This repository holds runnable examples for runkora, an open source control plane for running agent factories in production.
make up installs kora on a local k3d cluster. Run it from the repository root:
make upSee quickstart for the prerequisites and the install stages.
A coding agent can also run the quickstart. Ask it to set up runkora from this repository, and it follows .agents/skills/runkora-quickstart/SKILL.md.
An example is a set of Kubernetes manifests that declare tenants, models, agents, triggers and roles for kora. Choose one when make up asks, or apply one later with make example EXAMPLE=<name>:
single-tenant: One tenant with an agent that summarises pull requests.multi-tenant: An engineering team and an SRE team, each in a separate tenant and repository.factories/autofix: A factory of four agents that finds, verifies and fixes problems in a GitHub repository and hands a person the pull request.factories/sre: A factory that triages alerts in the cluster, sweeps its costs with OpenCost and proposes fixes as pull requests.factories/feature: A factory that turns a feature request into an approved spec, splits it into tasks and builds each task in a pull request.
Each example splits its manifests by owner:
platform/holds the tenants and models that the platform team owns.- A tenant folder, such as
tenant/orengineering/, holds one team's agents, triggers and roles.
In production, each folder is a separate Git repository that a GitOps tool such as Argo CD or Flux applies to the cluster.
versions.mk pins the kora release and the Istio, SPIRE, External Secrets Operator and k3s versions every example uses. The examples on main work with the latest kora release.
make validate checks every manifest against the Kubernetes, kora and External Secrets Operator schemas at those versions. CI runs it on every pull request. It needs helm, kubectl, kubeconform and Python with PyYAML.
Licensed under the Apache License, Version 2.0.