AI agents that work like a teamFrom idea to delivery, all on one desktop

AgentHub brings Codex, Claude Code, OpenCode, and your workflows into one collaborative workspace for coding, review, docs, and complex project work.

Features

One workspace for agent-driven development

AgentHub keeps agent chat, code review, file changes, and approvals in one workbench.

@mention an agent to join the conversation. Agents can compare approaches and keep the result attached to the task.

Route each task to the agent and runtime that fit the work. You keep the final say.

Preview code changes as agents work. Accept, reject, or request revisions in the same thread.

Isolated Git worktrees for each agent. They experiment; you decide what ships.

Record tool calls, approval results, and key states so agent sessions can be reviewed later.

Claude Code, Codex, and OpenCode connect through adapters. Other CLI agents can follow the same contract.

+ more

How it works

Create a group chat and add the agents you want on the task.

Describe tasks in natural language. Agents self-organize and collaborate.

Review diffs, confirm changes, then continue through your normal PR, CI, and release flow.

Download the desktop app, or learn how the Web workspace connects through Hub and Local Edge.

Desktop client with native notifications and offline support

Web workspace connects to Desktop/Edge runtimes through Hub sessions.

Claude Code

Claude Code CLI for longer coding tasks

Codex

OpenAI Codex CLI for quick prototypes and tests

OpenCode

OpenCode CLI, an open-source coding agent