Jozu
Jozu secures and governs AI artifacts from supply chain through runtime, providing an on-prem AI model registry and zero-trust runtime for enterprises and government agencies deploying models, AI agents, and MCP servers.
- Company typePrivate
- Founded2023
- HeadquartersNew York, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Jozu does
Jozu (operating as Akara Technologies Inc., a Delaware C-corporation headquartered in New York) secures and governs AI artifacts from supply chain through runtime for organizations deploying models, AI agents, and MCP servers. Its core platform, Jozu Hub, is an on-prem AI model registry and secure ML deployment system that installs behind the firewall via Helm chart in approximately one hour and stores, scans, signs, and deploys AI assets as OCI-compliant ModelKits with cryptographic provenance, five-layer AI-specific threat scanning, SPDX 3 SBOMs, and tamper-evident audit logs. Jozu Agent Guard extends the platform into runtime with a zero-trust AI runtime that enforces tool-level policy via a two-layer defense (self-protection CEL policies plus a fail-closed hook), preventing AI agents from disabling their own guardrails or audit trails.
The technology foundation rests on open standards: Jozu's CTO authored the ModelPack CNCF specification, and the company created KitOps and donated it to the CNCF Sandbox, giving the platform native compatibility with OCI registries (JFrog Artifactory, Harbor, Sonatype Nexus, Amazon ECR, GitLab Registry, Docker Hub), major Kubernetes distributions (EKS, AKS, GKE, OpenShift, Tanzu, Rancher, bare metal), and CI/CD systems (Jenkins, GitHub Actions, GitLab CI). KitOps has accumulated 235,000+ downloads and is supported by Red Hat, PayPal, ANT Group, and ByteDance. An MCP Registry API exposes centrally curated, security-scanned MCP servers and agent skills to IDEs such as VS Code, Cursor, and Claude Desktop, while the open-source PromptKit adds auditable prompt management.
Jozu generates revenue through five streams: Jozu Hub Enterprise (custom-priced self-hosted subscription with custom scanning, signed audit logs, and unlimited repositories), Jozu Hub SaaS (hosted subscription with SAML/SSO/SCIM), Sandbox (freemium with unlimited public repos and 250 GB storage as a top-of-funnel), Support Services for KitOps/ModelPack (managed-services subscription with 24/7 priority, 4-hour SLA, and quarterly health checks), and Professional Services for paid POC engagements (default 14-day). Distribution runs through direct enterprise field sales, on-demand demos, a free-trial/POC funnel, and two U.S. government marketplace listings (CDAO Tradewinds and Air Force Platform One). Named reference customers include DSV (logistics), Pacific Northwest National Laboratory, Vantedge, DVS, and a German government-focused IT services provider.
Jozu firmographics
Firmographics- Name
- Jozu
- Legal name
- Akara Technologies Inc.
- Website
- https://jozu.com
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Jozu secures and governs AI artifacts from supply chain through runtime, providing an on-prem AI model registry and zero-trust runtime for enterprises and government agencies deploying models, AI agents, and MCP servers.
- Ownership category
- akta.pro rank
Jozu industry classification
Industry- Product category
- AI Security and Governance Platform (MLOps / AI Supply Chain)
- NAICS
- Custom Computer Programming Services (541511), Computer Systems Design and Related Services (5415)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Programming Services (7371)
- akta.pro primary industry
- Model Deployment, Serving & Inference Platforms (HDAAABAF)
- akta.pro secondary industry
- Model Governance, Approval Workflows & Auditability (HDAAABAJ)
Keywords
Where Jozu is headquartered
LocationHeadquarters
- HQ city
- New York
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Jozu business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Marketing or Sales, Operations, Infrastructure
Revenue model
- Jozu Hub Enterprise (Self-hosted On-Prem / Private Cloud): Recurring enterprise license for self-hosted Jozu Hub deployed on the customer's Kubernetes infrastructure; includes custom security scanning & policies, signed centralized audit logs, unlimited repositories/storage, integration with customer's own registry and RBAC, and email/Slack support.
