Augment launches Cosmos platform for multi-agent software development workflows

Augment has released Cosmos, a platform that coordinates multiple AI agents across the software development lifecycle, now available to all team plans. The platform moves beyond single-agent code generation to orchestrate agents across triage, specification, implementation, review, testing, deployment, and feedback.

Platform architecture

Cosmos operates as what Augment calls an "operating system" that coordinates both AI agents and human developers. According to the company, the platform distinguishes itself from single-agent tools or workflow engines by enabling specialized agents to coordinate, delegate work, and share memory across long-running tasks.

The system includes:

• Multi-interface access through web, mobile, CLI, Slack, and Linear
• Deployment options including Augment's cloud sandboxes, self-hosted VMs, or local execution
• Integration capabilities via APIs, Model Context Protocol (MCP) support, and webhooks
• Shared filesystem and memory system that persists patterns and corrections across sessions

Natural language workflow configuration

Teams can configure agent workflows using natural language commands. Augment provides the example: "When feedback lands in #feedback-billing, triage it, open a Linear ticket, take a first pass at the fix, and open a PR." The platform then deploys the necessary agents to execute these multi-step workflows.

According to Augment, Cosmos agents can also modify the platform itself, helping teams build automations and debug existing workflows. This self-improvement mechanism is designed to expand platform capabilities as teams use it to solve new problems.

Deployment and pricing

Cosmos is available immediately to all Augment team plans. Pricing details for team plans were not disclosed in the announcement. The company positions the platform as infrastructure for what it terms "AI-native engineering teams"—organizations restructuring their development processes around agent-first workflows.

Chris Kelly, Product Lead for Cosmos, states the platform incorporates best practices refined through hundreds of customer engagements by Augment's applied team.

What this means

Cosmos represents a shift from point-solution AI coding assistants to platform-level orchestration of multiple specialized agents. The key technical challenge will be whether the shared memory and coordination mechanisms can maintain coherence across complex, multi-stage workflows without excessive human intervention. The platform's success depends on whether the abstraction of "teams of agents" provides genuine productivity gains over simpler, single-agent tools, or introduces coordination overhead that negates the benefits of parallelization. Augment's approach of making the platform self-modifiable by its own agents is technically ambitious but raises questions about drift and debugging in production environments.