1
Open a local agent session
Choose a mode that can run commands. If your assistant only runs code in a hosted sandbox,
use MCP instead.
2
Ask the agent to set up Glasser
Paste this into the chat:Approve the required file and install actions. If the client has no persistent Skill
support, it can use the instructions in this conversation, but you will need to provide
the Skill URL again in future conversations.
3
Approve sign-in
The agent runs
glasser login and gives you a browser link and an eight-character code.
Open the link, sign in, check the code, choose your Workspace, and approve access.
Create an account and Workspace if you do not have them yet.The CLI saves the Key locally. You do not need to paste a Key into the conversation.4
Check the connection
Ask the agent to use the Glasser Skill to check your available balance without making
paid Runs. If it saved the Skill, repeat this check in a new session to confirm that the
client can find it. A zero balance is a successful connection too.