[ connect — sdks & agents ]
runaii.cloud speaks both the OpenAI and Anthropic dialects, so every major coding agent works with one environment variable — no proxy daemon, no config DSL. Create a key, set two env vars, ship.
Point Claude Code at runaii (env vars — set before launching):
$ export ANTHROPIC_BASE_URL=https://api.runaii.cloud/apiexport ANTHROPIC_AUTH_TOKEN=sk-runaii-… # your API keyThen launch normally — requests are metered to your runaii balance:
$ claudeModel routing: use runaii/glm-5.3 or runaii/kimi-k3 with /model inside Claude Code.
Settings → Models → OpenAI API key → override base URL:
$ # Cursor settings JSON (Models → OpenAI API Key → override){ "apiKey": "sk-runaii-…", "baseUrl": "https://api.runaii.cloud/v1"}Enable the runaii models you want (they appear after Verify):
Add model IDs exactly as runaii/glm-5.3-flash, runaii/kimi-k3, runaii/deepseek-v4-pro.
Cline settings → API Provider → OpenAI Compatible:
$ Base URL: https://api.runaii.cloud/v1API Key: sk-runaii-…Model ID: runaii/glm-5.3-flash~/.codex/config.toml — point the OpenAI provider at runaii:
$ [model_providers.runaii]name = "runaii"base_url = "https://api.runaii.cloud/v1"env_key = "RUNAII_API_KEY" # then: profile.default_provider = "runaii"One env var pair — aider reads OpenAI-compatible endpoints natively:
$ export OPENAI_API_BASE=https://api.runaii.cloud/v1export OPENAI_API_KEY=sk-runaii-…aider --model openai/runaii/glm-5.3The two env vars every OpenAI-compatible tool understands:
$ export OPENAI_API_BASE=https://api.runaii.cloud/v1export OPENAI_API_KEY=sk-runaii-…Anthropic tools: ANTHROPIC_BASE_URL=https://api.runaii.cloud/api + ANTHROPIC_AUTH_TOKEN.
One curl from the same shell — if this returns a completion, your agent will too:
$ curl https://api.runaii.cloud/v1/chat/completions \ -H "Authorization: Bearer $RUNAII_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"runaii/glm-5.3-flash","messages":[{"role":"user","content":"ping"}]}'Using a tool that's not listed? Anything that speaks OpenAI or Anthropic works — check the docs for the base URLs, or hit FAQ for limits.