Skip to content

aws-samples/sample-aws-india-compliance-mcp

BranchesTags

Repository files navigation

AWS India Compliance MCP Server

An MCP server that assesses AWS infrastructure against four Indian regulatory frameworks:

  • DPDP Act 2023 (Digital Personal Data Protection) + Rules 2025
  • RBI Master Direction on IT Governance 2023
  • SEBI CSCRF 2024 (Cybersecurity and Cyber Resilience Framework)
  • CERT-In Directions 2022 (Incident reporting, log retention, NTP sync)

Works with Kiro, Claude Desktop, Cursor, or any MCP-compatible client.

Important: This tool provides automated assessment guidance based on published regulatory frameworks. It does not constitute legal advice or compliance certification. Organizations should consult qualified compliance and legal professionals for definitive regulatory compliance determinations.

Quick start

1. Install

git clone https://github.com/aws-samples/sample-aws-india-compliance-mcp.git
cd sample-aws-india-compliance-mcp

# Create a virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install
pip install .

# Or with pinned dependencies
pip install . -c constraints.txt

2. Configure AWS credentials

The server needs read-only AWS access. Use AWS IAM Identity Center (SSO) — this is the recommended approach for teams.

First-time SSO setup:

aws configure sso

Follow the prompts:

  • SSO session name: my-sso (any name)
  • SSO start URL: your org's SSO URL (e.g., https://d-xxxxxxxxxx.awsapps.com/start)
  • SSO region: us-east-1 (or wherever your Identity Center is)
  • It opens a browser — sign in and authorize
  • Select the account and role (e.g., ReadOnlyAccess)
  • CLI profile name: my-sso-profile (remember this)

Login before each session:

aws sso login --profile my-sso-profile

This opens a browser for authentication. Once approved, your CLI has temporary credentials for ~8 hours.

Verify it works:

aws sts get-caller-identity --profile my-sso-profile

Alternative: static credentials (not recommended for teams) If you have long-lived access keys in ~/.aws/credentials, those work too — but SSO is preferred for security.

3. Add to your MCP client

Add to .kiro/settings/mcp.json (or claude_desktop_config.json for Claude Desktop):

{
  "mcpServers": {
    "aws-india-compliance": {
      "command": "/path/to/your/.venv/bin/python3",
      "args": ["-m", "aws_india_compliance.server"],
      "env": {
        "PYTHONPATH": "/path/to/sample-aws-india-compliance-mcp/src",
        "AWS_PROFILE": "my-sso-profile",
        "LOG_LEVEL": "INFO"
      }
    }
  }
}

Replace /path/to/your/.venv/bin/python3 with the actual path to your venv Python, and my-sso-profile with your SSO profile name.

Tip: Run which python3 inside your activated venv to get the exact path.

4. Verify

Ask your MCP client:

"List the DPDP control domains"

If it returns 10 domains, you're set. To scan your AWS account:

"Scan my AWS account in ap-south-1 for DPDP and RBI compliance"

Prerequisites

  • Python 3.10+
  • AWS Config recorder enabled in target accounts/regions
  • IAM credentials with read-only access (see IAM policy below)
  • For org-wide scans: a Config Aggregator name

Tools (11)

Tool Purpose
scan_aws_account Discover resources via AWS Config, assess against all frameworks. Pass aggregator_name for org-wide scans. Supports tag filtering, exception rules, and SEBI entity tiering.
scan_control_tower_tool Enumerate enabled guardrails across OUs, recommend missing ones per framework including CERT-In. Per-OU breakdown with domain coverage.
parse_architecture Parse CloudFormation (JSON/YAML), Terraform (HCL), or draw.io (XML) templates. Max 10 MB.
assess_compliance Assess a component list against control domains. Supports tag filtering, exception suppression, and SEBI entity tiering.
generate_report Full report with posture scores, gap list, and phased remediation timeline.
search_regulatory_text Search regulatory text from government sources. Falls back to bundled mappings when sites are unreachable.
list_control_domains List domains for a framework: dpdp (10), rbi (7), sebi (6), or certin (4).
check_regulatory_updates Staleness check + content hash monitoring + new circular detection. Flags when mappings may be outdated.
propose_mapping_update Feed new regulatory text to the LLM client for analysis. Returns current mappings + structured prompt for proposing changes.
apply_mapping_update Validate and apply LLM-proposed mapping changes to control_mappings.json. Supports review-then-apply workflow.
format_report Convert scan report JSON into Markdown or production-grade DOCX. Pass output_format="docx" for a styled Word report with color-coded risk levels, posture scores, cover page, and Control Tower guardrails.

