Skip to main content

AgentHub Docs

AgentHub is a single-binary control plane for long-lived AI agents, Team workflows, and optional remote execution nodes.

What AgentHub Gives You​

Most agent tools are optimized for one terminal and one short-lived session. AgentHub is built for the operational side of agent work:

  • keep agent sessions alive after the browser tab closes
  • inspect structured ACP history instead of relying on raw scrollback
  • manage Team workflows with explicit shared conversation and task surfaces
  • route execution to remote nodes with actor p2p control and mailbox relay when one machine is not enough
  • keep runtime state, history, and operator control in one place

Why This Matters​

AgentHub is useful when you want the browser to be an operator surface rather than the runtime itself.

That changes a few things:

  • reconnecting to a session should be normal, not a recovery edge case
  • output should stay structured and reviewable after the original run ends
  • multi-agent collaboration should use explicit coordination surfaces instead of ad-hoc shell copy/paste
  • remote execution should preserve the same actor and mailbox model instead of introducing a separate control plane

Core Surfaces​

AgentHub has four primary user-facing surfaces:

  • Agents: create, start, stop, reconnect, and inspect single-agent runs
  • Conversation and ACP timelines: review plans, tool calls, command output, and debug events as structured history
  • Teams: coordinate leader/worker collaboration with shared conversation and task tracking
  • Agent Nodes and actor p2p: extend execution onto remote machines without moving the control plane, while keeping mailbox/control traffic on the same operational model

Choose Your Path​

Evaluate AgentHub​

  1. Product Overview
  2. Feature Overview
  3. Architecture Overview

Start Running Locally​

  1. Installation and Startup
  2. Configuration Basics
  3. Login and Access
  4. First Task Walkthrough
  5. Create Your First Agent

Operate Teams Or Distributed Execution​

  1. Team Workbench
  2. Agent Nodes and Remote Execution
  3. Deployment Overview and Topology
  4. Production Checklist
  5. Troubleshooting

Documentation Map​

  • Overview: what AgentHub is, which surfaces matter, and how the system is structured.
  • Getting Started: install, configure, log in, and complete one first task.
  • Core Workflow: create agents, choose workdirs/worktrees, run tasks, inspect output, and review changes.
  • Advanced Usage: Team workflows, OpenAPI-based automation, and connection recovery.
  • Deployment and Operations: rollout, production checks, security, and troubleshooting.

After The Basics​