nail-sdk 1.0.0

NAIL — AI Agent Liability SDK. Real-time monitoring, circuit-breaking, and insurance certification for AI agents.

nail-sdk

NAIL — AI Agent Liability SDK

Real-time monitoring, circuit-breaking, and insurance certification for AI agents. Drop-in integration with OpenAI, Anthropic, LangChain, CrewAI, and AutoGen.

Before You Start

You need two things from the NAIL Portal:

1. Register your agent — Dashboard → Agents → Add Agent. Note the agent name you choose.
2. Generate an API key — Dashboard → SDK & API Keys → Generate Key. Select the scopes you need (at minimum agent:read).

Store your key as an environment variable — never hard-code it:

export NAIL_API_KEY="nail_sk_YOUR_KEY_HERE"

Add that line to your .env file or deployment secrets (Render, Railway, Cloud Run, etc.).

---

Install

pip install nail-sdk                    # Core SDK
pip install nail-sdk[openai]            # + OpenAI wrapper
pip install nail-sdk[anthropic]         # + Anthropic wrapper
pip install nail-sdk[langchain]         # + LangChain callbacks
pip install nail-sdk[all]               # Everything

---

Quick Start

OpenAI

import os
from nail.sdk.openai_plugin import nail_wrap_openai
import openai

client = openai.OpenAI()
client, nail = nail_wrap_openai(
    client,
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-openai-agent",       # matches the name in your NAIL Portal
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)
print(nail.session_summary)

Anthropic

import os
from nail.sdk.anthropic_plugin import nail_wrap_anthropic
import anthropic

client = anthropic.Anthropic()
client, nail = nail_wrap_anthropic(
    client,
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-anthropic-agent",    # matches the name in your NAIL Portal
)

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}]
)

LangChain

import os
from nail.sdk.langchain_plugin import NAILGuardMiddleware

middleware = NAILGuardMiddleware(
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-langchain-agent",    # matches the name in your NAIL Portal
)
chain.invoke({"input": "..."}, config={"callbacks": [middleware]})

CrewAI

import os
from nail.sdk.crewai_plugin import nail_wrap_crew

crew, nail = nail_wrap_crew(
    my_crew,
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-crew-agent",         # matches the name in your NAIL Portal
)
result = crew.kickoff()

Any Python Agent (decorator)

import os
from nail.sdk import Neuravant

nail = Neuravant(
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-agent",              # matches the name in your NAIL Portal
)

@nail.monitor
async def handle_query(user_input: str) -> str:
    return await my_llm_call(user_input)

AutoGen

import os
from nail.sdk.autogen_plugin import NAILAutoGenMonitor

monitor = NAILAutoGenMonitor(
    api_key=os.getenv("NAIL_API_KEY"),
    agent_name="my-autogen-agent",      # matches the name in your NAIL Portal
)
for agent in [agent1, agent2]:
    monitor.register(agent)

---

No-Code / Non-Technical Setup (Proxy Mode)

If you use a no-code tool (Voiceflow, Make, Zapier, n8n, Dify, Flowise, etc.) and cannot modify Python code, use NAIL Proxy instead.

Step 1 — Get your proxy-enabled key

In the Portal → SDK & API Keys → Generate Key, tick the proxy:use scope.

Step 2 — Point your tool at the NAIL proxy

Instead of your LLM provider's base URL, use:

Provider	Original URL	NAIL Proxy URL
OpenAI	https://api.openai.com	https://neuravant.ai/proxy/openai/v1
Anthropic	https://api.anthropic.com	https://neuravant.ai/proxy/anthropic/v1
Mistral	https://api.mistral.ai	https://neuravant.ai/proxy/mistral/v1
Cohere	https://api.cohere.ai	https://neuravant.ai/proxy/cohere/v1

Step 3 — Add your NAIL key as a custom header

In your no-code tool's HTTP settings, add one custom request header:

X-NAIL-API-Key: nail_sk_YOUR_KEY_HERE

Your existing provider API key stays in the Authorization: Bearer <provider-key> header as normal. No other changes needed.

Test with curl:

curl https://neuravant.ai/proxy/openai/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "X-NAIL-API-Key: nail_sk_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o","messages":[{"role":"user","content":"ping"}]}'

All LLM traffic is transparently forwarded. NAIL monitoring (latency, tokens, pathology detection) runs in the background. No code changes required.

---

What You Get

• Circuit Breaker — auto-kills runaway agents on latency, cost, or error spikes
• Telemetry — latency, token usage, cost tracking with PII scrubbing
• Pathology Detection — sycophantic decay, consensus paralysis, shadow delegation
• Live Badge — embeddable SVG showing real-time NAIL safety rating
• Compliance — GDPR, SOC 2, EU AI Act scanning built in

Links

• Docs: neuravant.ai/sdk
• Portal: neuravant.ai/portal
• Support: [e-mail]