# Kontext

> Kubernetes-native control plane for running AI agents as production workloads.

Full docs: https://docs.kontext.run/llms-full.txt

## Pages

- [Introduction](https://docs.kontext.run/raw/docs/index.md): Kontext runs AI agents as Kubernetes workloads with Agent and AgentRun resources.
- [Quickstart](https://docs.kontext.run/raw/docs/quickstart.md): Install Kontext from a published release and run a keyless echo AgentRun without cloning the repository.
- [Resource model](https://docs.kontext.run/raw/docs/resources.md): How Agent and AgentRun map to familiar Kubernetes objects.
- [Task workload](https://docs.kontext.run/raw/docs/task-workload.md): Create a reusable Task Agent and invoke it through a sparse, user-named AgentRun.
- [Scheduled workload](https://docs.kontext.run/raw/docs/scheduled-workload.md): Run one-shot AgentRuns from cron slots with deadlines, overlap policy, suspension, and retained history.
- [Service workload](https://docs.kontext.run/raw/docs/service-workload.md): Deploy a persistent Service-mode Agent and watch the controller re-cast it after Pod deletion.
- [Operations](https://docs.kontext.run/raw/docs/operations.md): Alpha support matrix, failure boundaries, identity, secrets, network, budgets, and troubleshooting.
- [Releases](https://docs.kontext.run/raw/docs/releases.md): Version tags, GHCR images, digest-pinned install.yaml, upgrades, and uninstall.
- [Runtimes](https://docs.kontext.run/raw/docs/runtimes.md): Echo, reference, and bring-your-own runtime roles and result capture paths.
- [Evaluations](https://docs.kontext.run/raw/docs/evals.md): How kontext-eval grades AgentRuns outside the agent Pod.
- [API specification](https://docs.kontext.run/raw/SPEC.md): Agent, AgentRun, and runtime-image contract for kontext.dev/v1alpha1.
- [When not to use agents](https://docs.kontext.run/raw/docs/when-not-to-use-agents.md): Prefer deterministic Jobs and scripts when model judgment is not justified.