Key features

Confidence scoring

Every compliance gap includes a confidence level:

  • High — Direct technical check verifiable from AWS Config (e.g., encryption disabled, public access enabled)
  • Medium — Interpretive mapping from regulatory requirement to AWS control (e.g., data localization)
  • Low — Organizational requirement where infrastructure is only a proxy (e.g., DPO appointment, consent tracking)

Each gap also carries evidence (triggering property values), checked_at (ISO 8601 timestamp), and confidence_rationale.

CERT-In Directions 2022

Four control domains assessed:

  1. Incident Reporting Readiness (GuardDuty + EventBridge + SNS pipeline)
  2. Log Retention — 180 days (CloudWatch LogGroup retention check)
  3. NTP Synchronization (advisory — AWS uses Amazon Time Sync by default)
  4. Reportable Incident Awareness (Security Hub enablement)

Per-account breakdown

Org-wide scans group gaps by AWS account ID with individual DPDP, RBI, and SEBI posture scores per account.

Resource-level compliance tracking

Per-domain resource compliance percentages (e.g., "dpdp:6 — 482 checked, 320 passed, 66.4%").

Exception management

  • Auto-suppression of AWSControlTowerExecution roles (by design)
  • Custom exception rules via resource_pattern (fnmatch) and exclude_tag
  • Suppressed gaps tracked separately with suppression reason

Tag-based filtering

  • filter_tags: Include only resources matching ALL specified tag key-value pairs
  • exclude_tags: Exclude resources matching ANY specified tag key-value pair

SEBI entity tiering

  • MII tier: C-SOC readiness checks (GuardDuty + Security Hub + Detective), BYOK verification
  • qualified_re tier: BYOK check, skip C-SOC
  • other_re tier: Baseline SEBI checks only

Nuanced data localization

  • Storage resources outside India → high-risk gap (RBI 2018 Circular)
  • Compute resources outside India → medium-risk advisory (processing may be permissible)
  • Global services (IAM, CloudFront, Route53, WAFv2) → excluded from localization checks
  • DataClassification tag value included in gap description when present

Report formatting

The format_report tool supports two output formats:

Markdown (default): Structured text report for chat display and quick review.

DOCX (output_format="docx"): Production-grade Word document suitable for sharing with customers and auditors. Features:

  • Professional cover page with scan metadata and confidentiality notice
  • Color-coded posture scores (green/orange/red based on thresholds)
  • Dark blue styled table headers with alternating row shading
  • Risk-level cell coloring (red for critical, orange for high, yellow for medium)
  • Confidence-level cell coloring (green for high, orange for medium, red for low)
  • Landscape orientation with full-width tables (no truncation of resource names)
  • Control Tower section with guardrails, per-OU breakdown, and recommendations
  • Phased remediation timeline
  • Disclaimer page

Both formats include:

  • Executive summary with posture scores
  • Confidence distribution table
  • Gap summary by framework and risk level
  • Resource-level compliance pass rates
  • Per-account breakdown
  • Critical and high-risk gap tables with evidence
  • Suppressed gaps with reasons
  • Phased remediation timeline (Immediate / Short-term / Medium-term)

Large scan results (>100 gaps) are automatically saved to reports/ as JSON, with the MCP response trimmed to priority gaps.

Regulatory monitoring

Three-tier monitoring system:

  1. Staleness — Flags frameworks where last_verified exceeds threshold (default 30 days)
  2. Content hashing — Fetches regulatory source pages and compares SHA-256 hashes against baselines
  3. New circular detection — Scans RBI and SEBI circular listing pages for publications matching compliance keywords

LLM-assisted mapping updates via propose_mapping_update → apply_mapping_update workflow.

