Skip to main content
CoreWeave Mission Control Agent is in public preview. Availability, behavior, and tool coverage might change during the preview. Responses are AI-generated and can be incomplete or inaccurate, so verify them before you act.
Mission Control Agent is a conversational interface in the CoreWeave Cloud Console. Ask a question in plain language, and the agent gathers evidence from your metrics, logs, dashboards, resource state, and CoreWeave documentation. In organizations where CoreWeave has enabled write tools, the agent can also perform a limited set of supported resource changes with your permissions. The agent returns an answer with supporting evidence, such as charts, tables, log excerpts, and resource details. This evidence helps you verify the answer and continue an investigation without manually correlating several observability views.

When to use Mission Control Agent

Use Mission Control Agent when you are working in the Cloud Console and want to investigate an operational question or request a supported write action without configuring a separate client. For example, you can ask about the following areas:
  • Cluster, Node, and GPU health.
  • Workload scheduling and utilization.
  • CoreWeave SUNK clusters and Slurm jobs.
  • CoreWeave AI Object Storage buckets.
  • Usage and CoreWeave product documentation.
  • Supported resource changes, when write tools are enabled for your organization. For details, see Request a supported change.

How Mission Control Agent works

Mission Control Agent runs as the signed-in user. Every request uses your identity and permissions, so the agent can access only the resources available to you in your organization. It doesn’t use a separate service account or grant itself additional permissions. The agent uses Mission Control tools to connect questions to CoreWeave observability data, resource state, and documentation. The Cloud Console also provides the current page and any resources that you reference as context for the conversation.

Capabilities and limits

Mission Control Agent can gather evidence and explain what it finds. In organizations where CoreWeave has enabled write tools, you can also request supported resource changes, such as applying a Kubernetes manifest or updating selected Inference deployment fields. Write tools use the signed-in user’s permissions and aren’t self-serve. For instructions and safety considerations, see Request a supported change. The agent can access only data exposed through its supported tools and any context that you provide in the conversation.

Mission Control Agent and Mission Control MCP

Mission Control Agent and Mission Control MCP provide access to CoreWeave operational capabilities through different interfaces. Choose the interface that matches your workflow:
  • Mission Control Agent runs in the Cloud Console and uses your signed-in identity. Use it when you want to investigate while working in the Cloud Console without additional setup.
  • Mission Control MCP is a hosted Model Context Protocol (MCP) server that you connect to an MCP-compatible client with a CoreWeave API token. Use it to bring CoreWeave data into tools such as Claude Code, Cursor, or the ChatGPT desktop app.
For an overview, see Introduction to Mission Control MCP. For MCP setup instructions, see Configure Mission Control MCP.

Next steps

Use the following resources to continue with Mission Control:
Last modified on September 24, 2026