CoreWeave Inc. has unveiled Forge, a platform aimed at speeding up the iterative AI improvement loop for enterprises deploying intelligent agents. Forge provides a shared workflow for business and machine learning teams to continuously run, observe, evaluate, and improve AI models and agents.
- Forge facilitates faster AI agent iteration via a unified workflow.
- New features include Agent Lens for inspection and ARIA for pattern detection.
- CoreWeave builds a partner network with co-engineered integrations to accelerate adoption.
What happened
CoreWeave launched Forge, a full-stack AI platform designed to accelerate the continuous improvement loop for agentic AI systems used by enterprises. The platform enables teams to collaborate across the stages of running, observing, curating, improving, and evaluating AI agents, making the iterative process more efficient and transparent.
Forge includes tools such as Agent Lens for detailed agent behavior inspection and ARIA, an AI Research and Iteration Agent that helps surface improvement opportunities by spotting usage patterns. Additionally, CoreWeave offers Registry to manage model checkpoints and configurations, along with RL Rollouts and model distillation to drive agent enhancements.
Why it matters
The AI improvement loop is a never-ending cycle crucial for deploying productive, reliable AI agents. By focusing on speeding up this cycle, Forge helps organizations shorten the time between iterations, thus delivering value faster by continuously refining agent performance after initial deployment.
With AI workloads rapidly evolving, enterprises need platforms that support ongoing model improvements and real-time inference optimization. CoreWeave’s approach also addresses one of the toughest challenges in AI operations — bridging multiple teams and workflows — to reduce fragmentation and accelerate outcomes.
What to watch next
CoreWeave is rolling out a partner network with co-engineered integrations and pre-built playbooks to help enterprises rapidly implement Forge’s capabilities and avoid duplicative learning. This ecosystem expansion will be critical to adoption and real-world impact as organizations scale their AI efforts.
Future developments to monitor include further automation features within the AI improvement loop, enhancements to core tooling like Agent Lens and ARIA, and how well CoreWeave can support production workloads such as those launched by Cognition AI on new Nvidia infrastructure. These will indicate Forge’s maturity and effectiveness in accelerating AI model lifecycle management.