SableNetwork
Built for autonomous agents

Your agents' inference should be sovereign.

Sable is the inference layer for AI agents that handle private state — trading strategies, agent memory, internal source, user data. Drop-in OpenAI compatibility, cryptographic privacy guarantees.

# Before
from openai import OpenAI
client = OpenAI()

# After
from sable_network import Sable
client = Sable(api_key=os.environ["SABLE_API_KEY"])

# Everything below this line is unchanged.
response = client.chat.completions.create(
    model="sable-llama-3.3-70b",
    messages=agent_memory.recent_messages(),
)

One import. Every existing call keeps working.

The agent problem

Why agents need a different inference layer.

Provider logs are an exfil channel.

Your agent writes prompts that contain trade signals, internal source, customer data. Every centralized provider's privacy policy says 'we may use this for safety review' — which is a polite way of saying 'a human might read it'.

Memory needs more protection than chat.

Conversational privacy assumes one prompt per session. Agent memory persists, references prior reasoning, and accumulates value. The threat model is different — and worse — than a chatbot.

Trust-me APIs don't compose.

When agent A calls agent B which calls provider C, the trust assumptions multiply. Cryptographic guarantees are the only thing that scales.

What you get

Cryptographic privacy, OpenAI ergonomics.

Stop trusting. Start verifying.

Move your agents off the centralized loop.

One line of code, and every prompt is sealed. Create an API key, swap your base URL, and ship.