CLI Guide

Installation

pip install licit-ai-cli

Or from source:

git clone https://github.com/Diego303/licit-cli.git
cd licit-cli
pip install -e ".[dev]"

Invocation

# As an installed command
licit [options] <command> [arguments]

# As a Python module
python -m licit [options] <command> [arguments]

Global Options

licit --version
# licit, version 0.7.0

licit --verbose status
# Shows debug logs during execution

Commands

licit init

Initializes licit in the current project. Automatically detects project characteristics and generates the configuration.

licit init [--framework {eu-ai-act|owasp|all}]

Options:

What it does:

1. Runs ProjectDetector to detect languages, frameworks, CI/CD, AI agents, etc.
2. Generates .licit.yaml with a configuration tailored to the project.
3. Creates the .licit/ directory for internal data.
4. If architect or vigil are detected, automatically enables their connectors.

Example:

$ cd my-fastapi-project/
$ licit init

Initialized licit in my-fastapi-project
  Languages: python
  Frameworks: fastapi
  Agent configs: CLAUDE.md
  CI/CD: github-actions
  Config saved to .licit.yaml

Example with specific framework:

$ licit init --framework eu-ai-act
# Only enables EU AI Act, disables OWASP

licit status

Shows the current status of licit and connected data sources.

licit status

What it displays:

• Project information (name, languages, frameworks)
• Configuration status
• Enabled frameworks (EU AI Act, OWASP)
• Detected data sources (provenance, FRIA, changelog, etc.)
• Configured connectors (architect, vigil)
• Detected AI agent configurations

Example:

$ licit status

  licit Status
  ────────────────────────────────────────
  Project: my-fastapi-project
  Config: .licit.yaml

  Frameworks:
    [x] EU AI Act
    [x] OWASP Agentic Top 10
    [ ] NIST AI RMF (V1)
    [ ] ISO 42001 (V1)

  Data Sources:
    [x] Git history (142 commits)
    [x] Provenance tracking
    [ ] Config changelog
    [ ] FRIA document
    [ ] Annex IV documentation

  Connectors:
    [x] architect (.architect/config.yaml, enabled)
    [x] vigil (.vigil.yaml, enabled)
    Security findings: 3 total (1 critical, 1 high)

  Agent Configs (2):
    - CLAUDE.md (claude-code)
    - .cursorrules (cursor)

The status command now shows:

• Connector status as “enabled” or “detected”
• Security findings count when SARIF findings are present

licit connect

Configures optional connectors to integrate external data sources.

licit connect {architect|vigil} [--enable|--disable]

Arguments:

Options:

What it does:

1. Enables or disables the connector in .licit.yaml.
2. When enabling, auto-detects configuration paths if not already configured.
3. Verifies data availability on disk with available().
4. Shows feedback about whether connector data was found.

Example:

$ licit connect architect
  architect data found at: .architect/reports
  Connector 'architect' enabled.

$ licit connect vigil --enable
  vigil data found
  Connector 'vigil' enabled.

$ licit connect architect --disable
  Connector 'architect' disabled.

Data enriched by connectors:

licit trace

Tracks code provenance — identifies what was written by AI and what by humans.

Status: Functional (Phase 2 completed).

licit trace [--since DATE|TAG] [--report] [--stats]

Options:

What it does:

1. Runs GitAnalyzer to analyze commits with 6 heuristics (author, message, volume, co-authors, file patterns, time).
2. Optionally reads agent session logs (Claude Code).
3. Classifies each file as ai (score >= 0.7), mixed (>= 0.5), or human (< 0.5).
4. Stores results in .licit/provenance.jsonl (append-only).
5. If sign: true, signs each record with HMAC-SHA256.

Example:

$ licit trace --since 2026-01-01 --stats

  Analyzing git history...
  Records: 45 files analyzed
  AI-generated: 18 (40.0%)
  Human-written: 22 (48.9%)
  Mixed: 5 (11.1%)

  AI tools detected: claude-code (15), cursor (3)
  Models detected: claude-sonnet-4 (12), claude-opus-4 (3), gpt-4o (3)

  Stored in .licit/provenance.jsonl

Example with report:

$ licit trace --report
# Generates .licit/reports/provenance.md with a detailed per-file table

Heuristics used:

Only heuristics that produce a signal (score > 0) contribute to the weighted average.

licit changelog

Generates a changelog of changes in AI agent configurations with semantic diffing and severity classification.

Status: Functional (Phase 3 completed).

licit changelog [--since DATE|TAG] [--format {markdown|json}]

Options:

What it does:

1. Runs ConfigWatcher to retrieve the git history of monitored files.
2. Applies diff_configs() (semantic differ) between consecutive versions of each file.
3. Classifies each change with ChangeClassifier (MAJOR/MINOR/PATCH).
4. Renders the changelog with ChangelogRenderer (Markdown or JSON).
5. Displays the output in the terminal and saves it to output_path.

