Skip to main content

Deploy and operate Mecatl

Use these guides when you own the Mecatl service, its credentials, access policy, execution environment, and durable state. The clients connect to the service; they use the model access and workspaces you provide.

Start with a lightweight server​

Run mecated to separate the agent service from its clients. Start one instance, configure model access, and connect a client. Add local persistence when you need to resume sessions across restarts.

This is the simplest way to learn the service boundary or operate one controlled instance. Its process and storage are yours to maintain.

Operate a shared team deployment​

Use mecak8s for a shared team, including small teams. It supplies Redis-backed durable state, Kubernetes lease coordination, and disposable replicas. Start with Try Mecatl on Kubernetes to see a session survive replacement of its pod, then follow the deployment and maintenance guides.

Operating modelWhat you own
One mecated instanceHost lifecycle, workspace, local or configured session storage, and client access.
Shared mecak8s deploymentKubernetes deployment, Redis, access and network policy, execution placement, and capacity.
Custom multi-replica mecated compositionShared storage, a lease backend, routing, and replica lifecycle that mecak8s otherwise supplies.

Shared deployments are intended for trusted teams. Caller ownership and permissions govern access, while filesystem isolation depends on the workspaces and execution environments you provision. See Caller identity and Execution environments.

Configure and maintain the service​

The server CLI reference lists the complete server commands and flags. The capability matrix compares behavior and availability across deployment forms.

To integrate Mecatl into an application or CI workflow, follow Build with Mecatl.