Verified project record
agent-substrate/substrate
substrate is a secure-by-default agent execution runtime that runs large numbers of sandboxed, mostly-idle stateful agent workloads with 10x higher density than standard container runtimes. It multiplexes many actors onto a small pool of shared workers on Kubernetes, targeting millions of sandboxes.
Project overview
It demonstrates 30x+ oversubscription by exploiting that agents are idle most of the time, addressing the concrete inefficiency of running huge fleets of stateful sandboxes with standard container runtimes.
- Project type
- AI Agent · Infrastructure
- Use cases
- Automation
- Deployment
- Refer to project documentation
- License
- Apache-2.0
Best for
- Platform and infrastructure developers who operate fleets of stateful, mostly-idle sandboxed agents on Kubernetes and need dense multiplexing onto shared compute.
- Teams whose agents are built on any stack, since Substrate is framework- and harness-agnostic and manages standard OCI containers.
Key capabilities
- Maps many actor applications onto a smaller pool of ready workers, achieving heavy multiplexing (30x+ oversubscription demonstrated) by exploiting that agents are idle most of the time.
- Router holds inbound requests until a worker frees up instead of returning 503 during transient worker-pool saturation.
- A CLI tool (kubectl plugin) for managing Agent Substrate resources such as actors.
- Actor logging, metrics, and distributed tracing guides and support.
Limitations and risks
- Pre-1.0: no backward compatibility guarantees; APIs and behavior may still change significantly.
- Supports only the latest stable Kubernetes release and the previous minor release.
- The project is very young and may not review or merge contributions that don't align with core goals.
- Operating the system requires coding ability plus Kubernetes and developer tooling.
- The cloud deployment path provisions GCP resources (GKE, GCS, IAM); GPU requirements, telemetry behavior, and the end-to-end data boundary are not documented, so verify these against your own constraints before adoption.
Getting started
- Install Go, kubectl, and Docker; run hack/create-kind-cluster.sh, then hack/install-ate-kind.sh --deploy-ate-system and --deploy-demo-counter; install the CLI with go install ./cmd/kubectl-ate; create an actor with kubectl ate create actor my-counter-1 -a ate-demo-counter --template counter; and expose the router with kubectl port-forward -n ate-system svc/atenet-router 8000:80.
- For GKE, setup tooling provisions GCP resources (GKE, GCS, IAM, Cloud SQL), involving multi-step provisioning scripts and GCP resource setup.
Evidence and sources
- README: Agent Substrate is a secure-by-default agent execution runtime engineered to run millions of sandboxes with 10x higher density than standard container runtimes.
- README: **Agent Multiplexing:** Demonstrates 30x+ oversubscription by "juggling" a large registry of stateful actors onto a small pool of shared physical pods.
- README: Because it manages standard OCI containers at the kernel level (via gVisor), it can host agents built on any stack.
- README: A CNCF Sandbox project and Kubernetes-native framework for building, deploying, and managing AI agents that uses Agent Substrate to run sandboxed, stateful agent workloads
- README: Agent Substrate is pre-1.0. We are not making any guarantees about backward compatibility at this stage
AI Search
Find projects, verify facts, compare options, or turn a complex need into an actionable plan
Try a searchA click only fills the search box; you stay in control
Project Details
0