Cloud Agents

What are Cloud Agents?

A cloud agent is an AI agent, usually a general-purpose coding agent such as Claude Code, Codex or an open-source equivalent, that runs in a remote environment the organization controls rather than on an individual's laptop. The agent itself is generic. What makes it a cloud agent is where it runs and how it is started: a server or container with the right credentials, instructions and tools, launched by a trigger such as a schedule, a webhook, a new row in a database or a message in a queue.

Amol Kapoor of Nori Agentic uses the term for agents deployed in finance back offices. In his AI Engineer talk, a cloud agent starts whenever a cell changes in an asset manager's security master, runs an on-prem open-source model, reads PDFs in Box, calls Bloomberg and reads a slice of email, then posts an explanation and recommendation to the cell's owner in Slack.

Key Characteristics

• Event- or schedule-triggered - No human starts the run. Work begins when data changes or a timer fires.
• Auditable and observable - Because the agent runs on managed infrastructure, every run, tool call and output can be logged and reviewed, which matters in regulated industries.
• Generic agent, specific configuration - The same agent binary runs the sec master check, the audit job and code review. Instructions, tools, triggers and credentials differ; orchestration, security and integrations are shared.
• Configured in natural language - Behavior is set with skills and agent markdown files that say where documents live, which sources to trust and how to judge a result.
• Data stays put - The environment can run a local model so sensitive data never leaves the company's network.

Why Cloud Agents Matter

Agents on a laptop wait for a person to notice a problem and type a prompt. Cloud agents flip the direction: the agent notices, assembles the case and brings a recommendation to the person. Kapoor's framing is that the future of enterprise AI is "having the robots come to you with tasks, advice, analysis, and recommendations." For organizations, this is how a single deployment pattern turns into many agents as employees doing low-complexity, high-effort work across a firm, with a human in the loop only for decisions that need accountability.

Related Terms

• AI Agents - The general category
• Agent Skills - How cloud agents are configured for a specific job
• Human-in-the-Loop - The approval step most cloud agent workflows keep
• Sovereign Cloud - Running models where the data lives