Skip to main content
Work can use Glasser to research paid data and build reports, spreadsheets, and other results. Choose setup based on where your task runs:
  • Work locally in the desktop app: follow the Skill setup below.
  • Work in the cloud: follow the Work MCP guide. This also applies when you use the desktop app for a cloud task.
For a regular chat, use the ChatGPT Chat guide. For Codex, use the Codex guide.
1

Start a local Work task

Open the ChatGPT desktop app, select ChatGPT, then switch to Work. Choose Work locally or the local option in the composer before you start the task.The agent needs terminal and internet access on your computer. If you do not have a local option, use the Work MCP guide.
2

Ask Work to install the Skill

Paste the same setup prompt as on the Glasser home page:
Work should save the Skill for future tasks and install or update the Glasser CLI. Approve file and install actions when the app asks. If Node.js 22 or newer is missing, have Work help you install it before continuing.
3

Sign in to Glasser

Work starts glasser login and shows a browser link and an eight-character code. Open the link, sign in or create an account, check that the code matches, and approve access to your Glasser Workspace. Create a Workspace if prompted.Return to Work and let it finish the free balance check. The CLI stores the Key on your computer; you do not need to paste it into the task. An existing valid login can skip this step.
4

Check the Skill in a new local task

Open Skills in the app sidebar and check that Glasser is listed. Start a new local Work task, type @, select the Glasser Skill, and ask:
A balance response confirms that the Skill and sign-in work, even if the balance is zero. If Glasser is missing, return to the setup task and ask Work to check that it saved and enabled the full Skill for future tasks, then start a new local task.
A local installation and sign-in do not automatically become available in a cloud task. If you switch to the cloud, connect the Glasser MCP plugin.

Try your first request

After you approve, Work can retrieve the data and use it in your report. It should report the result and charge. See OpenAI’s Work guide and Skill guide for current app options.