Upsun AI Gateway

This gateway provides authenticated access to AI providers (Anthropic Claude, OpenAI, Google Vertex AI) for Upsun staff.

Quick Start (2 commands)

curl -fsS https://ai-gateway.plat.farm/install.sh | bash

ai-gateway setup

Manual Installation

If you prefer to run each step separately:

1. ai-gateway login — Authenticate via Okta SSO
2. ai-gateway service install — Install and start the proxy service
3. ai-gateway shell-init --add — Add eval "$(ai-gateway env)" to your shell profile

Usage

Once configured, AI tools will automatically use the gateway:

• Claude Code works automatically after shell configuration.
• OpenAI-compatible tools use the OPENAI_BASE_URL and OPENAI_API_KEY environment variables.

Commands

• ai-gateway setup — Complete guided setup (login, service, shell)
• ai-gateway login — Authenticate via SSO
• ai-gateway logout — Clear stored token
• ai-gateway env — Output environment variables
• ai-gateway shell-init [--add] — Detect and configure shell
• ai-gateway token — Output current access token
• ai-gateway service install|start|stop|status — Manage background proxy

Using Other AI Tools (BYOK)

For any OpenAI-compatible tool that supports "bring your own key", configure:

• Base URL: http://127.0.0.1:38089/v1
• API Key: proxy

The local proxy handles authentication automatically.

Available Models

Claude (Anthropic via Vertex AI)

• claude-opus-4-5
• claude-sonnet-4-5
• claude-haiku-4-5

OpenAI

• gpt-5.2, gpt-5.1, gpt-5
• gpt-5-mini, gpt-4.1, gpt-4.1-mini, gpt-4.1-nano
• gpt-5-codex, gpt-5.1-codex, gpt-5.1-codex-mini

Gemini (Google via Vertex AI)

• gemini-3-pro-preview, gemini-3-flash-preview
• gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite

Mistral

• mistral-large-latest, mistral-small-latest
• mistral-medium-3, mistral-small-2503, codestral-2

Other (via Vertex AI)

• deepseek-ai/deepseek-r1-0528-maas, deepseek-ai/deepseek-v3.1-maas
• meta/llama-4-maverick-17b-128e-instruct-maas, meta/llama-3.3-70b-instruct-maas
• qwen/qwen3-235b-a22b-instruct-2507-maas, qwen/qwen3-coder-480b-a35b-instruct-maas
• moonshotai/kimi-k2-thinking-maas, minimaxai/minimax-m2-maas

Alternative: Token Mode

If you prefer not to run a background service, use token mode:

eval "$(ai-gateway env --no-proxy)"

This configures tools to connect directly to the gateway. Tokens are refreshed on each invocation.