Overview
Agentic AI tools do more than answer questions. When fully utilized, they can inspect and modify files, write and test software, explore connected information, use external tools and carry out multi-step work. Florida State University’s Information Technology Services helps FSU faculty, staff, and research teams obtain and safely configure approved agentic AI services for their work.
FSU can facilitate access to Agentic AI tools from Anthropic and OpenAI. Anthropic account bundles include Claude Code and Claude Cowork. OpenAI account bundles include Codex and ChatGPT Work. To find the right account or service that fits your work, data, integrations, and budget, start with an AI Research Consultation. Once you have selected a service, ITS will help arrange institutional access, billing, and configuration.
Highlights
Multi-Step Work
give an agent a goal and relevant context, and it will plan the work, use approved tools, complete a sequence of steps and return a finished result for your review
Build and Maintain Software
use coding agents to understand codebases, develop features, debug problems, write tests, document systems and assist with repeatable engineering work
Create Deliverables
gather and synthesize information, analyze data and organize files to produce drafts of reports, documents, spreadsheets, presentations and other review-ready materials
Automate Repeatable Workflows
build scheduled or reusable agent workflows for tasks like project updates, recurring analysis, code review, document preparation, monitoring and administrative follow-up
Cost
Paid
Cost Details
Available To
Faculty
Departments
IT Pros
Getting Started
To purchase Anthropic (Claude) tools or OpenAI (ChatGPT) tools, request this service and ITS will facilitate the purchasing process for these tools.
If you need help detetermining which tool will best suit your needs request an AI Research Consultation for free guidance from the ITS AI Research team.
Additional Information
Before You Use an Agent:
Agentic tools can read project context and may be able to edit files, execute commands, install software, browse websites, send messages or act through connected applications. The user remains responsible for the agent’s actions and outputs.
- Classify the data. Identify the highest-risk information the agent could access through prompts, files, repositories, logs, connected applications or generated output. See FSU’s Data Classification Standards for guidance.
- Confirm the access path. Use the FSU-provided or ITS-approved account, license, cloud project and billing method for the work.
- Limit permissions. Connect only the files, repositories, applications, tools and accounts required for the task. Do not give an agent broad access out of convenience.
- Protect credentials. Keep passwords, API keys, private keys, tokens and other confidential information out of prompts and repositories.
- Preserve recoverability. Use version control, backups, test environments and change review to detect and reverse incorrect actions.
- Review consequential actions. Inspect generated code, commands, communications, purchases, account changes, research conclusions and other important outputs before accepting or sending them.
WARNING: FSU-supported agentic accounts are only enabled for up to Medium-risk data only. Do not use an agentic AI account with PHI, FERPA records, CUI, export-controlled information, identifiable human-subject data, restricted sponsor data, security-sensitive information or other High-risk data sets.
See Data Classification Standards for the full data classification guide. When in doubt, stop and request a review before providing the agent access.