en.Wedoany.com Reported - Augment Code Computing Inc. announced on Thursday the launch of Cosmos, a team-level AI platform designed to address the issue that individual coding assistants cannot improve overall team efficiency. Vinay Perneti, the company's Vice President of Engineering, stated in an interview with SiliconANGLE that AI applications are undergoing phased evolution, which he described as "chat in '24, agents in '25, and team-oriented in '26."

Perneti believes that many engineering team leaders are pushing agent tools from the top down, yet overall team throughput has not increased correspondingly. The reason is that while individual developers can achieve efficiency gains by outsourcing some tasks through agents, the team's context and knowledge are not shared across these agent suites, leading to an uneven overall experience. Cosmos was developed to break through this bottleneck.

Cosmos operates at a level above each developer's personal agent. While providing coding agents for individual developers, it also offers the entire internal organization the ability to plan, develop, build, and deploy using cross-team knowledge. The company defines it as a software delivery lifecycle platform designed to enable collaboration between humans and agents. Agents handle most of the execution and operate autonomously, but humans remain involved in prioritization, specification, intent verification, and high-risk decision-making.

In practical applications, developers first write specification documents, agents work based on these documents, and a verification agent checks the work against the codebase and accumulated company best practices. Cosmos is designed to learn over time through every touchpoint; for example, when a Slack agent receives a correction from a colleague, it retains a memory note, maintaining patterns and shared practices represented by trends across sessions, agents, and teammates.

Since Cosmos is cross-team, non-experts can also access knowledge and data that are typically siloed. Perneti cited an example where the sales team can access a "Data Analysis Expert" agent preset by the data team through Cosmos, without needing to set up their own Claude Code, BigQuery credentials, query tools, or talk to the data analysis team. This agent is described as a "read-only expert," configured with all credentials, fully controlled by the data team, ensuring integrity and real-time access to business knowledge, optimization, and governance, while eliminating approval bottlenecks in the organization. Perneti refers to this capability of Cosmos as "self-awareness," meaning the system is aware of all sessions and running agents.

Additionally, Cosmos includes a concierge feature that helps users select the appropriate expert and environment for a task within an organization that has dozens of experts and agent workflows. Users simply describe their needs, and the concierge routes the work to the correct agent and initiates execution.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com