Warp Agents: Run AI Coding Agents in Your Terminal | Warp

Introducing Warp Factories

open, flexible infrastructure for building your own cloud software factory. Learn More

Agents

Where agents meet the terminal

Run a team of agents — from Warp's SOTA agent to Claude Code, Codex, and Gemini CLI — all from the terminal.

[Download

Run Claude Code agents directly in Warp with shared context and fast iteration.

Invoke Codex from the terminal and keep edits, plans, and files in one place.

Use OpenCode to iterate on tasks with review-ready output and quick follow-ups.

Run Gemini CLI with Warp's agent tooling and keep everything in the terminal.

Trusted by over 800,000 developers and thousands of engineering teams at leading companies

Turn the terminal into an agentic development environment

Vertical tabs

Display terminal sessions as vertical tabs and configure metadata like git branch, worktrees, and pull requests.

Notifications

Agents notify you when they need attention — approving a command, reviewing a plan, or confirming they've finished a task.

Interactive code review

Take agent work from 80% to 100%. Review changes, leave comments, and send them back to agents with one click.

Run, manage, and orchestrate agents at scale

Everything you need to build with a team of agents.

Agents work where you do

Interact with cloud agents through Warp Terminal, Web, or Mobile. Trigger them from Slack, Linear, GitHub, or directly via CLI, API, or SDK.

Manage your agent workforce

Claude Code, Codex, or Warp Agent run in sandboxed environments hosted by Warp, or your own infrastructure. Share and observe sessions across your team, and steer your entire agent workforce.

Orchestrate fleets of agents

Move beyond interactive chats. Schedule and delegate tasks to agents from single-agent automations to complex multi-agent orchestration managed by Warp Agent.

Shared context for all your agents

Store and share agent context including conversation transcripts, artifacts, skills, rules, and Agent Memory. Context that compounds across sessions, agents, and harnesses.

Fully Programmable Agent Stack

One control plane, agent execution layer, and data plane for running and managing agents with shared context, permissions, observability, orchestration, and memory built in.