Evidence tier Source Confirmed · 4 captures on record
What the publisher says
As described on AWS Marketplace.
We built this platform to solve a specific problem: running AI Agents in production without the operational headaches that come with serverless glue code and third-party orchestration layers.
The runtime sits directly on Amazon EKS. Messages from end users hit the Pod through an in-process Channel Manager. No Lambda hops, no SQS detours, just a straight path from ingress to the Agent loop. This keeps latency predictable and makes the whole request lifecycle visible in a single trace.
Show the rest of the publisher’s description (5 more lines)
The orchestration engine is ours, not a wrapper around LangChain or AutoGen. It manages tool dispatch, memory retrieval, and LLM calls in one tight loop. Conversation history, Agent configuration, and working memory all land in PostgreSQL on every turn. If the Pod dies mid-response, you pick up exactly where you left off.
Execution is cleanly separated from orchestration through a pluggable Sandbox architecture. The default Sandbox spins up Docker containers with sub-second cold starts. For workloads that need hard isolation (multi-tenant or regulated environments), AgentCore provides Firecracker-backed micro-VMs with hardware-level isolation. Sandboxes are created lazily on the first tool call and automatically reclaimed after idle timeout. Pure-chat workloads consume zero Sandbox resources, and workspace state is persisted to S3 between sessions. The executor itself is roughly 200 lines of Go, small enough to audit in one sitting.
Concurrency is handled per-Pod with database-level session isolation keyed on agentID, accountID, and chatID. Sub-Agent delegation works out of the box, with cascading cancellation and timeout enforcement so runaway tasks do not pile up.
On the integration side, we support Telegram, Discord, Slack, LINE, WeChat, Feishu, and WeCom. Connect a channel, point it at your Agent, and you are live. Most teams get their first Agent into production in under five minutes.
Self-hosted, open source, no phone-home telemetry. If you need to own your Agent infrastructure end to end, this is the stack.
Highlights
Highlighted by the publisher on AWS Marketplace.
Runs AI Agents directly on Amazon EKS Pods. Messages go from channel ingress to the Agent loop in two hops, no Lambda or queue layers. Pod state is fully externalized to PostgreSQL, so Pods stay stateless, restarts lose nothing, and horizontal scaling is just adding replicas.
Pluggable Sandbox architecture with two isolation tiers. Docker containers spin up in sub-second for standard workloads. AgentCore delivers Firecracker-backed micro-VM isolation for multi-tenant or regulated environments.Sandboxes are lazily created and auto-reclaimed on idle. You pay only for actual execution.
Ship your first Agent in under five minutes. Connect any supported channel (Telegram, Discord, Slack, LINE, WeChat, Feishu, WeCom), configure your Agent, and go live. Sub-Agent task delegation, cascading timeout enforcement, and unified in-process tracing are included out of the box.
Agent build and provenance
See the full provenance
The layer-by-layer build, the evidence behind each claim, the risk basis and the cross-marketplace links are open to any account. Some rows are disclosed, some the source leaves Unknown; a free account shows you which.
Plans and pricing as listed
4 listed- Units
- Units
- Units
- Units
Refund terms
As stated by the publisher on AWS Marketplace.
All sales are final. No refunds are provided for contract-based purchases. If you experience technical issues or believe you were charged in error, please contact us at zhangyingting@hzcloud.com.cn within 30 days of purchase and we will review your case individually.
Sources
Publisher resources
2 linksLinked repositories
Unknown means this listing does not publish a repository. It is not a statement that the code is closed, and a linked repository is not a claim that the publisher wrote it: the registry computes that relationship privately and does not publish it.
Evidence risk is the share of the build you cannot see before you deploy, not a security rating. Sign in to see the layer-by-layer basis for this band.