- Jozu Hub SaaS (Jozu Hosted): Hosted SaaS deployment of Jozu Hub with enterprise authentication (SAML, SSO, SCIM), API and webhook access, and managed infrastructure.
- Sandbox (Freemium): Free tier for open-source and personal projects: unlimited public repositories, 1 private repository, Docker/Kubernetes usage, Hugging Face import, 250 GB storage, community Discord support; serves as top-of-funnel for enterprise conversion.
- Support Services for KitOps and ModelPack: Production support contracts for open-source KitOps/ModelPack users including 24/7 priority support, 4-hour response SLA, hot fixes and emergency patches, quarterly health checks, Governance Maturity Report, and Model Risk Audit.
- Professional Services / Proof-of-Concept Engagement: Paid POC evaluations (default 14-day period) with best-effort support; conversion path from POC to Master Software License Agreement for production deployments.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Annual | Sandbox - Free for open-source and personal projects |
| Subscription | Annual | Enterprise - Self-hosted on-prem or private cloud |
| Subscription | Annual | Support Services for KitOps and ModelPack |
| Subscription | Annual | SaaS (Jozu Hosted) |
Go-to-market motion1 record
Distribution channels7 records
Marketing channels9 records
Jozu product offering
Product offeringCore offering
Jozu sells an on-prem AI model registry and secure ML deployment platform (Jozu Hub) that packages AI/ML models, datasets, agents, MCP servers, and policies as OCI-compliant, tamper-proof ModelKits with five-layer security scanning, cryptographic signing, SPDX 3 SBOMs, and policy-gated deployment. Complementing the registry, Jozu Agent Guard enforces zero-trust runtime governance for AI agents and MCP servers via tool-level policies with fail-closed behavior. The portfolio is anchored by the open-source KitOps CLI/SDK and the ModelPack CNCF specification that Jozu created and contributes to.
Product overview
Jozu delivers a layered AI security and governance platform built on OCI standards, consisting of a core supply-chain platform (Jozu Hub) paired with a runtime enforcement product (Jozu Agent Guard) and supported by a portfolio of open-source projects and standards. Jozu Hub is the on-prem AI model registry and secure ML deployment platform that stores, scans, signs, and deploys AI artifacts as tamper-proof ModelKits, with five-layer threat scanning, policy gating, SPDX 3 SBOMs, and tamper-evident audit trails; it also hosts the Jozu MCP Registry that VS Code, Cursor, and Claude Desktop connect to directly. Jozu Agent Guard complements Hub by providing zero-trust runtime governance for AI agents and MCP servers, enforcing tool-level policy with fail-closed behavior so an agent cannot disable its own guardrails. Underpinning both is KitOps, the CNCF Sandbox open-source CLI and PyKitOps Python SDK project that creates the OCI-compliant ModelKits Hub consumes; KitOps implements ModelPack, the CNCF specification for vendor-neutral AI/ML packaging authored by Jozu's CTO and contributed to by Red Hat, PayPal, ANT Group, and ByteDance. Jozu's open-source PromptKit adds auditable prompt management, while the Jozu Rapid Inference Containers (RIC) layer and Jozu Hardened ModelKits offerings provide pre-built deployment acceleration and production-ready inference containers for major model families including Llama, Mistral, Phi, and Qwen.
Differentiator
Problem solved
Functional benefit
Brands
- Jozu Hub: Jozu's on-prem AI model registry and secure ML deployment platform for enterprise AI/ML workflows.
- Jozu Agent Guard
- KitOps
- ModelPack
- PromptKit
Products and services
- Jozu Hub On-prem AI model registry and secure ML deployment platform that installs behind the customer's firewall via Helm chart in approximately one hour. Stores, scans, signs, and deploys AI models, agents, MCP servers, datasets, and policies as OCI-compliant ModelKits with five-layer threat scanning, tamper-proof packaging, SPDX 3 SBOMs, policy gating, and tamper-evident audit trails. Exposes an MCP Registry API for VS Code, Cursor, and Claude Desktop. Available as Enterprise self-hosted on-prem/private cloud, Jozu Hosted SaaS, and a free Sandbox tier.
