iac-scanner MCP server
Commit Nov 14, 2026Evaluates Terraform and AWS CDK infrastructure against security benchmarks and generates AI-powered code fixes.
Also known as alphacrack/iac-scanner
- Stars
- 4
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 14, 2026
- License
- Apache-2.0
What is iac-scanner MCP?
iac-scanner evaluates Terraform and AWS CDK infrastructure against security benchmarks and generates AI-powered code fixes.
Summary generated from the project's README
About iac-scanner MCP server
Securing infrastructure-as-code by identifying vulnerabilities and generating automated corrections is what iac-scanner enables for DevOps engineers. This tool evaluates Terraform and AWS CDK configurations against security benchmarks to find risks and provide actionable fixes. It works as a local CLI or an MCP server that can be driven directly from Claude Desktop or Cursor.
Users can choose from several execution paths depending on their privacy and cost requirements. You can run scans using GitHub Models for free, use Ollama for fully local and offline processing, or connect to OpenAI and Anthropic via API keys. The tool also supports a hybrid mode by using Checkov to ground findings with CWE, CIS, and NIST tags. To prevent unexpected expenses, it includes a feature to abort runs if projected LLM costs exceed a specified dollar amount.
What you can do with it
- 1Automated security scanningScan a directory of Terraform files to identify security and best-practice violations.
- 2Integrate with CI/CD pipelinesFail a build process if the scanner detects any high or critical severity issues.
- 3Local offline analysisPerform infrastructure audits without an internet connection using local models.
Written from the project's README. Every use case is checked against a quote from it.
Configuration
GITHUB_TOKENOPENAI_API_KEY
| Name | Description | |
|---|---|---|
| GITHUB_TOKEN | — | |
| OPENAI_API_KEY | — |
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablealphacrack/iac-scanner is public and reachablegithub.com/alphacrack/iac-scanner ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-19github.com/alphacrack/iac-scanner ↗Pass
- OpenSSF ScorecardOpenSSF Scorecard 4.2 / 10; weakest: Pinned-Dependencies 0, Token-Permissions 0, CII-Best-Practices 0scorecard.dev/viewer/?uri=github.com/alphacrack/iac-scanner ↗Partial 6/15
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubIaC Scanner
Python CLI that scans Terraform and AWS CDK Infrastructure-as-Code, reports security + best-practice findings, and writes the fix. Runs locally, keylessly, or grounded by Checkov.
iac-scanner complements rule-based scanners like Checkov, tfsec, and KICS — it doesn't replace them. Its differentiator is AI-generated fixes alongside findings, and keyless operation via GitHub Models, Ollama, or an MCP server.
AI-generated output — review before applying. Fixed code is written to
scan-output/fixed/and must be human-reviewed before overwriting your working tree. No auto-apply.
License: Apache License 2.0.
Quickstart (30 seconds, no cost)
Pick whichever path matches your setup:
GitHub Models (free for any GitHub user)
pip install iac-scanner
export GITHUB_TOKEN=$(gh auth token)
iac-scan scan ./my-tf --provider github -o ./out
Ollama (fully local, offline, free)
pip install iac-scanner[local]
ollama pull qwen2.5-coder:7b-instruct # first run only
iac-scan scan ./my-tf --provider ollama -o ./out
MCP server — drive from Claude Desktop / Cursor
pip install iac-scanner[mcp]
# Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
# "mcpServers": { "iac-scanner": { "command": "iac-scan-mcp" } }
# Then in Claude: "Scan the Terraform in ~/work/infra"
OpenAI / Anthropic (bring your own key)
pip install iac-scanner
export OPENAI_API_KEY=sk-...
iac-scan scan ./my-tf --provider openai -o ./out
# or: iac-scan scan ./my-tf --provider auto (picks ollama → github → openai → anthropic)
No-network demo (no AI, just parse)
iac-scan scan ./samples/tf -o ./out --scan-only
Every mode writes scan-report.json (findings) and, when AI is enabled, fixed/ (corrected code).
Input (CLI)
- Terraform: a directory containing
main.tf, or the path tomain.tfitself. Sibling.tffiles are included. - CDK: a directory containing
index.ts/index.js, or the path to that file.lib/andbin/subdirectories are included.
What we skip (automatic)
terraform.tfstate*, *.tfvars, .env*, *.pem, *.key, id_rsa*, .terraform/, node_modules/, cdk.out/. See SECURITY.md for the full skip-list and threat model.
Process
- Factory creates the right scanner (
TerraformScannerorCdkScanner) from the given path. - Scan: load entry file(s), apply the skip-list, redact obvious secrets, enforce the 200 KB input cap.
- (Optional) Rule engine pre-pass (
--rules-engine=checkov) adds framework-mapped findings with CWE/CIS/NIST tags. - Analysis (LLM, structured output): findings as a Pydantic-validated JSON array with severity + location.
- Fix (LLM, text output): regenerates corrected code with a mandatory
AI-generated — review before applyingbanner. - Output: JSON and/or SARIF 2.1.0 report; fixed files under
fixed/.
Output
- JSON report (
scan-report.json):iac_type,entry_path,findings,metadata,provider,analysis_model,fix_model,prompt_version. - SARIF 2.1.0 (
--format sarif|both): consumed by GitHub Code Scanning, GitLab Security Dashboards, SonarQube. - Fixed code (
fixed/): multi-file output preserves the original layout; each file starts with the AI-generated banner.
