OSOpenSoftwareAPI

Live on OpenSoftware infrastructure

Open models.
One private API.

Run OpenAI-compatible inference against models hosted by OpenSoftware on GCP, with confidential-compute routing when privacy has to be enforced.

Login with OpenSoftware Read the docs

01

OpenAI compatible

Point the OpenAI SDK at one base URL. Chat completions, streaming, models, tool calls, and standard error shapes work as expected.

02

Keys you control

Create scoped credentials, set an expiration date and hard spend ceiling, inspect usage, and revoke access instantly.

03

Confidential routing

Require a TEE backend per request. If confidential compute is unavailable, the gateway fails closed instead of silently downgrading.

A familiar interface

Change the base URL.
Keep your code.

from openai import OpenAI

client = OpenAI(
    api_key="sk-os-…",
    base_url="https://api.opensoftware.co/v1",
)

response = client.chat.completions.create(
    model="qwen2.5-0.5b-instruct",
    messages=[{"role": "user", "content": "Hello"}],
)