- Jozu Agent Guard Zero-trust AI runtime that enforces tool-level policy on AI agents and MCP servers at execution time. Uses two-layer defense: self-protection CEL policies that block the agent from tampering with its own enforcement infrastructure, and a fail-closed hook that blocks all tool calls when the policy server is unreachable. Supports hypervisor isolation (micro-VM/Kata), human-in-the-loop approvals for high-risk actions, and tamper-evident cryptographically chained audit logs that sync to Jozu Hub. Targets enterprise security teams deploying AI coding agents such as Claude Code, Cursor, GitHub Copilot, and Gemini CLI.
- KitOps Open-source CLI and PyKitOps Python SDK project (CNCF Sandbox) that packages AI/ML models, datasets, code, configurations, MCP servers, agent skills, and documentation into OCI-compliant ModelKits. Includes build/push/pull operations, Hugging Face import with auto-scanning, and a dedicated GitHub Action for CI/CD pipelines. Reference implementation of the ModelPack specification. KitOps has reached 235,000+ downloads and is the entry point into the Jozu enterprise offering.
- PromptKit Local-first, open-source tool that provides auditable and production-safe prompt management. Uses policy-controlled workflows and verified prompt artifacts to prevent prompt injection incidents, treating prompts as first-class, signed release artifacts (introduced in KitOps v1.11 to close the supply-chain gap between behavior changes and governed deployments).
- Jozu Hardened ModelKits Pre-built container images produced by Jozu that bundle a specific model version, an inference runtime, and a REST API for serving the model. Each Hardened ModelKit ships with a software build-of-materials (SPDX 3) and a signed provenance attestation so users know exactly what was included and who built it. Designed to be used as-is or as a base for in-house fine-tuning and RAG pipelines, accelerating time-to-production for major open-weight model families.
- Support Services for KitOps and ModelPack Paid production-support contracts for open-source KitOps/ModelPack users, including 24/7 priority support, 4-hour response SLA, hot fixes and emergency patches, quarterly health checks, a Governance Maturity Report, and a Model Risk Audit. Custom-priced annually, this offering monetizes the open-source install base and serves enterprise customers who run KitOps in production.
- Professional Services / Proof-of-Concept Engagement Paid proof-of-concept engagements (default 14-day period) with best-effort support and engineer-led walkthroughs, serving as the conversion path from sandbox evaluation to a Master Software License Agreement for production Jozu Hub deployments.
Quantifiable outcome
- 10x faster model deployments vs. standard container builds (Llama 3.2 8B in 42.8 sec vs. 425.4 sec on Amazon EKS)
- +5 more outcomes
Companies that use Jozu
Customer profileNamed customers9 records
Segments5 records
Ideal customer profiles5 records
Jozu technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration25 records
AI capability6 records
Feature8 records
Jozu partnerships and signals
Strategic signalPartnerships
20 partnerships are on record, tiered flagship and core.
- Cloud Native Computing Foundation (CNCF)flagshipJozu is a CNCF Silver Member, created KitOps and donated it to CNCF Sandbox, and Jozu's CTO authored the ModelPack specification hosted at CNCF. KitOps 1.0 (Jan 2025) was submitted to CNCF Sandbox for vendor-neutral governance.
- PayPalcoreBacks KitOps and ModelPack open-source projects; contributes to the ModelPack specification alongside Red Hat, ByteDance, and ANT Group.
- ByteDancecoreSupports KitOps and ModelPack open-source projects; contributes to the ModelPack specification.
- ANT GroupcoreSupports KitOps and ModelPack open-source projects; contributes to the ModelPack specification.
- Red HatcoreSupports KitOps and ModelPack open-source projects; contributes to the ModelPack specification; Jozu lists Red Hat OpenShift as a supported Kubernetes distribution.
- U.S. Department of War CDAO (Tradewinds Solutions Marketplace)flagshipJozu achieved Awardable status in the Department of War Chief Digital and Artificial Intelligence Office's Tradewinds Solutions Marketplace, enabling rapid procurement of AI security solutions for mission-critical DoW systems.
