From the Lab to the Market
MC Project Lab, founded by Maurizio Cirio, develops software to control AI agents, protect data, and make results verifiable.
Armor Suite, Privacy Suite, and the MLDR method are operational. TCT is under development in two editions. We are seeking investors and industrial partners to turn technical work into products distributed, supported, and adopted by enterprises.
Core Solutions
The portfolio comprises tools developed by the laboratory, with local processing where provided by the architecture.
For each solution, we distinguish the development stage, available features, external integrations, and verification scope. Test results refer to the versions and configurations documented in the relevant reports.
TCT · Therminal Control Tower
Coordinate AI agents, restrict their actions, and preserve their audit trails
When an AI agent can execute tools, access data, or trigger operations, the enterprise must know what permissions it holds, what it is doing, and when human intervention is required. Risks include malicious instructions, out-of-scope access, and unauthorized data transfers.
TCT brings together permission controls, event logging, and human approval gates. Development proceeds on two editions: TCT AOI, tailored for professional practices and SMEs, and TCT AI Security, dedicated to security controls on agents and models. Prompt firewall features aim to detect or constrain dangerous instructions. Their efficacy must be assessed against documented attacks, configurations, and limitations.
Modular architecture based on FastAPI and React. Security testing includes authorized, simulated attack scenarios conducted by the laboratory's Red Team. Local processing, external model connections, and deployment modes must be specified for the customer configuration.
Enterprise licensing, pilot projects, systems integration, dedicated on-premise or sovereign cloud deployments, and verification retainers. Pricing may account for managed agents, feature sets, and required support.
Armor Suite
Identify sensitive information and inspect AI system artifacts
Logs, emails, and files used in AI systems often contain credentials, personal information, or other elements requiring inspection before storage, sharing, or processing.
LogArmor identifies and sanitizes credentials and sensitive data covered by its rulesets. Replay preserves immutable activity traces with cryptographic mechanisms designed to verify integrity; this does not mean records can never be altered under all conditions. MailArmor parses email payloads against implemented checks. WeightArmor inspects AI model files against supported threat indicators; results do not constitute an exhaustive guarantee of model security.
The suite prioritizes local execution. Supported file formats, check coverage, external connections, and test results are specified per module. Local execution is an architectural feature; detection effectiveness requires targeted empirical validation.
Licenses and integration into software CI/CD pipelines, security monitoring systems (SIEM), and enterprise audit workflows, backed by configuration and maintenance services.
Privacy Suite
Inspect links and mitigate tracking vectors
Online tracking relies on cookies, link parameters, and browser or device characteristics. Different vectors require different controls: no single tool guarantees complete anonymity or absolute protection.
URL Unmasker reconstructs redirect hops and identifies marketing parameters; link resolution requires network queries. Cookie Tracker supports inspection of cookie attributes and compliance baselines; observations inform privacy audits without substituting for formal GDPR compliance certifications. Data Obfuscator intervenes on supported fingerprinting surfaces to reduce device recognizability; effectiveness and compatibility vary with browser engines and configurations.
The suite combines Python tooling with a lightweight browser extension. On-device processing is kept distinct from external network calls needed to query web pages or resolve links. Verifications document software versions, browser environments, test cases, and functional impacts on audited pages.
Free baseline version, professional feature upgrades, and enterprise workstation licenses, accompanied by configuration, support, and technical reports.
MLDR
Machine Learning Demonstrable ResultsAn AI answer can assert that a task is “working”, “resolved”, or “verified” without presenting sufficient evidence. Decision-makers must distinguish an outcome verified by empirical testing from a consulted source, an inference, or memory recall.
MLDR categorizes statements according to their provenance:
Measured: grounded in recorded empirical proof or direct observation.
Read: derived from an actual verified source.
Deduced: obtained through documented reasoning, with stated premises and constraints.
Memory: recalled without fresh validation of the source or outcome.
The objective is linking claims to verifiable evidence and flagging what demands further inspection.
Automated checking identifies verifiable elements in available materials. It does not by itself prove that a source is accurate or a test comprehensive. Secondary review by independent AI models is a separate phase: it does not replace technical proof and may involve external API calls depending on configuration. Final decisions always rest with the responsible human operator.
Methodology rollout across enterprise engineering workflows, board-level compliance reports, and ongoing evidence audits for generative AI assistants.
What We Seek
Industrial investors and commercial partners who can help prioritize go-to-market strategies, scale engineering and sales teams, and reach enterprise customers or new market verticals.
Collaboration models include licensing, commercial distribution, white-label product integration, co-development, and targeted pilot projects.
The structure of any equity or capital investment is defined independently based on project scope and corporate vehicle.
What We Bring to the Table
Working source code and live demonstrations of available versions, detailed test reports, known architectural boundaries, and technology roadmaps.
Under confidential briefings, we can review deep technical architectures, deployment configurations, verification logs, and enterprise adoption prerequisites.
Let's Start with a Concrete Use Case
Our initial discussions focus on identifying a specific solution, a customer pain point, and clear, measurable success criteria.
Maurizio Cirio · Founder, MC Project Lab