AWS AgentCore vs semantic
AgentCore Runtime gives each session its own microVM, terminated and sanitized when the session ends. [1] It is a managed AWS service. [2]
AgentCore Identity keeps credentials in a KMS-encrypted token vault [3], and AgentCore Policy checks each tool call at the Gateway with Cedar rules. [4] Semantic runs on your own hardware, and its permissions cover files, Git and SSH, not only tool calls.
| AWS AgentCoremanaged AWS service | Semantic (smx)microVM + policy on your machines | |
|---|---|---|
| Isolation | microVM per session [1] | One microVM per agent, own kernel |
| Where it runs | Managed AWS service only [2] | Managed private beta; on-prem for enterprise (contact us) |
| Egress rules | Public or VPC mode; your security groups and routes [5] | Security profiles enforced by host firewall; metadata and link-local always closed |
| Credentials kept out | Partial: token vault [3] | Git tokens, SSH keys and model sign-ins stay on the host |
| File, Git, SSH permissions | Limited: tool-call policy at the Gateway [4] | Per path, repo, branch and SSH host; deny beats allow; enforced on the host |
| Tamper-evident agent log | Partial: CloudWatch and CloudTrail [4][6] | Hash-chained log of runs, events and refusals; read with smx agent logs |
| Snapshots, pause/resume | Stop and resume with session storage; memory — [1] | Snapshot, pause and resume with memory, disk and running processes |
| Local hosts (Linux / WSL2 / macOS) | Not applicable (managed) | Linux (KVM) and Windows via WSL2; macOS planned |
| License | Proprietary; Python SDK Apache-2.0 [7] | Proprietary (private beta) |
Which one fits
When AWS AgentCore fits better
- You are on AWS and want a managed runtime with a microVM per session. [1]
- You want tool-call policy in Cedar, including rules on session history. [4]
When Semantic fits better
- Code, data and credentials must stay in your environment; on-prem is available for enterprise (contact us).
- Agents work on your repos and you need branch-level Git and per-path file permissions enforced outside the agent.
- You need a tamper-evident log of what the agent did inside the sandbox, including refusals.
Tool calls vs actions
AgentCore Policy intercepts requests through its Gateway before the tool runs. [4] Semantic enforces at the resource: a file read, a Git push or an SSH session is checked on the host, whatever tool or shell the agent used.
Questions
Can AgentCore run on my hardware?
The runtime is a managed AWS service; Identity also works with self-hosted agents. [2]
Does AgentCore snapshot memory?
Its docs describe stop and resume with session storage for files, not memory snapshots. [1]
Sources
- AgentCore runtime sessions · last verified 2026-10-27
- What is AgentCore · last verified 2026-10-27
- AgentCore key features · last verified 2026-10-27
- AgentCore Policy · last verified 2026-10-27
- AgentCore VPC · last verified 2026-10-27
- AgentCore Code Interpreter · last verified 2026-10-27
- AgentCore Python SDK · last verified 2026-10-27