Governance spans the path and the plane above it.
Local enforcement sits in the tool path. Trust roots, signed policy, identity, and attestation remain outside the agent-controlled session so it cannot rewrite its own rules.
Company
Autonomy should expand only as accountability gets stronger.
mistwire builds controls for organizations that want the leverage of coding agents without turning policy, identity, and evidence into optional afterthoughts.
Why mistwire exists
We govern agent behavior and outcomes with probabilistic and deterministic rules, both inside and outside the agent’s context.
So that the rules an agent is asked to follow and the rules it is unable to break come from the same signed source, because either kind alone leaves a gap that is unkind to systems and unkind to the user.
Inside and outside the agent’s context
Some rules live where the agent can read them: the signed skills and instructions it loads before work, which shape what it tries to do. Those are probabilistic, because a model follows what it reads most of the time. Some rules live where the agent cannot reach them: hooks that check each request before it runs, and a service that compares what is installed against what it signed. Those are deterministic. Governance is both, installed from one signed source.
The rules, not the machine
We never look at your hardware, your memory, or your operating system. We look at one folder we created, the files we put in it, and the requests an agent makes through the hooks in it. That is the whole surface, and it is small on purpose.
Today, and next
Today the alpha installs signed policy and hooks on your machine that check each request before it runs. That already means the rules are yours to read, signed, versioned, and verifiable by anyone you choose to show them to. Next, more of the deterministic layer moves outside the agent’s harness entirely, so there is nothing to step around because the route was never there. Next
The witness
Anyone running software on hardware you do not control can lie about it. Our answer is to make a lie visible rather than to pretend it cannot happen: an independent witness, outside the agent’s own trust domain, so that a false claim is caught rather than merely blamed. That witness is planned, not shipped, and we say so. Next
A real read
We show a real read even when it is unflattering, and nothing at all rather than a number we invented. We ship rules that actually fire, and we publish what is not built as carefully as what is. The proof has to survive someone who has no reason to trust us, because that person is your auditor, and they are the only audience that matters.
Kind to people and to systems
We build all of this at the least cost to the people using it and the least footprint on the systems running it. It is kind to digital systems to have good ways of managing digital infrastructure, and anything less is debt handed to whoever comes next. The same instinct shapes what we keep: as little of your data as the job allows, because that is how trust is won.
The order of things
Governance for coding and agent sessions is the product today. Foil, which moves the network boundary itself, comes second. After that comes the runtime the agent lives in, which means rethinking how a system decides what to trust without breaking how the internet talks. We say the order out loud so you can hold us to it.
Human authority is structural.
Safe cases can move quickly, dangerous cases stop, and uncertain requests stay with the host and human.
Foil extends the boundary later.
The current product governs coding and agent sessions. Foil will bring moving-boundary protection to tailored enterprise environments.
Our values
Values
mistwire.io was built around a set of convictions about how software should work, how companies should operate, and what it means to build something worth trusting. They shaped the architecture and governance from the beginning.
01Kindness
mistwire.io believes in kindness. We build software and automations that are kind to the users and kind to the systems they run on. A system so computationally expensive to operate that it defeats its own purpose is unfair to users and unkind to the architecture supporting it.
We focus on lean, lightweight, modular, and useful features—tools and systems that feel natural and easy to use. Complexity that serves the user is welcome. Complexity that serves the product roadmap at the user’s expense is not.
02Human primacy
Autonomous AI agents that can modify live infrastructure are powerful—and that power requires a clear hierarchy. At mistwire.io, humans are always at the top of it. Our agents reason, decide, and act, but they do so within boundaries that human operators set, and they cannot override the consent model that governs them.
Automation should extend human capability, not replace human judgment. Human primacy is an architectural constraint, not something that should be accidentally configured away.
03Transparency and explainability
We believe that a security system you cannot interrogate is a security system you cannot trust. Explainability is a design constraint because opacity is a form of disrespect—toward users, regulators, and the people whose systems we are trusted to protect.
Governance decisions should be explained in plain language, with the evidence and limitations available for inspection. Claims must stay bounded by what the current system actually records and verifies.
04Mutual benefit over extraction
We don’t believe a business relationship should be structured so that one party wins by making the other party more dependent, more vulnerable, or less informed. mistwire.io builds products and relationships where the value is real, the terms are honest, and the customer’s success is the measure of ours.
We don’t gate capabilities that make customers safer behind premium tiers. We don’t build lock-in as a retention strategy. We earn continued trust by continuing to deserve it.
05Shared defense
Our direction for shared defense is that when one mistwire customer encounters a new threat, other participating customers can benefit from that knowledge without exposing the originating customer’s private material. A safer network for one should mean a safer network for all.
