Summary
The coding agent's provider configuration is described three different ways across the installers, the management CLIs, and the config template. The provider name, the connection URL, and the credential all disagree, so at most one path can actually connect the agent to the proxy — and it is unlikely the Linux command-line path has been verified end to end.
Motivation
When a user runs the tool, the agent needs to know where to send requests. Today:
- The installers register a provider under one name; the CLIs export settings under a different name; the config file names a third.
- The installers point the agent at one connection path; the CLIs point it at a slightly different path.
- One platform's CLI sends one placeholder credential; the other platform's CLI sends a different one.
These inconsistencies mean the "start" command and the "install" step may configure the agent differently, and a user could end up with a setup that silently fails to route through the subscription proxy. The naming and URL should be unified to a single source of truth so that install and start always agree.
Acceptance
🤖 Generated with Claude Code
Summary
The coding agent's provider configuration is described three different ways across the installers, the management CLIs, and the config template. The provider name, the connection URL, and the credential all disagree, so at most one path can actually connect the agent to the proxy — and it is unlikely the Linux command-line path has been verified end to end.
Motivation
When a user runs the tool, the agent needs to know where to send requests. Today:
These inconsistencies mean the "start" command and the "install" step may configure the agent differently, and a user could end up with a setup that silently fails to route through the subscription proxy. The naming and URL should be unified to a single source of truth so that install and start always agree.
Acceptance
🤖 Generated with Claude Code