Plug in.
Start building.

Authentication

Send your EliteAi key using the Authorization header. Keep keys private. Requests require an active plan and available tokens.

Authorization: Bearer YOUR_ELITEAI_KEY

Chat completions

POST /api/public/v1/chat/completions
Content-Type: application/json

{
  "model": "YOUR_CONFIGURED_MODEL_ID",
  "stream": true,
  "max_tokens": 1024,
  "messages": [{"role": "user", "content": "Hello"}]
}

The portable deployment also serves /api/v1/chat/completions. On hosting with site-level access protection, use /api/public/v1 as your base URL. Streaming returns OpenAI-compatible server-sent events.

Claude · Anthropic Messages

POST /api/public/v1/messages
Authorization: Bearer YOUR_ELITEAI_KEY
Content-Type: application/json

{"model":"anthropic/claude-haiku-4-5","max_tokens":1024,"stream":true,"messages":[{"role":"user","content":"Hello"}]}

Text, inline images, local tool calls/results and native thinking blocks are supported. Use the exact Claude model ID. Unsupported optional features are rejected rather than silently dropped.

Model catalog

Limits & errors

Each model consumes plan tokens at its administrator-configured rate. Input allowance is conservatively measured in UTF-8 bytes before generation; unused reserved allowance is refunded after final usage is reported. Inline PNG/JPEG/WebP images are supported for image-enabled models; Function tools and tool-result messages are supported when the upstream model supports them. Claude uses POST /api/public/v1/messages with native Anthropic requests and streams. Cached input tokens are included in usage; admin-configured model rates apply. Remote image URLs and OpenAI Responses are not supported.