metacog 1.0.0

Metacog Python SDK — Governed vertical AI for regulated industries

Metacog Python SDK

The world's first governed vertical AI SDK for regulated industries in India.

What makes Metacog different

Every other AI API gives you an answer. Metacog gives you a governed answer.

Every response includes:

• Confidence score — how certain is the AI? (M1)
• Source citations — which act and section? (V3)
• PII protection — no PAN, Aadhaar, or salary data leaked (M7)
• Hallucination risk — low / medium / high (M6)
• Drift status — are answers consistent over time? (M3)

Measured results:

• 0% hallucination rate across 37 Indian compliance questions
• 60% governance score vs 7% industry average
• +53 percentage points advantage over nearest competitor

Installation

pip install metacog

Quick start

import metacog

client = metacog.Metacog(api_key="mk-...")

response = client.governed.create(
    model=metacog.METACOG_M3,
    prompt="What is the TDS rate for salary of 12 lakhs under new regime?",
)

print(response.answer)
# TDS for ₹12 lakh salary under new tax regime for FY 2024-25...

print(response.confidence)
# 0.94

print(response.sources)
# ['Income Tax Act Section 192', 'Finance Act 2024']

print(response.governance_passed)
# True

print(response.hallucination_risk)
# 'low'

Governance parameters

response = client.governed.create(
    model=metacog.METACOG_M3,
    prompt="What is the PF contribution rate?",
    governance=metacog.GovernanceConfig(
        vertical="payroll",
        confidence_threshold=0.85,   # Block if confidence < 85%
        pii_scan=True,               # Scan for PAN, Aadhaar, salary
        cite_sources=True,           # Require act/section citations
        drift_check=True,            # Check for answer drift
    )
)

# Check if response is fully safe
if response.is_safe:
    print(response.answer)
else:
    print(f"Escalation required: confidence={response.confidence}")

Async usage

import asyncio
import metacog

async def check_compliance():
    client = metacog.Metacog(api_key="mk-...")
    response = await client.governed.acreate(
        model=metacog.METACOG_M3,
        prompt="GSTR-3B due date for May 2026?",
        governance=metacog.GovernanceConfig(vertical="gst")
    )
    return response

response = asyncio.run(check_compliance())
print(response.answer)

Available models

Model	Use case	Speed	Governance
metacog-m1	High-volume queries, simple lookups	Fast	Standard
metacog-m3	Daily compliance work (recommended)	Balanced	Enhanced
metacog-m7	Complex reasoning, filing decisions	Thorough	Maximum
metacog-f1	Finance & payroll specialist	Fast	Standard
metacog-b1	Banking & RBI compliance	Balanced	Enhanced
metacog-h1	Healthcare compliance	Balanced	Enhanced

Verticals supported

• payroll — TDS, PF, ESI, Form 16, Professional Tax
• gst — GSTR-1/3B/9, ITC, E-invoice, RCM
• labour_law — Factories Act, Maternity, Minimum Wages, Gratuity, Bonus
• banking — RBI, KYC/AML, Basel III, NPA, CRR/SLR
• healthcare — ABDM, Clinical Act, Drug Act, DPDP
• insurance — IRDAI, Claims, Motor, Life, Health

Environment variables

export METACOG_API_KEY="mk-..."
export METACOG_BASE_URL="https://api.metacog.ai"  # optional

Get an API key

Visit metacog.ai or email [e-mail]

Free tier: 1,000 API calls/month. No credit card required.

Licensing

For embedding Metacog in your product (ERP, HRMS, fintech), contact [e-mail]

Revenue share program available for CA firms and software partners.

Links

• Documentation
• API Reference
• Governance Score
• Hallucination Rate Report
• Pricing

---

Metacog — Governed Intelligence. Made in India. Built for the world.