Google Cloud made a major update to maintain AI ddevelopers on the Vertex AI platform to conceptualize, design, build, test, deploy and modify AI agents for enterprise applications.
New features announced today include additional enterprise management tools and expanded agent creation capabilities with just a few lines of code, faster performance with state-of-the-art context management layers and one-click deployment, and managed services to scale production and evaluation and support agent identification.
Agent Builder, released last 12 months at the annual event, Cloud Next provides enterprises with a no-code platform to create agents and connect them to orchestration platforms like LangChain.
Google Agent Development Kit (ADK), which allows developers to create agents “in less than 100 lines of code,” is also available through Agent Builder.
“These new capabilities underscore our commitment to Agent Builder and simplify the agent development process to meet developers where they are, no matter what technology stack they choose,” said Mike Clark, director of product management at Vertex AI Agent Builder.
Create agents faster
Part of Google’s proposal for new Agent Builder features is that enterprises can create orchestrations while building their agents.
“Building an agent from concept to working product requires complex orchestration,” Clark said.
New capabilities shipped with the ADK include:
-
SOTA context management layers, including static, turn-based, user and cache layers, giving enterprises greater control over agent context
-
Ready-made plugins with configurable logic. One of the new plugins allows agents to acknowledge failed tool calls and “self-repair” by retrying the task with a different approach
-
Additional language support in the ADK, including Go, in addition to Python and Java that launched with the ADK
-
One-click deployment via the ADK command-line interface to maneuver agents from on-premises to live testing with a single command
Management layer
Enterprises require high accuracy; security; observability and auditability (what the program did and why); and steerability (control) in their production-grade AI agents.
While Google provided observability features in the local development environment at launch, developers can now access these tools through a dashboard managed by Agent Engine.
The company said this provides cloud production monitoring to trace token consumption, error rates and latency. From this dashboard, enterprises can visualize actions taken by agents and reproduce any issues.
Agent Engine may also feature a new evaluation layer that may help “simulate agent performance across a wide range of user interactions and situations.”
This level of management may also include:
-
Agent identities, which Google says give “agents their very own unique, native identities in Google Cloud
-
An Armor model that may block rapid injections, screen tool calls, and agent responses
-
Security Command Center, so administrators can inventory their agents to detect threats akin to unauthorized access
“These native identities provide a deep, built-in layer of control and a clear audit trail for all agent activities. These certificate-backed identities further strengthen security because they cannot be spoofed and are tied directly to the agent lifecycle, eliminating the risk of inactive accounts,” Clark said.
Agent Builder Battle
It’s no surprise that model vendors are creating platforms to build agents and put them into production. The competition comes all the way down to how quickly new tools and features are added.
Google’s competition is Agent Builder OpenAIis open source software Agent Development Kitwhich allows developers to create AI agents using models aside from OpenAI.
Plus it’s recent announced AgentKitwhich incorporates an Agent Builder tool that permits corporations to simply integrate agents into their applications.
Microsoft has its own Azure AI Foundrylaunched last 12 months around this time to create AI agents, and AWS also offers agent builders on his Bedrock platform, but Google hopes a set of new features will give it a competitive advantage.
However, it isn’t just corporations with their very own models that are encouraging developers to create AI agents on their platforms. Every enterprise service provider with a library of agents also wants customers to create agents on their systems.
Capturing developer interest and keeping them in the ecosystem is a big battle among tech corporations today, and features make it easier to create and manage agents.