Tandem
Maxmedawar/tandem
Control Claude Code, Codex and Hermes agents on many computers from one MCP-capable chat
Tandem is a bridge that lets an MCP-capable AI chat start, steer and stop coding agents running on several of your computers. Hermes Agent is one of the engines it can launch, alongside Claude Code, Codex and a terminal shell.
What Tandem does
Tandem turns an MCP-capable AI chat into a control desk for coding agents running on several computers. From a browser chat, desktop AI client or another MCP agent, you can see which computers are online and which agents they can run, list live sessions, and start Claude Code, Codex, Hermes or a terminal shell on the machine that holds the project.
You can send more work to a running agent, read its new output, interrupt a stuck turn or close a finished session. Enrolled devices stay connected and reconnect automatically. The agents remain on their own computers, and Tandem acts as the secure connection between them and the chat. Setup is meant to be done by pasting a prompt into a trusted local coding agent. Claude is the only engine enabled by default; Codex, Hermes and shell access must be requested explicitly. The README uses Tailscale Funnel for the public MCP service and tailnet-only Tailscale Serve for device traffic.
Key features
- Lists online computers, the agents they can run and every live session
- Starts Claude Code, Codex, Hermes or a terminal shell on the machine with the project
- Sends follow-up work to a running agent without restarting it
- Reads new output, interrupts stuck turns and closes finished sessions
- Multiple enrolled computers that reconnect automatically
- Agent-driven setup prompt with explicit opt-in for extra engines
When to use it
- Starting a Hermes session on a server from a browser chat
- Sending failing test output from one machine to an agent on another
- Stopping a runaway shell command on a remote machine while keeping the session open
Who it is for: Developers who run coding agents on more than one computer and want to steer all of them from a single conversation.
How it fits with Hermes Agent
Supports Hermes Agent as one of several engines Tandem can start and steer, alongside Claude Code, Codex and a terminal shell; the Hermes engine stays off until you enable it.
Requirements: Node.js 22.6+, tmux, Tailscale and at least one supported coding agent
FAQ
What is Tandem?
Tandem is a bridge that lets an MCP-capable AI chat see, start, steer and stop coding agents on all of your enrolled computers. It does not replace Claude, Codex, Hermes or your preferred AI interface.
Does Tandem work with Hermes Agent?
Yes. Hermes is one of the engines Tandem can start on a machine, next to Claude Code, Codex and a terminal shell. Claude is the only engine enabled by default, so you must ask for Hermes during setup.
What do I need to run Tandem?
The setup prompt checks for Node.js 22.6 or later, tmux, Tailscale and at least one supported coding agent. You also choose which project folders Tandem may use.
Similar multi-agent for Hermes Agent
All multi-agentOpen-source project workspace where teams and AI agents share tasks, sandboxes and reviewed results
rundiffusion RunDiffusion AgentsOpen-source control plane for running OpenClaw, Hermes, Codex, Claude, Gemini and Pi agents side by side
Fearvox Multica Ultimate WorkbenchSkill pack for coordinating Codex, Claude Code and Hermes agents on Multica with review gates
jpalmae HermesHQDocker-first web control plane for running and operating multiple Hermes Agent instances
cyborg-garden Swarm MapMultiplayer admin and orchestrator for fleets of Hermes and Letta agents with multi-tenant security
galyarderlabs Galyarder FrameworkOpen-source agentic company kit of agents, skills and review gates that also installs for Hermes
Related guides: Build agent teams with the Hermes kanban board · Run multiple Hermes agents with profiles · Connect Hermes agents on several machines with Hermes Desktop