Skip to content
Warp

Automation Platform > Cloud Agents

Cloud agents overview

Start a team workflow with Warp Factories, or find the CLI, setup, and management references for existing standalone cloud agents.

Open in ChatGPT ↗
Ask ChatGPT about this page
Open in Claude ↗
Ask Claude about this page
Copied!

Cloud agents execute a request away from your laptop, on Warp-hosted compute or your own infrastructure. For example, an agent can investigate a failed build, review a pull request, or respond to an issue filed in your tracker.

Start with Warp Factories

Section titled “Start with Warp Factories”

For a standing team workflow, start with the Warp Factories quickstart. A factory brings together agents, automations, and evaluation.

See how Warp Factories work for the task lifecycle and human review points.

Continue a standalone workflow

Section titled “Continue a standalone workflow”

To start a cloud agent directly from Warp, use the cloud agents quickstart. Existing standalone runs, schedules, and integrations have their own configuration and management paths.

Already using the Oz web app, Oz CLI, or SDKs? See transitioning from Oz to Warp Factories before changing an existing workflow.

Monitor, inspect, and share cloud agent runs

Section titled “Monitor, inspect, and share cloud agent runs”

For factory work, open the factory’s Runs page. For standalone workflows, use the Agent Management Panel in Warp or the Oz web app Runs page.

Run records show the task’s status, conversation, and outputs so you can check what the agent actually did. Share a session with authorized teammates or send follow-up instructions when the workflow supports continuation.

See viewing cloud agent runs to inspect a transcript, share a session, or continue a standalone run.

What cloud agents are designed for

Section titled “What cloud agents are designed for”
  • Event-driven tasks - Investigate an issue, review a pull request, or examine failed-build logs when an integration or API request starts work.
  • Parallel work - Divide a larger task into bounded changes, give agents separate working copies, and coordinate their results for review.
  • Recurring maintenance - Check dependencies, generate tests, or propose documentation updates on a schedule.

Use factory automations for event-driven work and recurring maintenance in a factory. Standalone workflows have separate integrations and schedules. Multi-agent orchestration explains how parent and child agents divide work.

How cloud agents work

Section titled “How cloud agents work”

A trigger supplies the request and its context, such as an issue description or pull request diff. The agent works on its configured host with the repositories, tools, and credentials available to it, then produces results such as a branch, pull request, report, or transcript. Define the task’s scope and validation commands before starting; a completed run alone doesn’t prove the result is correct.

The Automation Platform overview explains the shared execution model in more detail.

For factory execution options, see Deployment patterns for Warp Factories.

To keep execution on your infrastructure, use a managed worker when the Automation Platform routes the run. For standalone agents started directly with oz agent run in CI, Kubernetes, or a dev environment, see Unmanaged execution.

Configure the runtime and tools

Section titled “Configure the runtime and tools”

A harness runs the coding agent and determines its runtime-specific configuration. An environment groups repositories, a container image, and setup commands; factories manage their workspace by default, with an existing environment available as an option. MCP servers connect the agent to external tools and data.

The shared references cover factory and standalone setup:

Codebase Context

Section titled “Codebase Context”

Codebase Context provides indexed repository search to help an agent locate relevant code. Its scope depends on indexing settings and the run’s harness. See Codebase Context in cloud agent runs for requirements and availability.

Centralized configuration

Section titled “Centralized configuration”

Reusable configuration keeps instructions, tools, and access settings with the workflow instead of repeating them in each request. For factory-wide and per-agent settings, use definitions as code and factory secrets. Standalone resources have their own scopes; the legacy secrets reference retains CLI and personal-secret workflows.

API access to runs

Section titled “API access to runs”

The API lets scripts, CI jobs, and internal services submit work and inspect its results without the desktop app. Use factory endpoints to send work to a factory. Use agent and run endpoints for standalone runs and to inspect run status, transcripts, and results.

Using cloud agents with or without the Warp app

Section titled “Using cloud agents with or without the Warp app”

For standalone runs outside the Warp app, use the Oz CLI, Oz web app, or API and SDKs. Session sharing and the Agent Management Panel provide inspection paths in Warp.

Use Handoff to continue work between local and cloud sessions.

Billing and plan requirements

Section titled “Billing and plan requirements”

Check access, billing, and identity requirements before starting runs or integrations. Platform usage owns the charge details, and factory infrastructure and security explains inference and credential options.

Section titled “Related pages”

Setup and configuration

Section titled “Setup and configuration”

Runs, interfaces, and access

Section titled “Runs, interfaces, and access”