Skip to main content
Add enterprise-grade governance, compliance, and human oversight to any AI agent in under 5 minutes. Katyar works with LangChain, CrewAI, LlamaIndex, custom agents, or any LLM tool-calling workflow.

​
Step-by-Step Installation

​
1. Install the Katyar SDK

Install the official Python SDK: 
pip install katyar
This gives you policy enforcement, tracing, observability, and compliance activation.

​
2. Create an API Key from the Dashboard

  1. Log in to your Katyar dashboard at https://app.katyar.ai
  2. Go to API Keys in the sidebar menu
  3. Click Create Key
  4. Give the key a descriptive name (e.g. my-agent-prod or dev-test)
  5. Copy the full secret key (kt_live_...) — it is only shown once
Creating an API key in the Katyar dashboard

​
3. Add the API Key to Your Environment

Create or update a .env file in your project root: 
KATYAR_API_KEY=kt_live_your_actual_key_here
Important:
  • Never hardcode the key in source code
  • Add .env to .gitignore
  • Use environment variables in production (Docker, Kubernetes, Vercel, etc.)

​
4. Initialize Katyar in Your Code (One Line)

Add this at the top of your agent script: 
import katyar

# This single line activates everything
katyar.init()  # Automatically loads KATYAR_API_KEY from environment
What happens automatically:
  • Agent registers itself with Katyar
  • 21 compliance controls are seeded
  • Real-time compliance evaluation runs
  • Policy gateway connection is established
  • OpenAI/Groq/Anthropic clients are auto-patched for tracing
  • Console shows your current compliance score
Example output: 
INFO: Katyar initialized – agent registered as "my-agent"
INFO: Compliance score: 12.4% (18 gaps) – visit dashboard to fix
Agent registered in Katyar dashboard

​
5. Secure Your Tools with the @katyar.tool Decorator

Wrap every function your agent calls: 
@katyar.tool("weather", "read")
def get_weather(city: str):
    # Your real implementation
    return {"temperature": 22, "condition": "sunny"}

@katyar.tool("database", "write")
def write_to_db(query: str):
    # This will be blocked/approved based on policy
    return execute_query(query)

@katyar.tool("payment", "transfer")
def transfer_funds(amount: float, to: str):
    # High-risk action — will likely trigger human approval
    return process_payment(amount, to)
Every decorated tool now has:
  • Policy checking before execution
  • Guardrail scanning (injection, PII, secrets, toxicity)
  • Tracing (latency, tokens, cost)
  • Audit logging
  • Blocking or HITL approval when required

​
Full Minimal Example

import katyar
from dotenv import load_dotenv

load_dotenv()

# Activate Katyar governance
katyar.init()

@katyar.tool("search", "web")
def web_search(query: str):
    return f"Search results for: {query}"

# Your agent logic
result = web_search("AI governance best practices")
print(result)

​