@kitana-sdk/server

An OpenAI-compatible HTTP server backed by your Claude subscription, from Kitana SDK. No API keys — it calls claude -p under the hood via @kitana-sdk/core.

Install

npm install @kitana-sdk/server

Run

npx @kitana-sdk/server
# or, if installed as a dependency:
PORT=4141 kitana-server

On first run it auto-installs the Claude CLI if missing and prompts you to sign in (claude auth login).

Endpoints

POST /v1/chat/completions   OpenAI-compatible chat completions (supports "stream": true)
GET  /v1/models             hardcoded model list
GET  /health                server + Claude CLI status

Example

curl -X POST http://localhost:4141/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"say PONG"}]}'

Works as a drop-in provider for anything that speaks the OpenAI Chat Completions API — n8n, OpenClaw, Continue.dev, the Vercel AI SDK, etc. See integrations.md in the main repo for verified setup guides.

Requirements

• Node.js 18+
• Claude CLI subscription (Claude Max, Pro, Team, or Enterprise)

See the main repo for full documentation.