Usage
# Basic scan (analysis + fix)
iac-scan scan ./my-tf-dir
# No-AI parse only (no keys needed)
iac-scan scan ./my-tf-dir --scan-only
# Findings only, skip fix generation
iac-scan scan ./my-tf-dir --no-fix
# Output SARIF for GitHub Code Scanning
iac-scan scan ./my-tf-dir --format sarif -o ./out
# Or both at once
iac-scan scan ./my-tf-dir --format both -o ./out
# Choose a provider explicitly
iac-scan scan ./my-tf-dir --provider github
iac-scan scan ./my-tf-dir --provider ollama
iac-scan scan ./my-tf-dir --provider openai
# Ground the LLM with Checkov rule findings (hybrid mode)
pip install iac-scanner[rules]
iac-scan scan ./my-tf-dir --rules-engine checkov
# CI gate: exit non-zero on any HIGH or CRITICAL finding
iac-scan scan ./my-tf-dir --fail-on high
# Cost cap: abort if projected LLM cost exceeds $0.50
iac-scan scan ./my-tf-dir --max-spend 0.50
# Force-refresh: skip the response cache for this run
iac-scan scan ./my-tf-dir --no-cache
Environment variables
| Variable | Purpose |
|---|---|
IAC_PROVIDER |
openai | anthropic | github | ollama (overrides auto-detect). |
OPENAI_API_KEY |
Required when --provider=openai. |
ANTHROPIC_API_KEY |
Required when --provider=anthropic. |
GITHUB_TOKEN |
Required when --provider=github. Any gh auth token works (free tier). |
OLLAMA_HOST |
Ollama endpoint. Default: http://localhost:11434. |
IAC_ANALYSIS_MODEL |
Override analysis model (e.g. gpt-4o, claude-3-5-sonnet-20241022). |
IAC_FIX_MODEL |
Override fix model. |
IAC_MAX_SPEND_USD |
Hard dollar cap per run. Abort if projected cost exceeds it. |
IAC_MAX_INPUT_BYTES |
Input size cap (default 200 KB, floored to 1 KB, ceilinged to 10 MB). |
IAC_NO_CACHE |
When set, skip the content-addressed response cache. |
IAC_NO_REDACT |
Disable secret redaction (not recommended — see SECURITY.md). |
IAC_CACHE_DIR |
Override cache directory (default ~/.cache/iac-scanner/). |
IAC_OUTPUT_FORMAT |
Default output format (json | sarif | both). |
Install
# Base install — includes OpenAI and Anthropic providers + GitHub Models
pip install iac-scanner
# With optional extras
pip install iac-scanner[local] # Ollama local-LLM provider
pip install iac-scanner[mcp] # MCP server mode for Claude Desktop / Cursor
pip install iac-scanner[rules] # Checkov hybrid mode
pip install iac-scanner[all] # all of the above
# From source
git clone https://github.com/alphacrack/iac-scanner
cd iac-scanner
pip install -e ".[dev]"
Blog and tutorial
Articles and a step-by-step tutorial are published on GitHub Pages at https://alphacrack.github.io/iac-scanner/. Source lives under docs/.
Contributing
See CONTRIBUTING.md for development setup, test strategy, and the release process (Trusted Publishing + SBOM + Sigstore signing).
Governance & support
- GOVERNANCE.md — roles, decision-making, branch protection policy.
- MAINTAINERS.md — canonical list of maintainers.
- SUPPORT.md — where to ask questions, file bugs, and request features.
- CODE_OF_CONDUCT.md — Contributor Covenant 2.1.
Security
See SECURITY.md for the threat model (prompt injection, secret exposure, hallucinated fixes, supply-chain, cost abuse) and private disclosure channel.
Project layout
src/iac_scanner/
cli.py # CLI entry (click)
factory.py # create_scanner(path) → TerraformScanner | CdkScanner
models.py # Pydantic: Finding, FindingsList, ScanReport, VerificationResult
cache.py # content-addressed SHA-256 response cache
cost.py # tiktoken preflight + IAC_MAX_SPEND_USD enforcement
mcp_server.py # iac-scan-mcp entry — MCP server for host LLMs
scanners/
base.py # IacScanner (abstract), ScanResult
_filters.py # skip-list, secret redaction, input size cap
terraform.py # TerraformScanner (main.tf)
cdk.py # CdkScanner (index.ts / index.js)
llm/
providers.py # LLMClient + OpenAI / Anthropic / GitHub Models / Ollama
orchestration/
tasks.py # analysis + fix LangChain tasks (structured output, XML fencing)
runner.py # run_pipeline: scan → cache → cost check → LLM → result
hybrid.py # rule-pre-pass + LLM augment + dedupe
rules/
engine.py # rule-engine dispatcher
checkov.py # Checkov subprocess adapter with CWE/CIS/NIST mapping
output/
report.py # write_report_and_fixes — JSON + fixed/ banner
sarif.py # SARIF 2.1.0 emitter
iac-scanner MCP questions
Does iac-scanner replace tools like Checkov?
No, it is designed to complement rule-based scanners like Checkov, tfsec, and KICS by providing AI-generated fixes.
Will the AI automatically apply fixes to my code?
No, fixed code is written to a specific directory and must be manually reviewed by a human before being applied.
Can I run this scanner without an internet connection?
Yes, you can use Ollama for a fully local and offline experience.