Monitored files (by default):

• CLAUDE.md, .cursorrules, .cursor/rules
• AGENTS.md, .github/copilot-instructions.md, .github/agents/*.md
• .architect/config.yaml, architect.yaml

Example:

$ licit changelog

# Agent Config Changelog

> 3 change(s) detected across 2 file(s): **1** major, **1** minor, **1** patch

## .architect/config.yaml

- **[MAJOR]** Changed: model from claude-sonnet-4 to claude-opus-4 (`a1b2c3d4`) — 2026-03-12
- **[PATCH]** Changed: budget.max_cost_usd from 5.0 to 10.0 (`a1b2c3d4`) — 2026-03-12

## CLAUDE.md

- **[MINOR]** Changed: section:Rules from 5 lines to 8 lines (+3/-0) (`e5f6g7h8`) — 2026-03-11

  Changelog saved to .licit/changelog.md

JSON example:

$ licit changelog --format json --since 2026-03-01
# Generates JSON with a "changes" array and saves to .licit/changelog.md

Severity classification:

Supported diff formats:

For detailed documentation on the changelog system, see Changelog.

licit fria

Completes the Fundamental Rights Impact Assessment (EU AI Act Article 27).

Status: Functional (Phase 4 completed).

licit fria [--update]

Options:

What it does:

1. Detects the project and gathers available evidence.
2. Runs an interactive 5-step questionnaire (16 questions).
3. Auto-detects answers where possible (8 fields: system_purpose, ai_technology, models_used, human_review, guardrails, security_scanning, testing, audit_trail).
4. Saves data to .licit/fria-data.json and generates a report at .licit/fria-report.md.

5 questionnaire steps:

Auto-detection: For fields marked with auto_detect, licit attempts to infer the answer from the project configuration. If successful, it displays the detected value and asks whether to accept it.

Generated files:

• .licit/fria-data.json — Raw assessment data (JSON, reusable with --update)
• .licit/fria-report.md — Human-readable Markdown FRIA report

Example:

$ licit fria

============================================================
  FUNDAMENTAL RIGHTS IMPACT ASSESSMENT (FRIA)
  EU AI Act -- Article 27
============================================================

──────────────────────────────────────────────────
  Step 1: System Description
──────────────────────────────────────────────────

  [1.1] What is the primary purpose of this AI system?
  -> Auto-detected: AI-assisted code development using claude-code
    Accept this value? [Y/n]:

licit annex-iv

Generates the Annex IV Technical Documentation (EU AI Act).

Status: Functional (Phase 4 completed).

licit annex-iv [--organization NAME] [--product NAME]

Options:

What it does:

1. Detects the project and gathers all available evidence.
2. Auto-populates an Annex IV document with 6 sections from project metadata.
3. Generates recommendations for sections with missing evidence.
4. Writes the result to .licit/annex-iv.md.

6 auto-generated sections:

Example:

$ licit annex-iv --organization "ACME Corp" --product "WebApp"

  Annex IV documentation saved to: .licit/annex-iv.md

Generated file:

• .licit/annex-iv.md — Complete technical documentation in Markdown

licit report

Generates a unified compliance report.

Status: Functional (Phase 6). Evaluates EU AI Act + OWASP Agentic Top 10. Supports Markdown, JSON, and HTML.

licit report [--framework {eu-ai-act|owasp|all}] [--format {markdown|json|html}] [--output PATH]

Options:

Example:

$ licit report --framework eu-ai-act

  Compliance Summary
  ─────────────────────────────────────────────
  Project: my-app
  Generated: 2026-03-15 12:00 UTC

  eu-ai-act (2024/1689)
    [##..................] 9.1%
    1 compliant | 4 partial | 6 non-compliant

  ─────────────────────────────────────────────
  Overall: [##..................] 9.1%
  1/11 controls compliant

  Report saved to: .licit/reports/compliance-report.md

Output formats:

Generated files:

• .licit/reports/compliance-report.md (or .json/.html depending on --format)

licit gaps

Identifies compliance gaps with actionable recommendations.

Status: Functional (Phase 6). Shows gaps with suggested tools and effort level.

licit gaps [--framework {eu-ai-act|owasp|all}]

Options:

Example:

$ licit gaps --framework eu-ai-act

  10 compliance gap(s) found:

  1. [X] [ART-27-1] Fundamental Rights Impact Assessment (FRIA)
     Missing: Before putting an AI system into use, deployers shall
     carry out an assessment of the impact on fundamental rights.
     -> Run: licit fria -- to complete the FRIA
     Tools: licit fria

  2. [!] [ART-12-1] Record Keeping — Automatic Logging
     Incomplete: AI systems shall be designed with capabilities enabling
     automatic recording of events (logs) over the lifetime.
     -> Enable structured audit trail (architect reports or manual logging)
     Tools: licit trace, architect (audit log)

Gaps are sorted by severity ([X] non-compliant before [!] partial) and each one includes a description, recommendation, and suggested tools.

licit verify

Verifies compliance and returns an exit code for CI/CD.

Status: Functional (Phases 4-5). Evaluates EU AI Act (11 articles) and OWASP Agentic Top 10 (10 risks).

licit verify [--framework {eu-ai-act|owasp|all}]

Exit codes:

Usage in CI/CD (GitHub Actions):

- name: Compliance check
  run: licit verify
  # The pipeline fails if exit code != 0

Command Summary Table