Updated November 2026Building ChatGPT Plugins and Workspace Agents
Class Duration
7 hours of live training delivered over 1-2 days to accommodate your scheduling needs.
Student Prerequisites
- Comfort with everyday office software: documents, spreadsheets, email, and a web browser
- Basic experience chatting with ChatGPT is helpful but not required
- Experience building or using custom GPTs is helpful for the migration topics
- No technical background or programming experience required
Target Audience
Business users, team leads, and custom GPT owners who need to rebuild their GPTs as plugins and workspace agents, plus the workspace administrators who govern them. Attendees work on ChatGPT Business, Enterprise, or Edu. No technical background is assumed.
Description
OpenAI is retiring custom GPTs and replacing them with plugins built from skills and apps. This course teaches business users to write skills, combine skills and apps into plugins, and build workspace agents, with no programming. Agents can run on demand, on a schedule, in Slack, or through an API trigger. Participants also learn how app actions are approved and how admins govern plugins, skills, and agents. A migration module moves existing GPTs to plugins. Custom GPTs retire on December 11, 2026, or February 11, 2027 for Enterprise workspaces with an approved deferral.
Learning Outcomes
- Explain how plugins, skills, apps, and workspace agents replace custom GPTs.
- Write skills with clear instructions, descriptions, and reference files.
- Combine skills and connected apps into a plugin.
- Connect apps and custom MCP servers with safe permission levels.
- Build a workspace agent that runs on demand, on a schedule, or in Slack.
- Share skills, plugins, and agents and apply workspace admin controls.
- Migrate a custom GPT to a plugin.
- Plan around what a migration does not carry over.
- Test migrated plugins and roll them out to the GPT's users.
Training Materials
Comprehensive courseware is distributed online at the start of class. All students receive a downloadable MP4 recording of the training.
Software Requirements
Students need a ChatGPT Business, Enterprise, or Edu workspace with plugins, skills, and workspace agents enabled by an administrator, plus a web browser. Existing custom GPTs to migrate are useful but not required. No installation required.
Training Topics
From Custom GPTs to Plugins
- Why custom GPTs are retiring
- Plugins, skills, and apps
- Plugins versus workspace agents
- Plugins versus projects
- The plugin directory
- One directory for ChatGPT and Codex
- Where a plugin or agent helps
Designing Skills
- What a skill contains
- Creating a skill in chat
- Uploading a skill folder
- Writing a skill description
- Writing skill instructions
- Adding examples and reference files
- Automatic and explicit skill use
Building Plugins
- Creating a plugin with Plugin Creator
- Combining skills and apps
- Naming and describing a plugin
- Installing plugins from the directory
- One plugin or several
- How connected apps work
- Read and write actions
- Approving an action
- App permission levels
- Custom MCP servers
- When IT sets up an MCP server
- What an app can access
Workspace Agents
- Workspace agents versus plugins
- Building an agent in the agent builder
- Instructions, skills, apps, and files
- Choosing model and reasoning effort
- Running on demand or on a schedule
- Agents in Slack channels
- API triggers
- Approvals for write actions
Admin Controls, Sharing, and Governance
- Sharing with people and groups
- Publishing to the workspace
- Plugin installation policies
- Role permissions for plugins and skills
- Admin controls for workspace agents
- Private marketplaces synced from GitHub
- Plugin verification status
- Compliance API and audit logs
- Protecting company data
Migrating Custom GPTs
- Inventorying your GPTs
- Instructions to skills
- Knowledge files to reference files
- Actions to apps or MCP servers
- The Migrate to plugin option
- Who can migrate a GPT
- What does not carry over
- Read-only GPTs after migration
- Retirement and deferral dates
Testing and Rolling Out
- Testing with the GPT's usual prompts
- Testing harder cases
- Migrated plugins start private
- Sharing with the original audience
- Telling users about the change
- Refining skills from feedback
- Maintaining skills, plugins, and agents