[Sandbox] agentregistry

[ilevine]

Project summary

agentregistry is a cloud-native registry for discovering, curating, and deploying MCP servers, agents, and skills across local development and Kubernetes environments.

Project description

As AI-powered development tools become core infrastructure, organizations face a challenge: these tools depend on MCP servers, agents, and skills, and there is not a standardized way to govern and control these agentic building blocks. Teams end up with scattered repositories, manual configuration, and inconsistent setups across local workstations and shared clusters.

Agentregistry brings governance and control to AI artifacts and infrastructure. It provides a centralized, cloud-native registry that gives platform teams and developers a single source of truth for agentic infrastructure. Key capabilities include:

• Registry and Curation: A single source of truth to package, discover, and curate AI artifacts.
• Control and Governance: Selective curation and control of custom collections of artifacts, enabling organizations to enforce policies over which AI components are available.
• Kubernetes-native deployment: Helm chart, controller-runtime integration, namespace-aware workload management, and in-cluster discovery
• Local development parity: Docker Compose orchestration with the same CLI and registry workflow used in production
• Client scaffolding: Auto-generate MCP setup for Claude Desktop, Cursor, VS Code, and Claude Code
• Data Enrichment: When AI artifacts are imported from any registry into agentregistry, the platform performs scoring and validation that enriches datasets.

Org repo URL (provide if all repos under the org are in scope of the application)

https://github.com/agentregistry-dev/

Project repo URL in scope of application

https://github.com/agentregistry-dev/agentregistry

Additional repos in scope of the application

N/A. All project code is contained in https://github.com/agentregistry-dev/agentregistry

Website URL

https://aregistry.ai/

Roadmap

agentregistry kanban

Roadmap context

The agentregistry roadmap is tracked via a public GitHub Projects board (linked above). Planned work is centered around reaching a stable 1.0 release. This will evolve as the community shares more feedback, but for now is focused on:

• first-class support for skills
• integration with evals
• formal catalog and packaging mechanism
• refactor daemon lifecycle management for local runtimes
• improve project stability
• expand the community

Contributing guide

https://github.com/agentregistry-dev/community/blob/main/CONTRIBUTING.md

Code of Conduct (CoC)

https://github.com/agentregistry-dev/community/blob/main/CODE-OF-CONDUCT.md

Adopters

N/A

Maintainers file

https://github.com/agentregistry-dev/community/blob/main/MAINTAINERS.md

Security policy file

https://github.com/agentregistry-dev/agentregistry/blob/main/SECURITY.md

Standard or specification?

N/A. Agentregistry is not a standard or specification. It implements and builds upon the Model Context Protocol (MCP) specification, but does not define its own standard.

Business product or service to project separation

agentregistry is an independent open-source project. While vendors may offer extended enterprise versions in the future, the core components remain entirely in the open-source repository. All development occurs in the public repo under a community-driven model.

Why CNCF?

Agentregistry aligns with the CNCF mission to make cloud native computing ubiquitous. As AI agents become core components of cloud native applications, the ecosystem needs governance and management tooling similar to what exists for containers and microservices. The CNCF can provide:

• Vendor-neutral governance: Ensuring the registry remains an open, community-driven project rather than being tied to a single vendor.
• Ecosystem alignment: The project already integrates with CNCF projects (Kubernetes, Prometheus, OpenTelemetry) and can benefit from deeper collaboration.
• Community growth: Access to the CNCF contributor community would accelerate development and adoption.
• Increased adoption and visibility: CNCF membership increases project visibility and adoption, particularly among enterprises evaluating open-source registries.
• Standards alignment: Positioning agentregistry as the cloud native standard for AI artifact management.

Benefit to the landscape

The cloud native landscape currently lacks a dedicated registry for AI agent infrastructure. As organizations adopt AI agents and MCP servers, they need tooling to manage these artifacts with the same rigor applied to container images and Helm charts. agentregistry provides:

• A governance layer for AI artifacts that mirrors container registry patterns familiar to cloud native practitioners.
• Integration with Kubernetes-native deployment that extends the cloud native deployment model to AI infrastructure.
• A unified proxy layer that aligns with patterns in the cloud native ecosystem.

