Enclawed, Inc.
Enclawed builds deterministic security and compliance infrastructure for AI agents, including deny-by-default controls, tamper-evident audit, data-loss prevention, prompt-injection defenses, and machine-readable audit evidence. Its stated goal is to make agents deployable and defensible in regulated and high-assurance environments.
At a Glance
- Individuals running assistants on their own machines
- Teams and businesses deploying agents
- Regulated enterprises
- Financial services
- +6 more
AI Tools by Enclawed, Inc.
(1)
omcp
Open AI Model Context Protocol
Discussions
No discussions yet
Be the first to start a discussion about Enclawed, Inc.
Latest News
omcp versions the attestation document as 1.0 and publishes the ATSA 1.0 extension
omcp clarifies its independent-fork status, Enclawed trademark ownership, upstream provenance, and non-affiliation
omcp publishes a generated versioned specification site in HTML and PDF
Enclawed publishes a snapshot of the enclawed-oss open-source hardened agent runtime
Products & Services
Proprietary production build of Enclawed's agent-security framework, distributed as signed binaries with vendor support and custom compliance/accreditation assistance under individually negotiated business agreements.
MIT-licensed open-source core and hardened OpenClaw fork for single-user AI assistants in high-trust environments. It provides classification-aware access control, DLP scanning and redaction, deny-by-default egress, signed-module loading, MCP attestation, prompt-injection defenses, transaction rollback, encryption, and tamper-evident audit.
Patent-pending sealed, tamper-evident USB hardware trust anchor that gives an agent runtime a unique hardware identity, authorizes and records actions, and fails secure when removed or swapped. It is designed to FIPS 140-3 Level 2 requirements and requires FCC certification before finished-product shipment.
Community-driven, drop-in-compatible hard fork of the Model Context Protocol: an open protocol, specification, schema, documentation, proposals, and extensions for connecting AI models to tools and data.
Market Position
Enclawed positions itself as a complete, compliance-oriented agent security layer rather than a content-only guardrail. Its comparison names vanilla OpenClaw and LangChain runtimes, NeMo Guardrails and Guardrails AI, and prompt-firewall products Lakera and Rebuff as categories that lack some combination of tool gating, deny-by-default egress/DLP, tamper-evident audit, boot-time integrity accreditation, multi-level access control, or machine-readable compliance evidence. omcp is positioned as a compatible, community-driven alternative specification and fork in the MCP ecosystem.
Leadership
Founders
Alfredo Metere
Founder and CEO of Enclawed. He authors the company's research program, including papers on hardened agent gateways, agent-skill verification, covert-channel egress monitoring, and attested MCP tool-server admission.
Executive Team
Alfredo Metere
Founder & CEO
Leads Enclawed and authors its six-paper research program on verifiable and containable agentic AI, including the foundational hardening framework and hardware-root-of-trust work.
Founding Story
Enclawed was started around the premise that autonomous AI agents can take consequential actions while conventional agent runtimes neither prevent hijacking nor provide trustworthy proof of what happened. Its initial vision was a sector-neutral hardening framework that gates the agent loop, uses deny-by-default access and egress controls, signs and admits modules, and maintains a tamper-evident audit trail; the company is extending that software foundation with an optional hardware root of trust.
Business Model
Revenue Model
The open-source enclawed-oss core is free under the MIT License. Enclawed monetizes the proprietary enclawed-enclaved build, signed binaries, vendor support, hardening reviews, dedicated engineering, OSCAL-emission help, and accreditation work through paid business-to-business subscriptions and custom engagements governed by individually negotiated order forms or MSAs.
Target Markets
- Individuals running assistants on their own machines
- Teams and businesses deploying agents
- Regulated enterprises
- Financial services
- Healthcare
- Defense and government
- Single-user AI assistants with access to email, files, and terminals
- Enterprise AI agents handling client data, production systems, or money
- Regulated healthcare workloads handling PHI
- Financial-services workloads handling material non-public information
- Defense contractors and government enclaves subject to CMMC, NIST 800-171, NIST 800-53, or FedRAMP
- Regulated R&D involving trade secrets or ITAR/EAR-controlled material