Container Images & Specifications¶

Images¶

AgentCube connectors are distributed as Linux container images hosted on GitHub Container Registry (GHCR):

Connector	Image
Essbase	`ghcr.io/caprusai/agentcube-essbase:{version}`
Planning	`ghcr.io/caprusai/agentcube-planning:{version}`

Authentication¶

Images are hosted in a private GHCR registry. Caprus AI will provide you with a GitHub Personal Access Token (PAT) for image access.

Authenticate with Docker:

echo "{github_pat}" | docker login ghcr.io -u {github_username} --password-stdin

Authenticate on managed container platforms:

For AWS ECS, OCI Container Instances, and similar platforms, store the provided credentials as a secret in your platform's secret manager and reference them in your deployment configuration. Each platform-specific guide covers this step.

Container Specifications¶

Setting	Value
Port	`8080`
Base image	Python 3.12 (slim)
Minimum CPU	0.25 vCPU
Minimum memory	0.5 GiB
Recommended replicas	1 or more (scaling to zero is not recommended)
TLS	Handled externally — container serves HTTP on port 8080

Scaling recommendation

Keeping at least one instance running at all times provides the best experience. Scaling to zero is possible but will result in cold start delays that can cause slower responses or timeouts during AI sessions while the connector restarts.