowa Data Classification Guide to IT Services before using university information in AI Gateway.

  • Public and University/Internal data: Permitted, subject to applicable university, research, sponsor and data-use requirements.
  • Restricted data: Complete the required consultation with the Information Security and Policy Office before use. Obtain any required supervisor, departmental or data-owner approvals. Supervisor approval does not replace the required consultation.
  • Critical data: Contact the Information Security and Policy Office before use to determine whether the specific data and proposed use are permitted and what controls or approvals are required. Do not proceed until the review is complete and applicable requirements have been met. Consultation does not override service-specific prohibitions, including the PHI restriction below.
  • Protected health information (PHI): As of November 2026, PHI and other information requiring HIPAA compliance must not be entered into or processed using the AI Gateway service provided through this program. This restriction remains in effect unless the university explicitly approves a designated service or environment for that use. Check this page for updates.

These requirements apply to prompts, uploaded files, agent reference materials and information accessed through integrations. Service-access approval does not, by itself, authorize the use of Restricted or Critical data. Follow applicable research, sponsor and data-use restrictions and permissions for connected services.

For help, submit an AI Gateway Question. Describe the data category and proposed use; do not include sensitive records in the initial message.

Using P3-funded Claude for Education? Follow the separate Claude for Education data-use guidance. PHI must not be entered into that service either.

Review AI-generated material for accuracy, bias and appropriateness before using or sharing it. AI output does not replace your responsibility for research integrity or professional judgment.

What should I do if I reach or expect to exceed my allowance?

To discuss additional allowance, submit an AI Gateway Question. Include your project, current funding route, and the reason for the request. Submitting a ticket does not automatically increase your limit.

Unused monthly allowance does not carry forward. Start with a focused task or a small sample before expanding to larger workloads.

What if my P3 project involves coding?

Coding-focused P3 requests are directed toward Claude. The application asks whether you will use AI to manage code or scripts, including writing, editing, and debugging. If you also need Gateway for a multi-model or API use case, submit an AI Gateway Question to discuss the request. Gateway access does not automatically include a standalone Claude license.

Can I request access for a group?

Yes. Identify each recipient individually. Each person receives an individual account and usage allowance. A group request does not create a shared login or pooled budget. Department-funded recipients on the same request must use the same funding account; use separate requests when funding accounts differ.

What feedback is required for P3 funding?

P3-funded AI Gateway participants agree to provide feedback at six months and one year about how they used the tool, what they accomplished and their next steps.

ITS handles monthly usage reporting. Participants do not need to submit monthly reports.

Where can I get help or training?

Submit a ticket through the ITS Self-Service Portal:

Include the time of the problem, the model or feature involved, the error message, and steps to reproduce it. Do not include passwords, API keys, or restricted source material in a routine support message.

For general AI skills and learning opportunities, explore AI at Iowa support and training and ITS generative AI training and resources. These resources cover several products, and examples may include features that are not available in AI Gateway.

For product documentation, see Cloudforce's Introduction to the Product, User Guide, and FAQ. These external knowledge-base articles may require access and may use NebulaONE, the vendor's name. Iowa's access, funding, and data-use requirements still apply.

Last updated
Article number
13811