CLIC by RegOps · Agentic compliance infrastructure for EU AI Act codebases

Classify. Implement. Ship.

CLIC is agentic compliance infrastructure for AI codebases. It detects AI systems in real repositories, classifies them under the EU AI Act, generates regulatory evidence and engineering tasks, and helps teams remediate the codebase.

Book a demo →

// phase 01 · classify · fastest path to AI Act classification

CLIC finds what the AI Act requires.

CLIC discovers AI system candidates in your repository, classifies them at system level under the EU AI Act, maps the obligations that apply, and turns them into concrete engineering tasks.

01scan/repositoryacme/lending-app

// 1,284 files · 2 candidates found

▾app/

·layout.tsx

·page.tsx

▾lib/

·utils.ts

·scoring.ts

·recommend.ts

▾models/

·credit-risk.pkl

·fraud-detect.onnx

·embeddings.bin

▾pipelines/

·training.py

▸tests/

▸node_modules/

·package.json

● credit-scoring [HIGH RISK · Annex III §5(b)]
● fraud-detection [LIMITED · Art. 50]
→ 2 candidates · 1 high-risk · 1 limited-risk

02$ clic check --repo . --json● ready

"status": "warn",
"candidateCount": 2,
"candidates": [

{"label": "credit-scoring","riskClass": "high" ●,"evidenceSummary": "Annex III §5(b) · Art. 9 · Art. 12","tasks": [{"title": "Add transparency notice", "priority": "high"}{"title": "Implement audit logging", "priority": "high"}{"title": "Risk management hooks", "priority": "medium"}]},

{"label": "fraud-detection","riskClass": "limited" ●,"evidenceSummary": "Art. 50 · transparency obligation","tasks": [{"title": "Add transparency notice", "priority": "medium"}]}

]

✓ 2 systems · 4 tasks

// or connect via MCP · zero config

// phase 02 · implement

CLIC Agent implements it.

The agentic layer turns EU AI Act obligations into implementation work, applies changes inside the codebase, and verifies whether the implementation satisfies the required evidence. Connect via MCP when you want it inside Claude Code, Cursor, or another compatible workflow.

01CLIC/scantasks.json

// 7 obligations · Annex III §5b

T-01

Add transparency notice

Art. 50(1) · user-facing
HIGH

T-02

Log inference decisions

Art. 12 · audit trail
HIGH

T-03

Risk mgmt hooks

Art. 9 · pre-deploy
MED

T-04

Data governance check

Art. 10 · training
MED

02CLIC Agent/agent● ready

analysing repository

[plan] 4 tasks · 12 files affected

[T-01] insert components/AIDisclosure.tsx
mount in app/layout.tsx ✓

[T-02] wrap lib/inference.ts
with audit logger · +34 lines ✓

[T-03] add CI gate
.github/workflows/risk.yml ✓

[evidence] writing .clic/
artefacts → conformity bundle

[mcp] git hook configured ✓

[mcp] tasks.json → agent context ✓

committing branch clic/eu-ai-act

~4m avg. runtime · MCP server included

03repo/pull request+ ready to merge

app/layout.tsx+8 −2

return (

<html>

lib/inference.ts+34 −6

return model.predict(input);

const out = await audit.wrap(

() => model.predict(input),

{ article: 12, system })

return out;

.clic/evidence.md+1 file

.clic/tasks.md+1 file

// 01

Any stack, zero config

Python, TypeScript, Go, Rust. CLIC reads your repository as-is. Regulated AI systems surface from what's actually in your code.

// 02

Fact-Linked Evidence

Every finding is cited to an AI Act article or recital, creating verifiable claims your engineers can act on.

// 03

CI/CD & MCP Native

Run headless in your CI/CD pipeline as a git hook, or connect via MCP server for interactive control. Classify on commit, configure hooks, pipe tasks into your issue tracker. Your workflow, your interface.

// classification outcomes

Certainty where possible.
Clarity where needed.

Where the AI Act is unambiguous, CLIC gives a definitive classification. Where one aspect of a system's context is unclear, CLIC returns needs_review with the exact question that determines the outcome: a deployment scope question, an intended-use ambiguity, a decision-chain dependency.

needs_review enables a precise classification. It surfaces the single factual detail that separates covered from not_covered, giving your legal or product team exactly what to confirm.

coveredAI Act obligations apply. Engineering tasks emitted.

not_coveredSystem outside AI Act scope. Reasoning on file.

needs_reviewOne open question. The exact detail that determines the classification is surfaced.

common questions

Can engineers use CLIC without legal expertise?

Yes. CLIC reads your codebase and outputs concrete engineering tasks. No legal background required. Your legal team reviews the findings.

How does CLIC fit into an existing pipeline?

CLIC runs via Git hooks or a single CLI command. No new infrastructure required.

Is CLIC just a rule pack or linter?

No. CLIC starts one layer earlier: AI system discovery and EU AI Act classification. Rule checks only make sense after the regulated system, risk class, obligations, and missing facts are understood.

// ready when you are

From AI codebase to audit-ready system.