I would like to suggest adding support for Workspace Agents inside ChatGPT Projects.
Projects are already very useful for organizing ongoing work around a specific client, initiative, product, case, or internal process. They allow users to keep related chats, files, context, instructions, research, decisions, and working materials in one place.
However, Workspace Agents currently appear to live outside that project context. This creates friction when a user has already developed all the relevant work inside a Project and then wants an Agent to execute the next step of the workflow.
Many real workflows are not isolated conversations. They often follow a pattern like this:
-
A user creates a Project for a client, initiative, or workstream.
-
Inside that Project, they gather context, upload files, discuss requirements, explore solutions, generate diagrams, perform calculations, review alternatives, and refine the final direction.
-
Once the work is mature, they need a repeatable Agent to take action based on that accumulated context: generate a formal document, update a template, create a deliverable, send an email, organize files, produce a report, or complete another structured workflow.
Today, if the Agent cannot be invoked directly inside the Project, the user has to move context manually into a separate Agent conversation. That weakens the value of both Projects and Agents. The Project contains the knowledge, but the Agent is the tool designed to execute the workflow. Keeping them separate makes the Agent feel less like an integrated assistant and more like a separate tool that requires manual handoff.
A strong implementation could allow users to:
-
Attach one or more Workspace Agents to a Project.
-
Invoke a Workspace Agent from any chat inside the Project.
-
Let the Agent access selected Project files, instructions, and relevant conversation context.
-
Define what parts of the Project the Agent can use.
-
Allow project-specific Agent runs for repeatable workflows.
-
Optionally allow Agents to create or update files, generate deliverables, send emails, or perform other approved actions based on the Project context.
This would make Projects much more powerful as persistent workspaces and make Workspace Agents more useful as true workflow executors.
The main need is simple: users should be able to keep their work, context, and execution inside the same Project. When an Agent is needed to complete a task, it should be available where the work already lives.
This would be valuable for many use cases, including client work, consulting, sales proposals, legal or operational reviews, internal reporting, product documentation, onboarding processes, project management, technical analysis, and any workflow where users build context over time and then need a repeatable Agent to produce or execute a final output.
Note: By “Workspace Agents,” I am not referring to GPTs. I am referring to the newer Agents feature available in ChatGPT for Pro, Business, and Enterprise users, where agents can be created to perform repeatable workflows with tools, files, apps, and approved actions.