Customers who prefer to operate independently must be able to opt out. We respect that choice. The broader federated-threat-intelligence model belongs to the Foil enterprise direction and is not a claim about the current Governance alpha.
These values are mistwire’s non-negotiables. They apply to the architecture, the governance documents, and the way we treat every person who works with us or depends on what we build.
The team
Security, governance, and product design around one table.
We bring security engineering, enterprise operations, and human-centered product design together to make governed agent work practical.
Paul DeJarnette
Founder & Chief Executive Officer
Technology and security executive with 20 years spanning agentic AI security, military cyber operations, cloud transformation, and enterprise infrastructure leadership. Paul founded mistwire to make autonomous action governable from the session to the enterprise boundary.
Career history & credentials
Career history
- mistwire.io · Founder & CEODesigned and built the moving-target-defense proof of concept, agent orchestration, and FROST wire protocol; led product strategy, business formation, and go-to-market.
- Starbucks · Manager, InfrastructureLed infrastructure work across 11 manufacturing and distribution sites and built the first fleet-wide QBR platform.
- SAIC · Project Manager, USAF EITaaSLed a $10M cloud and ServiceNow migration across multiple workstreams serving more than 60,000 users.
- Amazon · Program Manager, Brand ProtectionLed a globally distributed product team across nine swimlanes and scaled operations from 45 to 99 personnel ahead of plan.
- U.S. Air Force · Cyber operations, network operations, and program managementEleven years spanning offensive cyber operations, an enterprise network serving 8,500 customers, and an $18M Network Attack System across 11 areas.
Credentials
- B.S., Computer Science & Network Security — Park University
- Military Cyber 200 / Joint Intermediate Cyber Operations — AFIT
- PMP, CompTIA Security+, ITIL v4
“The attacker needs to be right once. The defender needs to be right every time. mistwire inverts that asymmetry.”
Quinton Taylor-Garcia
Co-Founder & Chief Operating Officer
Builder of autonomous defense systems with 20+ years across defense cyber operations, Zero Trust architecture, cloud security, and enterprise risk leadership. Quinton leads operational strategy, product execution, partnerships, and commercialization.
Career history & credentials
Career history
- mistwire.io · Co-Founder & COOLeads operational strategy, product execution, go-to-market, and commercialization for Governance, FROST, and the broader autonomous-defense direction.
- Omni Federal · Cybersecurity Executive / Product OwnerLeads cyber risk, Zero Trust, DevSecOps, ATO acceleration, and enterprise governance initiatives supporting defense missions.
- Bamboo Village Books · Founder & CEOBuilt a community-driven publishing platform supporting independent authors and veterans.
- The Buffalo Group · Cyber Risk & Accreditation LeadLed RMF accreditation and ATO processes for mission systems and managed enterprise security posture and compliance.
- U.S. Army · Cyber operations & IT leadershipEleven years supporting mission-critical operations for more than 25,000 users, managing a $1.3M budget, and training more than 200 personnel.
Credentials
- MBA, Project Management — University of Phoenix
- B.S., Business Management — University of Phoenix
- CISM, PMP, PMI-RMP, SecurityX (CASP+), SC-100, CSPO
“Autonomous defense begins when security stops reacting to threats and starts shaping the conditions attackers depend on.”
Jiyae Choi
Product Manager — Governance (Interim)
Product leader focused on AI governance, digital identity, and trustworthy human-agent systems. Jiyae brings human-centered design, responsible AI, and regulatory fluency to making governed agent experiences clear, usable, and enterprise-ready.
Career history & credentials
Selected experience
- mistwire.io · Product Manager, GovernanceShapes the product experience, trust model, and human-facing workflows for governed agent sessions.
- KFTC · Product ManagerOwned an OAuth-based identity platform across 60+ financial institutions, scaling to approximately 500,000 daily active users; launched cross-industry biometric check-in across 14 airports and 12 banks.
- University of Washington · Product and accessibility workLed an AI-assisted research-budgeting workflow redesign and remediated digital resources against WCAG 2.1 accessibility standards.
- PKNIC · Product Manager + Product DesignDesigned a trust architecture and accountability flows for a mentoring community, pairing conduct standards with human-reviewed enforcement.
- Seeds of Empowerment · Product ManagerDefined scaling strategy for a 600,000-user education product and unified fragmented design systems to reduce development complexity.
Standards & education
- Contributor to ISO 19092:2023 biometric security standards
- M.S., Human-Centered Design & Engineering — University of Washington
- Bachelor’s studies in Business Administration and French Language & Literature — Korea University
- Experience across GDPR, eIDAS, PSD2, responsible AI, accessibility, and adversarial-attack mitigation