Open source · Built for GitHub Actions

See the cloud APIs your code actually uses.

CCE Action reads your source code and turns cloud SDK calls into clear, reviewable evidence. Add one workflow to inventory APIs, review changes, and prepare IAM access requests.

  • No account required
  • Diagnostics off by default
  • Reports stay with your workflow
Cloud API inventory

payments-api

Ready
AWS
Amazon S3PutObject
2 call sites
GCP
Cloud Storagebucket
1 call site
Azure
Blob StorageUpload
3 call sites
Code in. Evidence out.
Made for code review
Works across your stack
  • Go
  • Python
  • Java
  • JavaScript
  • AWS
  • Azure
  • Google Cloud

Useful on the first run

From “what does this service use?” to evidence you can share.

Cloud usage is often spread across code, tickets, and IAM policies. CCE gives developers and reviewers a common starting point.

01

Build an API inventory

Find provider, resource, operation, and source-location evidence across supported languages. Export a machine-readable report for your own tools.

02

Review what changed

See added and removed Cloud API calls in one pull-request comment. Keep the JSON, Markdown, and SARIF evidence for later.

03

Start an access conversation

Give IAM reviewers a concise list of operations found in code. They decide the final policy and resource boundaries.

Quick start

One small workflow.
A useful first report.

Create .github/workflows/cce.yml in your repository. The workflow checks out your code, runs CCE, and uploads the results.

  1. 1Add the workflow and commit it.
  2. 2Open the CCE run in GitHub Actions.
  3. 3Download the cce-results artifact.
Open the complete example
.github/workflows/cce.ymlYAML
name: CCE
on: [push, pull_request]

permissions:
  contents: read
  packages: read

jobs:
  scan:
    uses: stackgenhq/cce-action/.github/workflows/scan.yml@v1
    with:
      language: AUTO
      filter: cloud
Prefer a fixed version? Use @v0.3.0 instead of @v1. The major tag receives compatible updates; an exact release stays fixed.

A real repository, a real result

See what the JSON tells you.

Scan aws-samples/aws-lambda-extensions at a pinned commit. CCE 0.0.38 examined 313 Python files and emitted 10 AWS evidence rows, including this Secrets Manager call.

Try it without setting up AWS

One workflow. One downloadable report.

The example checks out the public repository, runs the Python scan, and uploads cce-public-repo-results. No AWS credentials, deployed application, or CCE account is needed.

  1. 1Copy the workflow into any repository.
  2. 2Run it manually from GitHub Actions.
  3. 3Download the JSON and inspect the source evidence.

Copy the public-repo workflow

Read the full field guide

Download the complete JSON (all 10 rows)

cce-results.json (verified excerpt)JSON
{
  "files_scanned": 313,
  "summary": {
    "total_entitlements": 10,
    "by_provider": { "AWS": 10 }
  },
  "example": {
    "provider": "AWS",
    "resource": "secretsmanager",
    "operation": "get_secret_value",
    "file": "python-example-telemetry-opensearch-extension/python-example-telemetry-opensearch-extension/telemetry_dispatcher.py",
    "line": 19,
    "column": 19,
    "method": "boto3.client('secretsmanager').get_secret_value"
  }
}
WHAT

Normalized API evidence

The row identifies AWS, Secrets Manager, and get_secret_value. The ten summary rows are evidence records, not ten unique permissions.

WHERE

A source location you can inspect

File, line, column, method, and signature connect the classification to the code. The full guide links directly to the pinned source line.

WHY

A better review starting point

An IAM reviewer can evaluate secretsmanager:GetSecretValue, then choose the secret ARN and validate the role. CCE does not grant access automatically.

Evidence, not a runtime guarantee.

The scan does not execute code, read secrets, prove runtime coverage, or test effective IAM. Use it alongside human review and runtime validation.

Choose your starting point

Focused workflows for everyday engineering.

Start with a scan, then add only the capabilities your team needs.

Scan

Create a Cloud API report, IAM-related exports, and a cloud-neutral E-SBOM.

View example

Pull-request review

Post a sticky Cloud API diff, publish SARIF, and optionally fail when new operations appear.

View example

Access request

Package code-evidenced operations as versioned JSON for review in your IAM process.

View example

Quickstart bundle

Share an HTML report, API inventory, SARIF, lens lockfile, and suggested CI workflow.

View example

Lens verification

Test a proposed lens against committed fixtures and catch regressions before merging.

View example
Typical lens inputsYAML
- uses: stackgenhq/cce-action@v1
  with:
    language: AUTO
    lens: .cce/internal-sdk.yaml
    filter: all
Local lensesContent hashes are calculated automatically.
Remote lensesHTTPS and an expected SHA-256 are required.

Make CCE speak your language

Your code has its own vocabulary. A lens teaches CCE to recognize it.

A lens is a small YAML file that describes the libraries or patterns you want to find. Use one to inventory internal SDKs, legacy clients, or technology-debt patterns alongside built-in cloud mappings.

  • Keep lens definitions with your code.
  • Pin remote content so changes are visible.
  • Test changes against fixtures before rollout.
Read the plain-language lens guide

Designed for responsible review

Useful evidence. Clear boundaries.

CCE supports engineering decisions; it does not replace testing, provider documentation, or an IAM reviewer.

Private by default

No CCE account is needed. Field diagnostics are off unless you explicitly enable them. Basic scans use read-only repository permissions.

Traceable results

Reports connect operations to source locations. Lens hashes and versioned artifacts help reviewers understand what produced a result.

No guessed permissions

Strict GCP and Azure exports do not publish unverified permission mappings. A missing role export is not silently replaced with a guess.

Static-analysis limits

Dynamic, reflected, or unrecognized wrapped calls may be missed. An empty result does not prove that a service needs no permissions.

IAM requests are review input, not ready-to-deploy policy.

CCE identifies operations in source code. Your reviewers remain responsible for resource scope, runtime behavior, organizational policy, and final approval.

Good to know

Practical details

Runner
Linux with Docker; ubuntu-latest is the simplest choice.
Languages
Go, Python, Java, and JavaScript. Use AUTO for detection.
Results
JSON reports, workflow artifacts, optional HTML and SARIF, and selected IAM exports.
Custom providers
Use filter: all when a lens reports categories outside the cloud filter.
Large repositories
Scan a subdirectory or set scan-timeout to keep CI bounded.
Fork pull requests
Review artifacts remain available; comments and SARIF publication are skipped when the token cannot write safely.

Common questions

Do I need to install CCE separately?

No. The reusable workflows and composite actions provide a tested CCE version for the run.

Can I use CCE inside an existing job?

Yes. Check out your repository, then use stackgenhq/cce-action@v1 or a focused subaction such as access-request@v1.

Where do my reports go?

They are written to the configured output directory and, in the reusable workflows, uploaded as GitHub Actions artifacts.

Is this a compliance or least-privilege verdict?

No. It is source-code evidence for review. Validate policies against runtime needs and your organization’s requirements.

Start with one repository

Make cloud usage easier to see, share, and review.