Platform Status: Production Ready

The Avarion Governance Platform

The first governance platform built for the age of AI coding agents. Observe, control, and audit everything Copilot, Claude, Cursor, and Devin generate in your codebase.

Request Demo See How It Works

Progressive Governance Adoption

Start with visibility, progress to control. Three operational modes that match your team's maturity.

Observe Mode

Passive monitoring. See all AI-generated code across your repositories without blocking anything.

Real-time AI activity dashboard
Agent attribution (which AI wrote it)
Historical analytics & trends

Audit Mode

Active tracking with provenance. Every AI-generated line gets a cryptographic audit trail.

ProofBundle generation
Compliance report exports
Merkle-hashed provenance

Enforce Mode

Active blocking. Permission schema enforcement prevents unauthorized AI code from entering your codebase.

Pre-commit hooks
Permission schema rules
Automatic rollback capability

Core Feature

The Permission Schema

Define exactly which files, directories, and operations each AI agent can touch. Version-controlled, auditable, and enforceable at the pre-commit level.

Path-Based Rules

Define allow/deny rules per file path or glob pattern

Agent-Specific Policies

Different rules for Copilot vs Claude vs Cursor

Operation Controls

Allow create but deny delete, or vice versa

.avarion/permissions.yaml

version: "1.0"
default_policy: deny

agents:
  copilot:
    allow:
      - path: "src/utils/**"
        operations: [create, modify]
      - path: "tests/**"
        operations: [create, modify, delete]
    deny:
      - path: "src/auth/**"
        reason: "Authentication code requires human review"
      - path: "src/payments/**"
        reason: "Financial logic is restricted"

  claude:
    allow:
      - path: "docs/**"
        operations: [create, modify]
      - path: "src/components/**"
        operations: [create, modify]

  cursor:
    inherit: copilot
    deny:
      - path: "infrastructure/**"
        reason: "Infrastructure changes need approval"

Visual Policy Map:

src/utils/AI Allowed

tests/AI Allowed

src/auth/Human Only

src/payments/Human Only

infrastructure/Approval Required

Dual-Track Architecture

Avarion operates on two parallel tracks to ensure complete coverage of your AI development lifecycle.

Track 1: CLI Pipeline

For developers and agents. The avarion CLI enforces the "Hermetic Generation" protocol locally.

Frozen Snapshots: Code is generated solely from audited contexts.
Merkle Hashing: Every file change is cryptographically signed.
Pre-commit Hooks: Enforced permission checks before code enters git.

Track 2: Control Plane

For security teams and auditors. A centralized SaaS layer that provides visibility and governance over all AI activity.

Shadow Scanning: Auto-detects AI code in repos without config.
Provenance Tracking: Full lineage history for every commit.
Blast Radius Analysis: Visualize impact before merging.

Risk Assessment

Blast Radius Analysis

Before AI-generated code merges, visualize its impact. See which modules depend on the changed files and assess the risk of the change.

Dependency Graph Visualization
See how changes ripple through your codebase
Risk Scoring
Automatic risk assessment based on affected modules
Pre-Merge Approval Workflows
High-risk changes require additional review

ML Detection

AI Code Detection Engine

Our Contrastive CodeBERT model achieves 95.6% F1 Score in detecting AI-generated code. Multi-signal detection combines five independent methods for maximum accuracy.

95.6%

F1 Score

<2%

False Positive Rate

Read the Full Technical Report

Detection Signals:

ML CodeBERT

98%

Annotation Detection

95%

Pattern Detection

65%

Timing Heuristics

50%

Git Metadata

40%

Weighted signal aggregation for final confidence score

Ready to govern your AI agents?

The EU AI Act enforcement begins August 2026. Start building your audit trail today.

Request Demo View Pricing