Agent Server [2/3]: Where Should Your Agent Server Run?

Where should your data agent server run, Snowflake, Databricks, AWS, Azure, or on-prem Kubernetes? In this video we compare deployment options for enterprise data agents, why many teams run agents inside their own environment, and how to lock down access control using service accounts and least-privilege principles.

You’ll see how containerized agent servers (Docker) can be deployed on Snowflake Snowpark Container Services (via a native app), Databricks, AWS (EC2/ECS), Azure (Containers/VMs), or on-prem Kubernetes—plus how Terraform helps standardize installs. We also break down two layers of security:

1. what systems and data an agent can touch
2. what users can do with agents (RBAC), monitored just like a new employee.

You’ll learn
- Pros/cons of SaaS-hosted vs self-hosted agent servers
- Why colocating agents with your data improves security and governance
- Containerized deployment paths: Snowflake, Databricks, AWS, Azure, K8s
- Using Terraform for repeatable installs
- Treating agents like “people” for IAM: service accounts, keys, and auditing
- Right-sizing read/write access with least privilege and RBAC

Technologies Mentioned: Docker, Snowflake Snowpark Container Services, Databricks, AWS EC2/ECS, Azure Containers/VMs, Kubernetes, Terraform, Jira, GitHub, Google Drive/SharePoint, Slack, Teams.

⏱️ Chapters
00:00 Intro: where should agents live
00:07 SaaS vs self-hosted overview
01:02 Agents replacing some SaaS tasks
01:48 Why run agents inside your environment
02:00 Containers: packaging and portability (Docker)
02:21 Snowflake native app & Snowpark Container Services
02:47 Databricks container options
02:55 AWS deployments (EC2/ECS) + Terraform
03:12 Azure deployments (Containers/VMs) + Terraform
03:27 Kubernetes & on-prem (incl. hedge funds)
04:01 Security case for running agents where data resides
04:55 Two layers of access control (agent access vs user access)
05:17 Treat agents like people/service accounts
06:05 Monitoring agent activity like a human user
06:27 Least-privilege access for data agents
06:59 Agent server manages roles/keys securely

Key takeaways
1. For data agents, run close to your data for simpler security and auditing.
2. Package the agent server in containers to deploy across cloud/on-prem.
3. Use service accounts + least privilege and monitor like any employee.
#AgentServer #DataAgents #EnterpriseAI #LLMOps #Snowflake #Databricks #AWS #Azure #Kubernetes

Видео Agent Server [2/3]: Where Should Your Agent Server Run? канала Genesis Computing