The Bridge

What does connecting Claude Desktop to Claude Code actually do?

Claude Desktop is the conversational interface where you think and plan. Claude Code is the execution layer that reads, writes, and runs code. Connecting them through an MCP server lets Claude Desktop spawn, monitor, and coordinate multiple Claude Code processes as autonomous agents.

Without an MCP bridge, you can only use one Claude Code session at a time, manually switching context between tasks. With chAIrman's MCP server, Claude Desktop becomes a CEO that hires agents, assigns tasks with dependency chains, tracks costs, and auto-retries failures — all through natural language.

Claude Desktop sends structured commands through 54 MCP tools
chAIrman translates those commands into Claude Code agent processes
Each agent works autonomously in its own directory with its own file boundaries
Results, costs, and handoff notes flow back to Claude Desktop in real time

// Your MCP config (~/.claude/mcp.json) { "mcpServers": { "chairman": { "command": "chairman", "args": [], "env": {} } } } // That's it. Restart Claude Desktop. // 54 tools appear instantly. // Start talking: // "Hire a frontend lead and // a backend lead for my-app"

Setup Steps

How to connect Claude Desktop to Claude Code

Four steps from zero to multi-agent orchestration.

1

Install chAIrman

Download the native binary for your platform. chAIrman ships as a single executable — no Node.js, no npm, no dependencies. Run the install script or download directly from your account page after subscribing.

2

Add MCP Configuration

Open ~/.claude/mcp.json and add chAIrman as an MCP server. The config is three lines: the server name, the command to run it, and an empty args array. No API keys needed in the config — authentication is handled via your chAIrman session.

3

Restart Claude Desktop

Quit and reopen Claude Desktop. It discovers the MCP server on startup and loads all 54 tools. You can verify by asking Claude: "What chAIrman tools are available?" It will list every tool with parameters.

4

Start Orchestrating

Tell Claude Desktop what you want to build. It will create a backlog with milestones and tickets, hire specialized agents, assign tasks with file boundaries and success criteria, monitor progress, and report back. You're the chairman — Claude is your CEO.

Capabilities

What the connection unlocks

Everything Claude Desktop can do once chAIrman is connected.

Capability	Without MCP	With chAIrman MCP
Claude Code sessions	One at a time, manually	Unlimited concurrent agents
Task coordination	Copy-paste between chats	Dependency pipelines with auto-ordering
Error recovery	Start over from scratch	Auto-retry with handoff inheritance
Domain knowledge	Whatever's in the prompt	857 skills auto-injected via matching
Cost tracking	Check billing page manually	Per-agent, per-task, per-project budgets
Monitoring	No visibility after launch	Real-time dashboard with WebSocket updates
Project memory	Lost between sessions	Alumni archives, handoffs, CEO notes

54

MCP Tools

857

Auto-Injected Skills

<5min

Setup Time

10

Agent Templates

Common Questions

Claude Desktop + Claude Code FAQ

What is MCP and why do I need it?

MCP (Model Context Protocol) is a standard that lets AI applications like Claude Desktop communicate with external tools and services. chAIrman is an MCP server that exposes 54 tools for orchestrating Claude Code agents. Without MCP, Claude Desktop has no way to spawn or manage Claude Code processes.

Do I need Claude Code installed separately?

Yes. Claude Code must be installed and accessible in your PATH. chAIrman spawns Claude Code as child processes — it does not bundle its own copy. You need both Claude Desktop (the interface) and Claude Code (the execution engine) installed on the same machine.

How many agents can run at once?

The default limit is 10 concurrent agents, configurable via the CHAIRMAN_MAX_AGENTS environment variable. chAIrman's smart scheduler calculates optimal concurrency based on your CPU cores and available memory. The practical ceiling is your hardware and Anthropic API rate limits.

Does this work on Windows?

chAIrman currently supports macOS (ARM and Intel) and Linux (x64). Windows support is not yet available, but you can run it in WSL2 (Windows Subsystem for Linux) with the Linux binary.

Can I use this with other AI models?

chAIrman is built specifically for Claude. It uses Claude Desktop as the command interface and Claude Code as the agent execution layer. Agent tasks can be assigned to different Claude model tiers — Opus for complex architecture work, Sonnet for standard development, Haiku for quick tasks.

What happens if Claude Desktop disconnects?

Running agents continue working independently. The MCP server persists state to disk (state.json) and saves agent handoffs every 60 seconds. When you reconnect, all agent progress, costs, and outputs are available. No work is lost.

Connect Claude Desktop to Claude Code: The Complete Guide