How scanning works

  1. AWS Config Advanced Query pulls resource configurations in a single API call.
  2. The scanner extracts compliance-relevant properties per resource type (encryption, public access, logging, retention, key rotation, TLS enforcement, VPC flow logs, security group rules, secrets rotation, backup plans, etc.).
  3. Fallback API checks cover Security Hub, GuardDuty, CloudTrail, WAF, AWS Backup, and Amazon Inspector.
  4. The assessment engine evaluates each resource against applicable DPDP, RBI, SEBI, and CERT-In control domains.
  5. Results include risk-rated gaps with confidence levels, evidence, specific remediation steps, and regulatory section references.

For org-wide scans, pass the Config Aggregator name (e.g., aws-controltower-ConfigAggregatorForOrganizations). The aggregator must be configured in the management or delegated admin account.

Resource checks

Resource What gets checked
S3 Encryption at rest, lifecycle policies, Block Public Access, versioning, access logging, Object Lock (audit buckets), TLS enforcement (bucket policy)
RDS Storage encryption, public accessibility, Multi-AZ, audit logging, SSL enforcement
DynamoDB KMS encryption, TTL, point-in-time recovery
Lambda Secrets in environment variables, dead letter queue
EC2 Public IP assignment, IMDSv2 enforcement, EBS encryption
EKS Secrets envelope encryption, API server endpoint visibility, control plane logging
ECS Container Insights
CloudTrail Log file validation, KMS encryption, CloudWatch Logs integration
KMS Automatic key rotation for customer-managed keys, BYOK verification (SEBI)
API Gateway WAF association
CloudFront WAF association, access logging
SQS/SNS Encryption at rest
SageMaker Direct internet access, KMS encryption, VPC configuration, consent tracking for ML data
IAM Roles Overprivileged policies (AdministratorAccess, PowerUserAccess, IAMFullAccess)
VPC Flow Logs enablement and destination
Security Groups Open SSH (22) and RDP (3389) to 0.0.0.0/0
Secrets Manager Automatic rotation configuration
AWS Backup Backup plan existence, Vault Lock status
Amazon Inspector Enablement status (SEBI VAPT requirement)

RBI-regulated scans additionally flag resources deployed outside ap-south-1 and ap-south-2 per the RBI Data Localization Circular 2018, with nuanced classification (storage vs compute vs global services).

Control domains

DPDP Act (10 domains): Lawful Processing, Data Minimization, Privacy Notices, Data Principal Rights, Breach Notification, Reasonable Security Safeguards, Data Retention Limits, Cross-Border Data Transfer, Children's Data Protection, Significant Data Fiduciary Obligations.

RBI Master Direction (7 domains): IT Governance, IT Infrastructure, IT Risk Management, Information Security, Cyber Security, Business Continuity/DR, Information Systems Audit.

SEBI CSCRF (6 domains): Cyber Governance, Cyber Risk Identification, Cyber Protection, Cyber Detection, Cyber Response, Cyber Recovery.

CERT-In Directions (4 domains): Incident Reporting Readiness, Log Retention (180 days), NTP Synchronization, Reportable Incident Awareness.

Regulatory version tracking

Framework Version Source
DPDP Act As enacted August 11, 2023 + Rules 2025 (Nov 14, 2025) dpdpact.in
RBI Master Direction DoS.CO.CSITE.SEC.3/31.01.015/2023-24 (April 7, 2023) rbi.org.in
SEBI CSCRF Circular SEBI/HO/ITD/ITD-SEC-1/P/CIR/2024/113 (August 20, 2024) sebi.gov.in
CERT-In Directions Directions dated April 28, 2022 cert-in.org.in

Control mappings are maintained in control_mappings.json with manifest_version, last_verified dates, and source URLs per framework. Run check_regulatory_updates to see when mappings were last verified and where to check for new publications.

Security

This server performs read-only operations. It does not modify AWS resources.

Input validation:

  • AWS region format validated via regex (^[a-z]{2}(-[a-z]+-\d+)?$)
  • Config Aggregator name validated (^[a-zA-Z0-9_-]{1,256}$)
  • Report file paths constrained to reports/ directory (path traversal blocked)
  • top_k parameter capped at 50
  • Error messages sanitized to avoid leaking filesystem paths

Transport: stdio by default (local process, no network exposure). For remote deployment over HTTP, set MCP_API_KEY and use OAuth 2.1 or equivalent authentication.

Credentials: Use IAM roles or SSO profiles. Do not hardcode credentials in config files or source code.

