The Forge

The Forge is AgentMux’s agent configuration manager. It’s where you create, configure, and manage your AI agents.

Opening The Forge

Click the hammer icon in the top bar
Or press Cmd+Shift+F / Alt+Shift+F
Or right-click any pane header and select Forge from the widget menu

Views

The Forge has three main views:

Agent List

The default view shows all configured agents, grouped into sections:

Host Agents — Run directly on your machine
Container Agents — Run in Docker containers
Custom Agents — User-defined configurations

Each agent card shows the agent name, type, and status. Click a card to open the detail view, or click + New Agent to create one.

Create / Edit Form

The agent form has the following fields:

Field	Description
Name	Human-readable agent name
Agent Type	`host` or `container`
Provider	Claude Code, Codex CLI, or Gemini CLI
Command	Launch command (auto-filled from provider)
Working Directory	Project directory the agent operates in
Connection	Remote connection (SSH host) or local

Detail View

The detail view for a specific agent has three sections:

Content

Four tabs for configuring the agent’s context:

Tab	Purpose
Soul	System prompt — defines the agent’s personality, priorities, and behavior
Instructions	Project-specific instructions loaded into context on launch (like `CLAUDE.md`)
MCP	Model Context Protocol server configuration (JSON)
Env	Environment variables passed to the agent process

Content is persisted per-agent and saved to the backend automatically.

Skills

Reusable capabilities attached to an agent:

Skill Type	Description
`prompt`	A reusable prompt template
`command`	A shell command or script
`workflow`	A multi-step automation sequence
`mcp-tool`	An MCP tool configuration

Create, edit, and delete skills from this tab. Skills are stored per-agent.

History

Session history for the agent. Shows past sessions with:

Session ID and timestamp
Model used
Message count and token usage
Working directory and git branch
First user message (preview)

Provider Configuration

The Forge pre-configures launch commands for each provider:

Provider	Default Command
Claude Code	`claude --output-format stream-json`
Codex CLI	`codex --full-auto`
Gemini CLI	`gemini --yolo`

You can customize these commands in the agent edit form.

MCP Server Configuration

Configure MCP servers per-agent in the MCP content tab using JSON:

{
  "filesystem": {
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/dir"]
  },
  "github": {
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-github"],
    "env": {
      "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..."
    }
  }
}

MCP servers configured here override the global MCP settings for this specific agent.

Import from Claw

If you manage container agents with Claw, click Import from Claw to pull existing agent configurations into The Forge. This imports:

Agent name and identity
Working directory
Environment variables
MCP server configuration