Cloud native 'fit'

Agentregistry embodies cloud native principles:

• Containerized: Both the server and proxy are distributed as container images (published to ghcr.io), with multi-architecture support
• Kubernetes-native: Integrates with kagent for Kubernetes deployment.
• Microservices-oriented: Clean separation between the registry server, proxy, database, and CLI client.
• Observable: Prometheus metrics and OpenTelemetry instrumentation for runtime metrics.
• Declarative configuration: Uses YAML-based resource definitions

Cloud native 'integration'

Agentregistry integrates with the following CNCF projects:

Project	Project Status	Integration Context
Kubernetes	Graduated	Uses client-go, controller-runtime, and the Kubernetes API for managing deployments. Tests run against KinD clusters.
Prometheus	Graduated	Metrics exposed via the Prometheus client library.
OpenTelemetry	Incubating	Integrates OTel for metrics collection, including runtime instrumentation and SDK-based metric pipelines.
containerd	Graduated	Uses container registry libraries compatible with OCI standards for artifact distribution.
kagent	Sandbox, applying to Incubation	Used to handle the actual runtime execution of agents and MCP servers on the cluster.

Cloud native overlap

There is no direct overlap with existing CNCF projects. Agentregistry occupies a new category: AI agent infrastructure management. It shares some similarities with projects:

Project	Project Status	Overlap Context
Harbor	Graduated	Both are registries, but Harbor focuses on container images and Helm charts. Agentregistry manages MCP servers, agents, and skills. These are a different type of artifact, with different metadata, validation, and deployment requirements.
Artifact Hub	Sandbox	Artifact Hub is a discovery portal for cloud native packages. Agentregistry goes beyond discovery to include deployment, governance, and runtime management of AI artifacts.

Agentregistry is complementary to these projects rather than competitive.

Similar projects

Project	Website	Similarity
MCP Registry	https://github.com/modelcontextprotocol/registry	The upstream MCP registry schema that Agentregistry implements and extends. The MCP Registry project defines the data model, and agentregistry provides the operational platform including deployment, governance, and management capabilities. We believe the MCP registry is complementary to agentregistry, since it is designed to be an upstream registry for others projects to build on top of.
Smithery	https://smithery.ai/	A hosted registry for MCP servers. Unlike agentregistry, Smithery is not open source and does not provide self-hosted deployment, Kubernetes integration, or governance capabilities.

No existing CNCF project provides equivalent functionality.

Landscape

cncf/landscape#4755 is a PR to add the project to the landscape. The project has not yet reached the required 300 GitHub stars to qualify. We continue to expand the reach of the community and impact of the project to achieve this milestone.

Insights

N/A. Agentregistry is not yet listed on LFX Insights

Trademark and accounts

• If the project is accepted, I agree to donate all project trademarks and accounts to the CNCF

IP policy

• If the project is accepted, I agree the project will follow the CNCF IP Policy

Will the project require a license exception?

N/A. The project uses the Apache License 2.0, which is the preferred license for CNCF projects. All dependencies use licenses compatible with the CNCF Allowlist License Policy.

Project "Domain Technical Review"

• General Technical Review: https://github.com/agentregistry-dev/agentregistry/blob/9cbab504223d1bd81d9fab026267fec9578f64af/docs/governance/cncf/technical-review.md
• Security Self Assessment: https://github.com/agentregistry-dev/agentregistry/blob/9cbab504223d1bd81d9fab026267fec9578f64af/docs/governance/cncf/security-self-assessment.md

Application contact email(s)

idit.levine@solo.io;
lin.sun@solo.io;
sam.heilbron@solo.io

Contributing or sponsoring entity signatory information

Name	Address	Type (e.g., Delaware corporation)	Signatory name and title	Email address
Solo.io	222 Third Street, Suite 3300, Cambridge, MA 02142	Delaware Corporation	Solo.io Legal	legal@solo.io

CNCF contacts

Additional information

Project statistics (as of application date):

Metric	Value
First commit	October 29, 2025
Total commits	~239
Releases	v0.0.1 through v0.3.2
Unique contributors	29
Primary language	Go
License	Apache License 2.0

CI/CD:
The project has comprehensive GitHub Actions CI/CD:

• Build and Test: Runs unit and integration tests on push to main and on PRs.
• Lint: Runs golangci-lint (Go) and eslint (UI) on all PRs.
• Verify: Ensures generated code (OpenAPI spec, TypeScript client) is up to date.
• E2E Tests: End-to-end tests running against Kind Kubernetes clusters.
• Release: Automated multi-platform binary builds (linux/amd64, linux/arm64, darwin/amd64, darwin/arm64, windows/amd64) and multi-arch Docker image publishing to ghcr.io on tag push.
• Stale bot and labeler: Automated issue/PR management.