interseptix 0.2.0b1

Zero-config IAM and policy enforcement for AI agents

Interseptix Python SDK

Zero-config IAM and policy enforcement for AI agents.

Two lines of code. Your agent runs unchanged. Interseptix silently intercepts every outbound HTTP call — enforcing policies, redacting PII, and logging everything with a cryptographic signature.

Install (beta)

pip install interseptix

Quickstart

import os
from interseptix import InterseptixClient

# Initialise once at the top of your agent entrypoint
sdk = InterseptixClient(api_key=os.environ["INTERSEPTIX_API_KEY"])

# Activate the agent — base scopes loaded automatically from your dashboard
sdk.set_agent(agent_id=os.environ["AGENT_ID"])

# Run your agent exactly as before — nothing else changes
agent = create_react_agent(llm, tools=[...])
agent.invoke({"messages": [("user", "Process refund for order 123")]})

# Behind the scenes:
#   • every HTTP call is intercepted before it executes
#   • calls outside the agent's scopes → blocked, PermissionError raised
#   • PII in payloads → redacted before logging
#   • high-value actions → held for human approval
#   • everything logged with a cryptographic signature

How it works

The SDK monkey-patches httpx and requests at import time. No changes to your agent tools or LLM calls are needed.

• Scopes — configured in the Interseptix dashboard, fetched automatically by set_agent()
• Policies — deny_if, rate_limit, require_approval_if, redact_pii rules evaluated on every call
• PCI DSS — card numbers and CVVs are hard-stripped at the SDK before any data leaves your machine
• HIPAA — payloads containing PHI field names are withheld from logging
• Audit ledger — every intercepted call is HMAC-signed and stored immutably

Requirements

• Python 3.9+
• httpx (installed automatically)

Links

• Dashboard: interseptix.com
• Docs: interseptix.com/#/docs
• Issues: github.com/getinterseptix/interseptix-python/issues