AWS unveils Bedrock AgentCore, a new platform for building enterprise AI agents with open source frameworks and tools


Want smarter ideas in your inbox? Sign up for our weekly newsletters to get what is important for businesses, information and security leaders. Subscribe now


Giant Amazon Web Services (AWS) He believes that the AI agents will need to interact with how we all work and interact with information and allow enterprises to allow and place agents on a scale.

Today at his New York summit, AWS has been announced Amazon Bedrock Agentcore, A new enterprise platform designed to settle and build, place and operate AI agents.

Agent Vice President of the EU Sivasubramanian, Ai Sivasubramanian, helps to transfer agents to the manufacturer-ready-made agent systems that can trust the most critical business processes. “

AgentTore, a modular stack in advance, it provides the main infrastructure for the development, including the prototype, including the prototype production, memory, identity, observation, API integration and tools for web reviews and web views.


AI impact series returns to San Francisco – August 5

The next stage of AI is here – are you ready? Block, GSK and SAP to how autonomous agents change the workflows of enterprises – Join the leaders to take an exclusive look after the real-time decision for the latest automation.

Take your place now – Location is limited: https://bit.ly/3guplf


“We believe how agents will change how we use the Internet,” Deepak Singh, Vice President of Deepak Singh and AI “The line between the agent and an application is mixed.”

AgentCore is built within the existing Bedrock agents launched in late 2024, but by supporting any agent framework or foundation – not only in the bed.

This includes Crewai, Langchain, Langgraph, Langgraph and AWS compatibility with a set of open-source tools like SDK in their own strands agents.

What does Bedrock agentcore contain

  • Agent Walking Duration: Sesimodal business loads and a server-free, low-delaying environment that supports long-term sessions with insulation of session.
  • Agency Memory: Long and short-term memory services that continue the agents from past interactions and continue contextual knowledge between the meetings.
  • AGENTTORE ID: OAuth-based personality and access management, agents provide for moving on behalf of users in systems such as Github, Slack or Salesforce.
  • Agentitor Observation: Mounted tables, bears and telemetry vehicles installed with OpenTelemetree, Langsmith and Datadog support.
  • Agentcore Gateway: The model uses the context of the protocol (MCP) into the internal API, lambda functions and third party services, agents to suitable tools.
  • Agentcore Browser: Provides headless browser access to agents to interact with websites with websites.
  • Agentcore Code translator: A valid environment to perform the code generated by agents for analysis and visualization.

AgentCore, also combines with AWS Marketplace, allows you to discover and place pre-built agents and tools from teams.

According to Singh, the agency was developed with a reciprocal contact. MCP and Google supports developing industrial standards such as an agent-2-agent (A2A) protocol. The properties such as agentic personality and gateway allow the agents to clearly monitor and can interact safely with internal systems and third parties APIs.

The launch of AWS puts it in the center of those who quickly become one of the most competitive segments in the enterprise.

Openai agents SDK and Google’s Gemini-based agents SDK We push similar images of agent development platforms until the end.

Writer AI HQ and as the beginnings Cognition (Devin manufacturer) It is also tools for managing autonomous program agents.

“Agents are the most effective change in centuries,” Sivasubraman said. “As an application with agents, it comes to service. This is a tectonic change in how the program is built, placed and employed.”

CUSTOMER SETTLEMENT AND EARLY USE

Several companies, which gave an early access to Agency, are already building production-grade agent applications in industries, including financial, health, marketing and content management.

Cloud document and file storage company box, strands agents and bed agents are investigating the ways to expand content management using the processing period.

CTO Ben Kus provides “the highest level security and compatibility” mobile customers “highest level of security and compatibility” when measuring AI capabilities in integration, enterprise environment.

Brazilian ITAU uses agentcore to support the development of hyper, safe digital banking experiences of hyper, safe digital banking. Chief Technology Officer Carlos Eduardo Mazzei said that the new platform will help provide an intuitive bank experience with the effectiveness of automation and individual customers. “

In the field of health, Innovaccer has built a new protocol-HMCP (Health Model Model Protocol) – AgentCore Gateway. CEO and co-founder Abhinav Shashank, while maintaining the company’s trust, compatibility and operational efficiency, the company is called “game change” that allows the company to turn the existing API into the agent.

The marketing company Epsilon uses agentcore to accelerate the time and improve the time of building the campaign. Prashanth Athota, Software SVP, the company has increased its construction times to 30% and increases the personalization of customer travel.

Availability and Price

AgentCore is in advance in the AWS regions, including the US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney) and Europe (Frankfurt). It is free to test until September 16, 2025, which will start after this.

Prices for AgentCore are completely consumption based, no obligation or minimum payment. Each module-working time, memory, personality, observation, gateway, browser and code translator – independently, or can be used in a la carte.

According to the execution period, browser and code translator services, CPUs and memory usage, prices for prices and $ 0.0895 per hour and $ 0.00945 and $ 0.00945 and $ 0.00945.

Gateway 1000 tool for API invitations for $ 0.005, $ 0.025 for each 1,000 search queries, $ 0.02 in 100 vehicles.

Memory costs are based on data volume: $ 0.25 for short-term memory events, $ 0.75 for 1000 long-term memories (or 0.25 dollars in special strategies) and 1,0.50 dollars for $ 1,0.50.

Even if the agentCore ID is used by 1,000 token or API key desires, if not included in additional payment when used by 0.010, working hours or by a gateway.

The observation is calculated by the Amazon CloudWatch ratios.

To get more information or start work, AWS developers directly for him AgentCore documents, GitHub samples and dedicated Disagreement.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *