For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending .md to the page URL.
Primary navigation

Plugin controls

Understand plugin availability, connector-backed capabilities, actions, and connected-system access

A plugin extends ChatGPT and Codex by packaging skills and optional connectors so teams can distribute workflows and knowledge. The products share one universal plugin directory, while admins control availability and installation for their workspace. Learn more about plugins, skills, and apps and connectors.

When a plugin includes a connector, workspace admins must make the plugin available through plugin controls and configure connector access before members can use the connector-backed capability.

Plugins are available with ChatGPT Work on the web, and with ChatGPT Work and Codex in the ChatGPT desktop app, and through the Codex CLI plugin browser. Availability on those surfaces doesn’t make plugins available in Chat, the IDE extension, or mobile.

For the complete administration model, see Roles and workspace permissions.

Understand the capability chain

Each layer has a separate scope and control surface:

LayerWhat it determinesWhere to manage it
Plugin availability and installationWhether the plugin bundle is available to the userWorkspace settings for supported web and desktop surfaces; the CLI plugin browser for CLI
Bundled skillsWhich reusable instructions the installed plugin contributesThe plugin package and Skill controls
Connector accessWhether users can use a connector-backed capabilityWorkspace apps and Permissions & roles
Connector actions and permissionsWhich actions users can run and when ChatGPT asks before using the connectorThe connector’s Action control and App permissions in Workspace apps
Source-system authorizationWhich external data and actions the authenticated identity can accessThe connected service and its identity provider
Runtime permissionsWhat an agent can do after it receives data or a toolThe runtime, sandbox, and approval controls for the active surface

Depending on the workflow, admins can govern plugin availability, connector access, connector actions and permissions, provider authorization, and runtime policy independently.

Plugin availability controls

Workspace plugin controls determine whether a plugin is available or installed for supported workspace roles. The Codex CLI plugin browser controls CLI installation through its own path. See Build plugins for packaging and distribution.

Connector-backed capability controls

Plugins in ChatGPT and Codex can include connectors that search, retrieve, sync, or act on external systems. Workspace admins configure plugin availability separately from the access and actions granted to each connector.

Manage connector-backed capabilities from Workspace apps and Permissions & roles. Available controls let admins:

  • Enable reviewed connectors and assign access by workspace role.
  • For connectors that support Action control, allow read-only actions or an approved custom set, including how the workspace handles newly added actions.
  • Set App permissions that determine when ChatGPT asks before using a connector.
  • Keep access within the scopes and permissions granted by each connected service and authenticated user.

For current availability and procedures, see Admin controls, security, and compliance in apps.

Choose a starting set of plugins

For a broad initial rollout, consider plugin categories teams use every day: email, calendar, and file or document systems such as Google Drive or Notion. Use the Plugins Directory to confirm current availability and capabilities across supported ChatGPT and Codex surfaces.

Start with read actions. Enable write actions only after reviewing the plugin’s owner, each connector’s requested scopes, data access, external effects, and recovery path.

Understand data flow and security

When ChatGPT uses a connector-backed plugin, the connector sends a request to the connected service and returns data or action results allowed by the authenticated user’s provider permissions. Custom MCP servers expose these operations as tools through Model Context Protocol (MCP).

For non-synced connector use, ChatGPT processes data from Chat and deep research transiently and doesn’t index it. Connectors with sync index selected connected content in advance. This indexing distinction doesn’t replace normal chat-retention controls; chats that use plugins remain available through the Compliance API.

OpenAI’s current connector guidance also documents encryption in transit and at rest, per-user authorization, role and action controls, restricted network access for chats that use plugins, and no model training on information accessed through plugins for Business, Enterprise, and Edu customers. Review the connected service’s scopes, retention, and data-residency policies because those policies apply when a request reaches that service.

See app security and compliance and apps with sync for the current data-handling details. For locally configured MCP servers in the ChatGPT desktop app, Codex CLI, or IDE extension, see Codex MCP configuration.

Use current procedures