Programmable compliance infrastructure for agentic AI.
Redact, simulate, and audit sensitive data — with policies, trustScore, and role-based logic.
Built for teams using LLMs, agents, and AI-native workflows.
Read the Docs marvisvault.com Apply for Vault Plus
- Redaction Engine — Mask sensitive fields using role + trustScore
- Policy Language — Declarative conditions with
&&
,||
, and field logic - Simulation CLI — See what would be masked before sending to the model
- Audit Logging — Structured JSONL logs for every mask/unmask decision
- Python SDK — Use Vault in agents, pipelines, or AI assistants
- Policy Templates — GDPR, PII, finance, healthcare — ready to drop in
git clone https://github.com/abbybiswas/marvis-vault-oss.git
cd marvis-vault-oss
pip install -e .
See Local Setup Guide for detailed environment setup instructions
# Simulate policy evaluation
vault simulate --agent examples/agent.json --policy policies/pii-basic.json
# Redact sensitive data
vault redact --agent examples/agent.json --data examples/data-pii.json --policy policies/gdpr-lite.json
# View audit logs
vault audit --log logs/audit.log --format table
from vault.sdk import redact
from vault.engine.policy_parser import load_policy
# Load policy and agent context
policy = load_policy("policies/healthcare.json")
agent = {"role": "analyst", "trustScore": 75}
# Redact sensitive data
result = redact(
content='{"name": "John Doe", "ssn": "123-45-6789"}',
policy=policy,
agent_context=agent
)
print(result.content) # {"name": "John Doe", "ssn": "[REDACTED]"}
marvis-vault-oss/
├── vault/ # Core library code
│ ├── cli/ # CLI commands
│ ├── engine/ # Policy engine
│ ├── sdk/ # Python SDK
│ └── utils/ # Security utilities
├── examples/ # Simple examples to get started
├── policies/ # Pre-built policy templates
├── tests/ # Test suite
├── dev/ # Development resources
│ ├── test-data/ # Comprehensive test data
│ ├── scripts/ # Testing & demo scripts
│ └── instructions/ # Development guides
└── docs/ # Documentation
Ready-to-use templates in policies/
:
- pii-basic.json — Basic PII protection (name, email, SSN)
- healthcare.json — HIPAA-compliant medical records
- finance-trust.json — Financial data with trust-based access
- gdpr-lite.json — GDPR-inspired data protection
For production-grade testing:
# Run comprehensive API tests
python dev/scripts/api_test_runner.py
# Test security hardening
python -m pytest tests/security/ -v
# Performance benchmarks
python dev/scripts/benchmark.py
See dev/test-data/
for:
- Production agent profiles
- Industry-specific test data (healthcare, financial, HR)
- Security attack vectors
- Complex nested structures
Feature | OSS | Vault Plus |
---|---|---|
Policy engine (mask, simulate) | [x] | [x] |
Full CLI + Python SDK | [x] | [x] |
Hosted API (FastAPI) | [ ] | [x] |
Secure role-based unmasking | [ ] | [x] |
Interactive TUI playground | [ ] | [x] |
Telemetry + usage analytics | [ ] | [x] |
Policy Marketplace (Q3 2024) | [ ] | [x] |
Vault Plus is free during early access — Apply here
We welcome contributions! See CONTRIBUTING.md for guidelines.
This project includes security hardening against:
- SQL/NoSQL injection
- XSS attacks
- Command injection
- Path traversal
- DoS attacks
- Type confusion
- Special value attacks (Infinity, NaN)
See SECURITY.md for details.
MIT License - see LICENSE.md
- GitHub Issues: Report bugs or request features
- Documentation: docs/
- Community: Coming soon
Built with love by the Marvis team