DEVELOPER RESOURCES

Build with
Sevren.

Uncensored LLMs for the things you want to build. Connect an agent, write your own client, or start a conversation in the terminal.

Make your first request Connect an existing agent

YOUR FIRST CONVERSATIONGet an API key ↗

PYTHON

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://api.example.com/v1",
    api_key=os.environ["SEVREN_API_KEY"],
)

reply = client.chat.completions.create(
    model="sevren-lite-27b",
    messages=[
        {"role": "user", "content": "Hello."}
    ],
)
print(reply.choices[0].message.content)

Examples use api.example.com. Replace it with your deployed API host. Use the exact model ID from the catalog.

ONE MODEL COLLECTION

Two familiar APIs.

Text, streaming and function tools. Use the client your workflow already speaks.

OpenAI compatiblePython, JavaScript & Chat Completions clientsPOST /v1/chat/completions

Anthropic compatibleClaude Code & Anthropic SDKsPOST /v1/messages

Find your next step.

GUIDES & REFERENCE

Start here

API reference

Account & access

CONNECTION DETAILS

Point your client here.

OpenAI clients take the base URL with /v1. Anthropic clients use the host without it.

OPENAI BASE URL

https://api.example.com/v1

ANTHROPIC HOST

https://api.example.com

API coverage & hosts

The official OpenAI Python and JavaScript packages work with a custom base URL. Claude Code and Anthropic SDKs use the Messages API.

Model endpoints
Endpoint	Credentials	What it does
`GET /v1/catalog`	None	The public model catalog with prices, limits and tags.
`GET /v1/models`	Optional	OpenAI-style list of the models your credentials can use.
`POST /v1/chat/completions`	API key for paid models	Chat completions with tool calling, streamed or as one JSON response.
`POST /v1/messages`	API key for paid models	Anthropic-compatible messages for Claude Code and the Anthropic SDKs.
`POST /v1/messages/count_tokens`	API key for paid models	The prompt size of an Anthropic-style request. Never charged.

Images and other non-text content, structured output, log probabilities, more than one choice per request, legacy functions fields, embeddings and the legacy completions endpoint are unavailable. Unsupported request fields return 400; missing endpoints return 404. See Chat completions for details.

Sevren hosts
Host	Serves
`https://api.example.com`	The model API under `/v1` and account API under `/account`.
`https://auth.example.com`	Wallet sign-in and holder plans under `/auth`.
`https://app.example.com`	The terminal, model browser and account page.
`https://docs.example.com`	This documentation.

These are example hosts. The API accepts browser requests only from configured Sevren origins. Call it from your server, script or command-line tool, not a page on another origin.