- U.S. Air Force Platform One (P1 Solutions Marketplace)flagshipJozu achieved Awardable status in the U.S. Air Force Platform One (P1) Solutions Marketplace, enabling rapid acquisition by U.S. Department of Defense customers for AI assurance across connected, on-prem, and air-gapped environments.
- MLflow (Databricks)coreDocumented integration between Jozu/KitOps and MLflow: PyKitOps SDK packages MLflow-tracked models into ModelKits automatically; many teams use both MLflow for experiment tracking and Jozu for production governance.
- Docker (Docker Hub / Docker Model Runner)coreModelKits are OCI-compliant artifacts that work alongside Docker Hub; Jozu explicitly positions Jozu Hub as the AI model layer complementing Docker Hub for container images and uses the same OCI standard as Docker Model Runner.
- Weights & BiasescoreDocumented integration: W&B Artifacts can be packaged into ModelKits via KitOps for governance, security scanning, and deployment through Jozu Hub; complementary tools for development vs. production lifecycle.
- JFrog (Artifactory)coreJozu Hub integrates with JFrog Artifactory as a supported OCI registry for storing and serving ModelKits alongside other OCI artifacts.
- HarborcoreJozu Hub integrates with Harbor as a supported OCI registry; Harbor is positioned as a customer-deployable registry option alongside ECR, GCR, and Artifactory.
- GitLab (Registry & CI)coreJozu integrates with GitLab Container Registry and GitLab CI/CD pipelines for ModelKit storage, scanning, and deployment.
- GitHub ActionscoreJozu integrates with GitHub Actions pipelines; KitOps provides a dedicated GitHub Action for using Kit CLI in MLOps pipelines.
- JenkinscoreJozu integrates with Jenkins CI/CD pipelines for ModelKit packaging, validation, and deployment workflows.
- Amazon Web Services (EKS)coreJozu supports Amazon EKS for Kubernetes deployment; Amazon ECR is a supported OCI registry; AWS SageMaker-trained models can be packaged via KitOps CLI and deployed to EKS via Jozu.
- Microsoft Azure (AKS)coreJozu supports Azure AKS for Kubernetes deployment; AKS is listed alongside EKS, GKE, OpenShift, Tanzu, Rancher, and bare-metal Kubernetes as supported deployment targets.
- Google Cloud (GKE)coreJozu supports Google GKE for Kubernetes deployment; GCR is a supported OCI registry.
- Hugging FacecoreJozu Hub supports one-click import from Hugging Face with automatic AI-specific security scanning; KitOps 1.0 added Hugging Face import capabilities and automatic Kitfile generation.
- Linux Foundation (governance of OCI standard)flagshipJozu's products are built on the OCI Image Manifest Specification, governed by the Linux Foundation; KitOps and ModelPack are positioned as openly governed standards to align with this.
Scale indicators10 records
Recent moves6 records
Expansion highlights6 records
Jozu competitors and assessment
Company assessmentDirect peers
- Seldon: Open-source and enterprise MLOps platform for model deployment, monitoring, and governance on Kubernetes. Overlaps with Jozu on Kubernetes-native model packaging, deployment, and audit/observability for production ML workloads.
- MLflow (Databricks): Open-source ML lifecycle platform for experiment tracking, model registry, and deployment. Complementary to Jozu (documented PyKitOps integration) but also a direct competitor for production model packaging and registry, especially within the Databricks ecosystem.
- Weights & Biases: Provides a model registry and MLOps platform widely used by ML teams for experiment tracking, artifact management, and model governance. Directly comparable to Jozu Hub for production model registry and packaging, with documented integration between W&B Artifacts and KitOps/Jozu.
- Protect AI: AI/ML security platform focused on model scanning, supply chain, and LLM/agent security. Most directly comparable competitor on AI-specific threat scanning (backdoors, poisoning, serialization attacks) and supply chain attestation for AI artifacts.
- Wallaroo Labs: Enterprise platform for deploying, observing, and governing ML models at scale, including in air-gapped and edge environments. Comparable to Jozu Hub on enterprise model deployment and governance with similar on-prem and regulated-industry focus.
