Code trust, with an expiration date.

Certify continuously evaluates every code unit in your repository, scores it against versioned policies, and assigns time-bound certification you can actually trust.

CI tells you whether code passes right now. Certify tells you whether code should still be trusted.

📋 Report Card → · 🌐 Full Report (Site) → · 📖 Documentation → · 🧩 VSCode Extension →

Code that once passed review doesn't stay trustworthy forever. Standards evolve, dependencies change, systems grow more complex.

Certify introduces continuous code certification — measurable quality scores with certification that expires intentionally. When certification lapses, code must be re-evaluated against current standards.

Instead of treating quality as a one-time event, Certify makes it a continuous process of trust, verification, and renewal.

go install github.com/iksnae/code-certification/cmd/certify@latest

Or build from source:

git clone https://github.com/iksnae/code-certification.git
cd code-certification
go build -o certify ./cmd/certify/

Requires: Go 1.22+, Git

See certification grades inline on every function, open an interactive dashboard, and configure AI providers visually.

code --install-extension iksnae.certify-vscode

Features: CodeLens grade annotations (Go, TypeScript, Python, Rust) · Dashboard with deep analysis metrics · Tree View sidebar · Status Bar badge · Diagnostics for failing units + dead code + high fan-in · Provider configurator with 10 presets + custom endpoints · Model browser with live discovery · VS Code Settings integration

Extension guide →

cd your-repo

# 1. Bootstrap — creates config, policies, and CI workflows
certify init

# 2. Discover — finds every function, method, type, and file
certify scan

# 3. Certify — collects evidence, evaluates, scores
certify certify

# 4. Report — generates your report card
certify report --format full

Your report card is at .certification/REPORT_CARD.md.

For large repos, generate the interactive site:

certify report --site
open .certification/site/index.html

A complete per-unit certification of your entire codebase:

# 🟢 Certify — Full Report

## Summary
| Overall Grade | 🟢 B |
| Total Units   | 474  |
| Pass Rate     | 100% |

## Dimension Averages
| correctness              | 95.0% | ██████████████████░░ |
| maintainability          | 93.3% | ██████████████████░░ |
| readability              | 92.4% | ██████████████████░░ |
| testability              | 90.0% | █████████████████░░░ |
| security                 | 80.0% | ████████████████░░░░ |

## All Units (organized by directory)
| Score         | function | B | 86.7% | certified | 2026-06-07 |
| CertifyUnit   | function | B | 85.6% | certified | 2026-06-07 |
... every unit in your repo

Add a live badge to your README — it updates automatically:

certify report --badge

Outputs:

[![Certification](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/OWNER/REPO/main/.certification/badge.json)](https://github.com/OWNER/REPO/blob/main/.certification/REPORT_CARD.md)

Click the badge → full report card.

For repos with hundreds or thousands of units, generate a self-contained HTML site:

certify report --site
open .certification/site/index.html

The site includes:

• Dashboard — summary stats, grade distribution, dimension averages, top issues
• Package pages — per-directory roll-ups with sortable unit tables
• Unit pages — detail view with dimension scores, AI observations, actions
• Client-side search — find any unit instantly across the entire codebase

Works offline via file:// — no server needed, no external dependencies. At 559 units, generates 584 pages in under 2 seconds.

certify init generates GitHub Actions workflows:

• PR — Certifies changed files, posts review summary
• Nightly — Sweeps for expired certifications
• Weekly — Full certification run + report card update

Every code unit is scored across 9 quality dimensions:

Dimensions are weighted and combined into a single score → grade (A through F).

Certifications are time-bound — they expire by design.

Default certification window: 90 days. Risk factors adjust the window — high churn shortens it, stable code extends it.

certify certify --skip-agent         # deterministic only, no LLM review
certify certify --batch 20           # process 20 units at a time
certify certify --diff-base main     # only changed files (for PRs)
certify certify --target internal/   # scope to specific paths

certify report --format full         # complete report card (markdown)
certify report --format card         # terminal report card
certify report --format json         # machine-readable
certify report --site                # interactive HTML site (584+ pages)
certify report --badge               # print README badge snippet
certify report --output report.md    # write to file

certify init creates .certification/config.yml:

mode: advisory        # advisory (report only) or enforcing (block on failure)

scope:
  include: []         # empty = everything
  exclude:
    - "vendor/**"
    - "node_modules/**"
    - "**/*_test.go"

expiry:
  default_window_days: 90

Add YAML policy packs to .certification/policies/:

name: my-team-standards
version: "1.0.0"
language: go

rules:
  - id: no-todos
    dimension: readability
    description: "No TODO comments in certified code"
    severity: warning
    metric: todo_count
    threshold: 0

  - id: low-complexity
    dimension: maintainability
    description: "Cyclomatic complexity under 15"
    severity: error
    metric: cyclomatic_complexity
    threshold: 15

Certify auto-detects available AI providers and enables conservative review — no config changes needed:

Just set an env var or start a local server — Certify handles the rest. Any OpenAI-compatible endpoint works.

For full control, explicitly configure in .certification/config.yml:

agent:
  enabled: true
  provider:
    type: openai-compatible
    base_url: https://api.openai.com/v1    # or any provider URL
    api_key_env: OPENAI_API_KEY
  models:
    prescreen: gpt-4o-mini
    review: gpt-4o-mini
    scoring: gpt-4o-mini

Agent review supplements — it never overrides — deterministic evidence. Certify works fully without it.

Run certify doctor to see which analysis tiers are available for each language in your environment.

.certification/
├── config.yml          # configuration
├── policies/           # policy packs
├── records/            # per-unit certification records (gitignored)
├── overrides/          # human governance overrides
├── state.json          # certification state snapshot (tracked)
├── runs.jsonl          # certification run history (tracked)
├── REPORT_CARD.md      # ← the report card
├── badge.json          # shields.io badge endpoint
├── reports/            # per-unit markdown reports (gitignored)
└── site/               # interactive HTML report (gitignored)
    ├── index.html      # dashboard
    ├── packages/       # per-directory roll-ups
    ├── units/          # per-unit detail pages
    └── search-index.js # client-side search data

• Brand Guide — identity, terminology, visual direction
• Architecture — system design and package structure
• Policy Authoring — writing custom policy packs
• Troubleshooting — common issues and solutions
• Contributor Guide — development setup

MIT