<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Kelos is a Kubernetes-native framework designed to manage autonomous AI coding agents as infrastructure, moving beyond simple command-line tools. It enables users to define agents, their environments, and complex workflows declaratively using YAML, integrating them seamlessly into CI/CD pipelines and event-driven systems. This empowers developers and researchers to orchestrate AI agents at scale, automating coding tasks, code reviews, and development processes within a secure, version-controlled Kubernetes environment.

How It Works

Kelos treats AI agents and their operational context (code repositories, credentials, models, compute resources) as Kubernetes Custom Resources. Agents are defined via AgentConfig, their execution environments via Workspace, and individual runs as Task resources. Workflows are triggered by TaskSpawner reacting to external events (e.g., GitHub issues, webhooks) or chained together using dependsOn for complex pipelines. This Kubernetes-native approach allows for declarative management, GitOps integration, and scalable, isolated execution of AI agents.

Quick Start & Requirements

• Installation: Requires a Kubernetes cluster (v1.28+) with cert-manager installed. Installation involves kelos install for the controller and CRDs, followed by kelos init to configure agent credentials. Local development can use kind.
• Prerequisites: Kubernetes cluster, cert-manager, Docker (for kind), AI model API keys/tokens (e.g., OpenAI, Anthropic, Google), GitHub Personal Access Token or GitHub App credentials.
• Links: See the Quick Start section in the README for detailed steps, Examples for practical use cases, and the Reference for resource details.

Highlighted Details

• Broad Agent Support: Integrates with Claude Code, OpenAI Codex, Google Gemini, OpenCode, Cursor, and custom agent images via a standardized container interface.
• Declarative Workflows: Define agent tasks, triggers, and pipelines entirely in YAML, enabling GitOps and version control for AI operations.
• Orchestration & Integration: Supports event-driven automation via TaskSpawner (GitHub Issues, Webhooks, Jira, etc.) and task chaining (dependsOn) for complex pipelines.
• Secure Execution: Agents run in isolated, ephemeral Kubernetes Pods with restricted access, limited only by injected credentials, enhancing security.
• Observability: Agent runs are first-class Kubernetes resources, monitorable via kubectl, the kelos CLI, and integrate with GitOps tools like ArgoCD.

Maintenance & Community

The project appears actively developed, with mentions of self-development patterns and sibling projects. Specific community channels (Discord, Slack) or notable contributors are not detailed in the README.

Licensing & Compatibility

• License: Apache License 2.0.
• Compatibility: Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

• Kubernetes Dependency: Kelos is strictly Kubernetes-dependent and cannot be run outside of a cluster environment.
• Development Stage: Custom Resource Definitions use v1alpha2, indicating potential for API changes and breaking updates as the project matures.
• Credential Management: Requires careful setup and management of various API keys and tokens for different AI models and repository access.