oktsec Python SDK

Python client for the oktsec security proxy — identity verification, policy enforcement, and content scanning for AI agent communication.

Install

pip install oktsec

Quick start

from oktsec import Client

with Client("http://localhost:8080", "my-agent") as c:
    resp = c.send_message("recipient", "hello from Python")
    print(resp.status, resp.message_id)

With Ed25519 signing

from oktsec import Client, load_keypair

kp = load_keypair("./keys", "my-agent")
with Client("http://localhost:8080", "my-agent", keypair=kp) as c:
    resp = c.send_message("recipient", "signed message")
    print(resp.verified_sender)  # True

Async support

import asyncio
from oktsec import AsyncClient

async def main():
    async with AsyncClient("http://localhost:8080", "my-agent") as c:
        resp = await c.send_message("recipient", "async hello")
        print(resp.status)

asyncio.run(main())

Error handling

from oktsec import Client, PolicyError

with Client("http://localhost:8080", "my-agent") as c:
    try:
        resp = c.send_message("target", "suspicious content")
    except PolicyError as e:
        print(f"Rejected: {e.response.policy_decision}")
        print(f"HTTP {e.status_code}")
        for rule in e.response.rules_triggered:
            print(f"  - {rule.rule_id}: {rule.name} ({rule.severity})")

Generate keypairs

from oktsec import generate_keypair

kp = generate_keypair("new-agent")
kp.save("./keys")

API

Client(base_url, agent_name, keypair=None, timeout=30.0)

• send_message(to, content, metadata=None) → MessageResponse
• health() → HealthResponse
• get_quarantine(quarantine_id) → dict

AsyncClient(base_url, agent_name, keypair=None, timeout=30.0)

Same methods, all async.

load_keypair(directory, name) → Keypair

Loads Ed25519 keys from PEM files. Compatible with oktsec keygen output.

generate_keypair(name) → Keypair

Generates a new Ed25519 keypair.

License

Apache 2.0 — same as oktsec.