Kaigents Platform (GA Release)

Kaigents is a production-ready, Kubernetes-native platform for building, running, and operating AI agents in enterprise environments. It is optimized for low total cost of ownership (TCO) with a strong focus on AMD Ryzen AI hardware.

Current Version: 1.0.0 (General Availability)

This repository is the Kaigents platform (distinct from any future marketing/community site).

Canonical docs

• docs/product/kaigents-prd.md
  • Product goals, MVP scope, functional requirements, UX requirements (run timeline), and milestones.
• docs/architecture/kaigents-architecture-and-design.md
  • Canonical system design: boundaries, data flows, and the role of tool plane, run timeline, artifacts, and RAG.
• docs/implementation/kaigents-implementation-tracker.md
  • Milestone tracker and push/review gates (when we are allowed to push working code).
• docs/CODING_STANDARDS_AND_DOD.md
  • Coding standards, CI quality gates, and definition of done.
• docs/research/technology/itd-register.md
  • Important Technical Decisions (ITDs) that constrain implementation choices.
• docs/research/technology/oss-components-commercially-permissible.md
  • OSS due diligence list and licensing posture (redistribute vs integrate-only vs exclude).

Ecosystem Projects

Kaigents is supported by a suite of specialized ecosystem projects:

• KaiCatalog: A curated catalog service for MCP servers with 40+ pre-vetted tools, security posture ratings, and air-gapped mirroring support.
• KaiManager: The management layer for Kaigents, providing first-class surfaces for agent personas, quality gates, and process health monitoring.
• KaiCLI: The operator interface for the Kaigents ecosystem. A unified command-line tool for managing MCP tools, personas, and platform health.

Production Hardened

Kaigents 1.0.0 is built for stability and enterprise operations:

• Durable Execution: Long-running workflows survive component restarts.
• Observability: Full Prometheus metrics and JSON structured logs (Loki) across all components.
• Enterprise Storage: Cloud-agnostic S3 support with large-object streaming.
• Identity: Built-in OIDC (Keycloak) and Kubernetes RBAC integration.

Features

• Kubernetes-native: Built on CRDs, standard RBAC, and GitOps-friendly workflows.
• Enterprise Identity: Full OIDC integration with Keycloak for platform-wide SSO.
• Durable Execution: Powered by Temporal for long-running, human-gateable agent workflows.
• Hybrid Execution: Declarative hardware pinning (CPU/GPU/NPU) via RoutingPolicy.
• Observable: Structured JSON logging (Loki-ready) and Prometheus metrics.
• Cloud-Agnostic Storage: S3-compatible artifact storage (AWS, MinIO, Ceph).

License and OSS posture

• Kaigents is MIT-licensed.
• Core dependencies must remain commercial-safe (redistribution-safe).
• Integrate-only components (user-supplied) are allowed only when clearly separated and documented.

Getting Started

Follow the Getting Started Guide to install Kaigents and deploy your first AI agent team.

Managed AI Teams

You can build any agent team on Kaigents yourself. You can also skip the build — we operate a growing catalog of production-ready AI agent teams as fully managed services. See docs/product/managed-services.md for an overview of available teams and how to get access.

Development

This repo uses a minimal Makefile to run formatting, linting, and tests when relevant toolchains are present.

make fmt
make lint
make test

Operations

• Temporal installation

Project status

Implementation scope, milestones, and acceptance criteria are tracked in docs/implementation/kaigents-implementation-tracker.md.