Agentrooms

Multi-agent workspace for collaborative development with Claude CLI. Route tasks to specialized agents (local or remote), coordinate complex workflows.

Current Status: This version supports one agent room. Multiple rooms support is planned for future releases - contributions welcome!

Forked from sugyan/claude-code-webui

0725.CodeByAgents.Demo1.mp4

Key Features

• @agent-name mentions: Direct execution, no orchestration overhead
• Multi-agent workflows: Automatic task decomposition and coordination
• Local + Remote agents: Mix local agents and remote machines (Mac Mini browser agent, cloud instances, etc.)
• Free orchestrator Anthropic usage: No API key required (uses my endpoint to cover your cost by default ) Sure you can bring your own API_KEY
• Custom API support: Configure your own endpoint in Settings
• Dynamic agents: Add/remove agents via web UI

API Design

• Planner: Uses API key for task analysis and coordination
• Agents: Use your local Claude CLI subscription for execution
• Default: Free public endpoint (zero setup)
• Custom: Set your API URL in Settings for private deployment

Quick Start

# Prerequisites: Claude CLI installed + authenticated

# Option 1: Desktop App
make install && make electron      # Electron desktop app

# Option 2: Web Development  
cd backend && deno task dev        # Backend: http://localhost:8080
cd frontend && npm run dev         # Frontend: http://localhost:3000

# 3. Start agent instances (local or remote)
cd path/to/agent1 && deno task dev --port 8081   # Local agent
cd path/to/agent2 && deno task dev --port 8082   # Local agent
# Remote agents: Run on other machines, expose ports

# 4. Configure agents in Settings

Usage

Single agent: @api-agent add user authentication

• Direct HTTP call to agent endpoint
• No coordination overhead

Multi-agent: "Create full auth system with frontend and backend"

• Orchestrator analyzes and creates execution plan
• Coordinates file-based communication between agents
• Manages dependencies automatically

Configuration

In Settings UI:

1. API Configuration:

   • Default: Uses free public endpoint
   • Custom: Set your API endpoint URL

2. Add Agents (local or remote):

   • Name: API Backend Agent
   • Description: Handles backend API development
   • Working Directory: /path/to/backend
   • API Endpoint: http://localhost:8081 (local) or http://mac-mini.local:8081 (remote)

3. Agent Routing:

   • First agent = orchestrator
   • @mentions route to specific agents
   • General requests use orchestrator

Architecture

Frontend → Main Backend (Orchestrator) → Local Agent 1 (localhost:8081)
                                      → Local Agent 2 (localhost:8082)  
                                      → Remote Agent 3 (mac-mini.local:8081)
                                      → Remote Agent N (cloud-instance:8081)

Single Agent Flow:

User → @agent-name → HTTP Request → Agent's Claude Instance → Response

Multi-Agent Flow:

User → General Request → Orchestrator Analysis → Execution Plan
                                                ↓
Agent 1 ← Step 1 ← File Dependencies ← Coordination Logic
Agent 2 ← Step 2 ← Read Step 1 Output  
Agent N ← Step N ← Read Previous Results

Development

# Desktop app
make electron           # Run Electron app
make dist              # Build production app  
make dmg               # Build macOS installer

# Quality checks
make check      # Format, lint, typecheck, test
make format     # Format code
make test       # Run tests

# Building
make build-backend   # Create binary
make build-frontend  # Build frontend

Contributing

• Lefthook: Pre-commit hooks ensure quality
• TypeScript: Full type safety
• HTTP APIs: RESTful agent communication
• Dynamic config: All agents configurable via UI

License

MIT License