Best ChatGPT connectors and plugins for work in 2026

Most “best ChatGPT connectors” lists rank integrations as if one of them wins. That framing does not survive contact with a real team. The best ChatGPT connector is usually the one attached to the system your team already treats as the source of truth — which means the right answer for a Google Workspace agency is the wrong answer for a Microsoft 365 hospital network, and both are wrong for a software team whose real knowledge base is a monorepo and a Slack channel.

So this guide is organized by job to be done: company documents and knowledge, communication and coordination, and engineering and work management. Every capability statement below is drawn from OpenAI's current documentation, and where Connector Scout holds a documentation-verified record for a connection, the guide links to it instead of repeating the detail.

First, the vocabulary changed: plugins, apps and connectors

As of 9 July 2026, OpenAI migrated the app directory to the Plugin Directory, and plugins are now the primary way to discover workflow capabilities across ChatGPT and Codex. A plugin can package skills, apps and app templates. Apps remain the integrations that connect ChatGPT or Codex to external data and actions, which is what people usually mean when they say “connector.”

That last sentence matters more than the renaming. A plugin appearing in the directory is not evidence that your account can use it, and it is not evidence that write actions are enabled. Availability still resolves per workspace.

Best ChatGPT connectors at a glance

Best for company documents and knowledge

This is the category where connectors pay for themselves fastest, because the work is retrieval: find the current version of a document, reconcile two versions of a number, summarize a folder of research, answer a question with a citation instead of a guess.

Google Drive is the default choice for teams whose working knowledge already lives in Google Workspace. ChatGPT can search and reference Docs, Sheets and Slides the connected account can access, and OpenAI documents supported Drive actions — creating, updating, moving, sharing or deleting content — as dependent on the account's permissions and the workspace's approved scopes. Sync is where teams get confused: an individual connection is live access, while admin-managed Google Workspace sync is a separate, eligibility-limited setup.

SharePoint is the equivalent for Microsoft-heavy organizations, and it is the better first connection wherever documents are governed rather than merely stored. OpenAI documents search and reference across SharePoint and supported OneDrive for work or school content, plus actions such as creating folders, uploading and updating files, managing sharing links and working with supported lists and pages, depending on which actions are enabled. Admin-managed sync here is restricted to eligible Enterprise and Edu workspaces.

Box is the sensible alternative for organizations that standardized on Box rather than Drive or SharePoint: ChatGPT can use authorized Box files, action availability varies by account and workspace, and individual sync is not part of the current Box setup flow. Notion suits lighter team knowledge and project documentation — authorized pages are searchable, individual sync is no longer supported, and page or database writes depend on the specific app or plugin, the connecting account's permissions, workspace settings and approvals.

Best for communication and coordination

Documents tell you what was written down. Messages tell you what was decided. That is the real job of this category: recovering the reasoning that never made it into a document.

Slack is the strongest example. ChatGPT can search messages, threads and channels; keyword search is broadly available, while semantic search depends on supported Slack AI Search plans and settings. Where actions are enabled, documented ones include joining channels, setting reminders, uploading files and updating profiles. One practical constraint: a connected account works with one Slack workspace at a time.

Microsoft Teams covers the same job for organizations whose operating context lives in Teams chats, channels and meetings. OpenAI documents searching chats and channels and reviewing scheduled meeting artifacts, with supported live actions — creating a Planner task, for instance — dependent on app permissions and workspace settings.

Outlook splits into Email and Calendar apps. Email can search and reference messages and supports available contact-management actions; Calendar can read and search events and availability, and supported shared-calendar actions can include create, update, RSVP, cancel and delete when permissions and actions allow. Do not generalize that into “ChatGPT can manage my inbox” — available actions depend on account permissions and Microsoft-side configuration.

Gmail is worth enabling for Google Workspace teams that already connected Drive. OpenAI's Google data controls documentation confirms Gmail as a connected Google service and describes event-triggered Work tasks for new messages for eligible users, with access governed by the scopes an administrator approved.

Best for engineering and work management

GitHub is the clearest single-purpose win on this list. Connected, ChatGPT can retrieve code, README files and other documentation from authorized repositories on demand and cite the relevant snippets, with available experiences varying by ChatGPT surface and plan. Two limits are worth stating plainly, because older write-ups get them wrong: live access does not create a synced GitHub index, and individually configured repository indexing is not available.

For work management, OpenAI documents a ClickUp app with sync that can access spaces, folders, lists, tasks and subtasks, comments and checklists, tags, assignees and due dates. That is genuinely useful for status questions and planning context — but the documented sync connector is read-scoped, so treat “ChatGPT can update our tickets” as unproven until a specific app documents the action.

Connector Scout also holds documentation-verified ChatGPT records for other work systems where the evidence supports publication, including project tracking and CRM.

How to choose: five questions in order

1. Which system is the source of truth? Connect that one first, even if a different tool has a more impressive integration.
2. Do you need retrieval, or do you need actions? Retrieval is where most documented capability sits today; actions require separate verification.
3. Live fetch or synced index? Live access reflects the connected account right now; sync supports broader retrieval but usually adds eligibility and administrator setup.
4. What can the connected account already access? A connection inherits that account's permissions — it does not widen them, and it can quietly narrow what ChatGPT can find.
5. What does your workspace admin allow? Installation, app access, actions and provider permissions can all be restricted before a user ever sees the app.

Why read support and write support are different decisions

Reading is reversible. If ChatGPT retrieves the wrong document, you notice and move on. Writing is not: a created folder, an updated property, a changed sharing link or a cancelled meeting propagates into other people's work. That asymmetry is why Connector Scout records read, search, sync and action capabilities separately, and why a connection with excellent search support should not be assumed to have any write support at all.

The practical rule for rollout: enable retrieval broadly, enable actions narrowly, and require a named owner for any action-enabled connection.

OpenAI's admin documentation is the right reference for that governance layer: plugin installation, app access, actions, provider permissions and supported indexing are all managed centrally in Business, Enterprise and Edu workspaces.

What I would enable first, by organization type

• Google Workspace-heavy small team: Drive first, then Gmail and Calendar if available on your plan. One document source, one message source, no duplication.
• Microsoft 365 organization: SharePoint for governed documents, Teams for operating context, Outlook for inbox and calendar work — in that order, because document retrieval delivers value before action support is configured.
• Software team: GitHub for code, whichever of Slack or Teams holds decisions, then a project tracker for status context.
• Knowledge-heavy operations or consulting: your primary document store (Drive, SharePoint or Box), plus Slack or Teams, plus Notion if it genuinely holds current knowledge rather than abandoned pages.

In every case the sequence is the same: one source of truth, one communication surface, then a specialist system — not ten connections enabled in an afternoon.