Data handling: No persistence, no telemetry, no caching by default. Scan results are held in memory and discarded on exit. Large scan reports are saved to reports/ (gitignored). Logs (stderr, INFO level) contain resource ARNs and type identifiers but not credential material or data values.

XML parsing: draw.io templates are parsed with defusedxml, which blocks XXE, DTD processing, and entity expansion.

Outbound network: HTTPS-only calls to regulatory sites (dpdpact.in, rbi.org.in, sebi.gov.in, cert-in.org.in). Domain allowlist enforced. Response size capped at 5 MB. Rate-limited to 10 requests/minute per domain. 30-second timeout.

Minimum IAM policy

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "ConfigAdvancedQuery",
      "Effect": "Allow",
      "Action": [
        "config:SelectAggregateResourceConfig",
        "config:SelectResourceConfig",
        "config:DescribeConfigurationRecorders",
        "config:DescribeConfigurationAggregators"
      ],
      "Resource": "*"
    },
    {
      "Sid": "SecurityServices",
      "Effect": "Allow",
      "Action": [
        "securityhub:GetFindings",
        "securityhub:DescribeHub",
        "guardduty:ListDetectors",
        "guardduty:GetDetector",
        "cloudtrail:DescribeTrails",
        "cloudtrail:GetTrailStatus",
        "wafv2:ListWebACLs",
        "wafv2:GetWebACL",
        "backup:ListBackupPlans",
        "backup:ListBackupVaults",
        "inspector2:BatchGetAccountStatus"
      ],
      "Resource": "*"
    },
    {
      "Sid": "ControlTower",
      "Effect": "Allow",
      "Action": [
        "controltower:ListLandingZones",
        "controltower:GetLandingZone",
        "controltower:ListEnabledControls",
        "organizations:ListRoots",
        "organizations:ListOrganizationalUnitsForParent"
      ],
      "Resource": "*"
    }
  ]
}

For org-wide scans, add config:SelectAggregateResourceConfig on the aggregator ARN. Consider adding an explicit Deny statement for destructive actions (DeleteTrail, StopLogging, DeleteDetector, etc.).

Project structure

src/aws_india_compliance/
  server.py              # MCP tool registration, entry point, input validation
  assessment.py          # Compliance assessment engine (confidence scoring, resource tracking, exceptions)
  aws_scanner.py         # AWS Config query + fallback API checks (Backup, Inspector)
  control_tower.py       # Control Tower scanner + per-OU guardrail mapping
  parsers.py             # CloudFormation/Terraform/draw.io parsers
  knowledge.py           # Live regulatory text search + content hash monitoring
  domains.py             # Domain definitions + manifest loader + staleness check
  report_formatter.py    # Markdown report generator (account scan + Control Tower)
  control_mappings.json  # Versioned control-to-AWS mapping manifest (DPDP, RBI, SEBI, CERT-In)
tests/
  test_assessment.py     # Assessment engine tests
  test_aws_scanner.py    # Scanner tests (mocked boto3)
  test_control_tower.py  # Control Tower assessment tests
  test_domains.py        # Domain definition + manifest tests
  test_parsers.py        # Parser tests
reports/                 # Scan report output (gitignored)

Tests

PYTHONPATH=src python3 -m pytest tests/ -v

28 tests covering assessment logic, scanner component extraction, Control Tower gap analysis, domain definitions, manifest integrity, and all three parsers.

Environment variables

Variable Default Purpose
AWS_PROFILE — AWS SSO profile name
LOG_LEVEL INFO Logging level (DEBUG, INFO, WARNING, ERROR)
REGULATORY_CACHE_TTL 0 Seconds to cache regulatory site responses. 0 = no caching.
STALENESS_THRESHOLD_DAYS 30 Days after last_verified before staleness warnings appear in scan results.
MCP_TRANSPORT stdio Transport mode: stdio for local, streamable-http for remote
MCP_HOST 127.0.0.1 Host for HTTP transport
MCP_PORT 8000 Port for HTTP transport (validated 1-65535)
REPORT_DIR reports/ Directory for full scan report JSON output

Disclaimer: This is a sample tool for educational and assessment purposes. It performs read-only operations and does not modify AWS resources. Users should validate compliance findings against their specific regulatory requirements and consult qualified compliance professionals.

License

Apache 2.0

About

No description, website, or topics provided.

Resources

Readme

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages

Generated from amazon-archives/__template_MIT-0