Drop-in for Claude Code & Codex
Use the keys you generate here as drop-in replacements for Anthropic or OpenAI keys. Streaming, prompt caching, and the Responses API all flow through unchanged.
One managed gateway for Claude Code, Codex, and SDK calls — with per-user budgets, audit trails, and team policies. Sign in with your work account and copy your key in under a minute.
What you get
Everyone at Bounteous gets one identity-bound key. The gateway handles provider routing, budgets, and audit. You handle the code.
Use the keys you generate here as drop-in replacements for Anthropic or OpenAI keys. Streaming, prompt caching, and the Responses API all flow through unchanged.
Every key is tied to an identity. Spend, request rate, and model access are enforced server-side — no surprise bills, no shared keys floating around in dotfiles.
Every request is logged with user, team, model, and token counts — shipped to CloudWatch and (optionally) Langfuse for trace-level inspection.
Use the smart-default alias and the gateway picks the
best model for the job. Anthropic outage? It transparently falls
over to OpenAI without you changing a line of code.
Regenerating from this portal blocks the old key first, then issues a new one — no race conditions, no leaked keys lingering after a mistake.
Claude (Opus, Sonnet, Haiku), GPT-5, GPT-4o, Llama, Mixtral, Qwen,
DeepSeek — all accessible by name. Add new models in
litellm/config.yaml.
In three steps
No tickets to file, no keys to email around. The portal does the rest.
One click — your Bounteous Entra ID account authenticates you. We never see your password, and your session is HTTPS-only with strict SameSite cookies.
The portal shows the key once and gives you ready-to-paste shell snippets for Claude Code, Codex, Python Anthropic SDK, and Python OpenAI SDK.
Your tools point at the gateway instead of the provider. Streaming, tool use, structured output — everything works. The gateway handles policy, budgets, and tracing.