Claude Code with ShareQuota
Claude Code speaks the Anthropic dialect, which is the proxy’s native one. Two environment entries point it at localhost; everything else about the tool stays as it was.
Claude Code reads its API base URL and its token from the environment or from ~/.claude/settings.json. Set the base URL to the local endpoint and the token to your sq- key, and every request goes to the proxy, which signs in to the Claude plan you connected — or fails over to another provider that can serve the same model when that plan is out of quota.
Setup
- 1Start the proxy and connect a planRun the binary, open the dashboard, and complete the OAuth sign-in for your Claude account. An API key works too.
- 2Point Claude Code at localhostAdd the two env entries to ~/.claude/settings.json, or export them in your shell. The dashboard’s one-click switch writes exactly these.
- 3Pick a model and run itStart claude with a routed model id. Any claude- prefixed id lands on the Claude connection; a provider prefix sends it elsewhere.
What to set
{ "env": { "ANTHROPIC_BASE_URL": "http://localhost:20130", "ANTHROPIC_AUTH_TOKEN": "sq-your-internal-key" }}Worth knowing
- Claude Code sends the Anthropic dialect, so streaming, tool use and thinking all pass through untranslated.
- If the Claude plan is exhausted, failover moves the request to another connection that serves the same model — the tool sees one endpoint throughout.
To undo it
Remove the entries above, or turn the integration off in the dashboard and the backup is restored. Nothing else was written.
Questions
Yes — that is the point. The proxy signs in to the plan you already pay for and Claude Code spends that quota, instead of a second bill for API tokens.
Yes. Send chatgpt/gpt-5.4 or gemini/gemini-3.1-pro-preview as the model and the proxy translates the Anthropic dialect to the provider’s own, including the stream coming back.
Checked against the tool’s own documentation on 2026-09-13. Settings move; if a field has been renamed, the docs link above is the source.