Emerging players
- Hugging Face: Operates the leading model hub with growing enterprise and security offerings. Competes on model distribution and import (Jozu supports one-click Hugging Face import) and increasingly on enterprise evaluation and governance features.
- Lakera: AI security company focused on prompt injection, LLM guardrails, and agent runtime protection. Overlaps with Jozu Agent Guard on runtime policy enforcement for AI applications and agents, with comparable positioning for enterprise security buyers.
Broad incumbents
- Docker: Operator of Docker Hub and Docker Model Runner, leveraging the same OCI standards Jozu builds on. Competes in container and now model packaging/distribution, with broader developer mindshare but no AI-specific scanning or supply-chain attestation focus.
- JFrog: Broad artifact management and software supply chain platform with native OCI registry (Artifactory) and security scanning. Competes with Jozu on supply chain integrity and SBOM generation at the artifact layer, but as part of a much wider DevOps portfolio.
- AWS SageMaker: Hyperscaler MLOps suite covering training, registry, deployment, and inference. Competes with Jozu on the cloud-native side, but lacks Jozu's on-prem/air-gap posture, AI-specific five-layer scanning, and OCI-portable ModelKit format.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Jozu social profiles
Digital presenceJozu compliance and trust
Trust signalCompliance8 records
Jozu financial estimates
Financial estimateRevenue estimate
Valuation estimate
Jozu leadership team
Management profileNumber of profiles
Profiles3 records
Jozu funding detail
Funding detailFunding overview
Funding rounds2 records
Investors6 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Jozu M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Jozu
What does Jozu do?
Jozu sells an on-prem AI model registry and secure ML deployment platform (Jozu Hub) that packages AI/ML models, datasets, agents, MCP servers, and policies as OCI-compliant, tamper-proof ModelKits with five-layer security scanning, cryptographic signing, SPDX 3 SBOMs, and policy-gated deployment. Complementing the registry, Jozu Agent Guard enforces zero-trust runtime governance for AI agents and MCP servers via tool-level policies with fail-closed behavior. The portfolio is anchored by the open-source KitOps CLI/SDK and the ModelPack CNCF specification that Jozu created and contributes to.
Is Jozu a public or private company?
Jozu is a private company. It is classified as venture growth investor backed and is currently operating.
When was Jozu founded?
Jozu was founded in 2023. It employs 11 to 50 people.
Where is Jozu based?
Jozu is headquartered in New York, United States, in the North America region.
How does Jozu make money?
Five revenue lines are on record. Jozu Hub Enterprise (Self-hosted On-Prem / Private Cloud) is the primary driver. The others are jozu Hub SaaS (Jozu Hosted), sandbox (Freemium), support Services for KitOps and ModelPack and professional Services / Proof-of-Concept Engagement.
Who are Jozu's main competitors?
Direct peers on record are Seldon, MLflow (Databricks), Weights & Biases, Protect AI and Wallaroo Labs. Emerging players are Hugging Face and Lakera. Broad incumbents are Docker, JFrog and AWS SageMaker.
Does Jozu have an API?
Yes. Jozu Hub offers REST API and full webhook access (Enterprise tier). Jozu also exposes a native MCP Registry API that AI development tools such as VS Code, Cursor, and Claude Desktop connect to directly for retrieving centrally curated MCP servers and agent skills. KitOps provides a Python SDK (PyKitOps) for programmatic ModelKit packaging and a Kit CLI command-line tool for build/push/pull operations. The MCP Registry API implements the MCP registry specification and serves as a central catalog for MCP servers, agent skills, and model artifacts. Developer documentation is at kitops.org/docs/overview.
What industry is Jozu in?
Jozu's product category is AI Security and Governance Platform (MLOps / AI Supply Chain). Its primary akta.pro industry code is HDAAABAF, Model Deployment, Serving & Inference Platforms, with a secondary code of HDAAABAJ, Model Governance, Approval Workflows & Auditability. Its NAICS code is 541511 and its SIC code is 7372.