Hermes Squad
Multi-agent AI orchestrator that runs parallel coding agents in isolated sessions with self-improving intelligence, exposed via an MCP server for task execution and management.
README
<p align="center"> <img src="docs/assets/hermes-squad-logo.png" alt="Hermes Squad" width="200" /> </p>
<h1 align="center">Hermes Squad</h1>
<p align="center"> <strong>Multi-agent AI orchestrator combining Claude Squad's session management with Hermes Agent's self-improving intelligence.</strong> </p>
<p align="center"> <a href="https://github.com/barnsl/hermes-squad/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/barnsl/hermes-squad/ci.yml?branch=main&style=flat-square&logo=github&label=build" alt="Build Status" /></a> <a href="https://github.com/barnsl/hermes-squad/releases/latest"><img src="https://img.shields.io/github/v/release/barnsl/hermes-squad?style=flat-square&logo=semanticrelease&color=blue&label=version" alt="Version" /></a> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" alt="License: MIT" /></a> <a href="https://github.com/barnsl/hermes-squad/stargazers"><img src="https://img.shields.io/github/stars/barnsl/hermes-squad?style=flat-square&logo=github" alt="Stars" /></a> <a href="https://github.com/barnsl/hermes-squad/issues"><img src="https://img.shields.io/github/issues/barnsl/hermes-squad?style=flat-square" alt="Issues" /></a> <a href="https://discord.gg/hermes-squad"><img src="https://img.shields.io/discord/1234567890?style=flat-square&logo=discord&label=discord" alt="Discord" /></a> </p>
<p align="center"> <a href="#-quick-start">Quick Start</a> • <a href="#-features">Features</a> • <a href="#-supported-agents">Agents</a> • <a href="#-integrations">Integrations</a> • <a href="docs/README.md">Documentation</a> • <a href="CONTRIBUTING.md">Contributing</a> </p>
🧬 What is Hermes Squad?
Hermes Squad is a multi-agent AI orchestrator combining Claude Squad's session management with Hermes Agent's self-improving intelligence. It integrates natively with Amazon Quick and Kiro via ACP/MCP.
Run multiple AI coding agents in parallel, each in isolated tmux sessions with full git worktree support, while a meta-orchestration layer learns from outcomes, routes tasks intelligently, and continuously improves agent performance.
┌─────────────────────────────────────────────────────────┐
│ HERMES SQUAD │
├─────────────┬─────────────┬─────────────┬──────────────┤
│ Claude Code│ Kiro CLI │ Gemini CLI │ Hermes Agent │
│ ┌───────┐ │ ┌───────┐ │ ┌───────┐ │ ┌───────┐ │
│ │ tmux │ │ │ tmux │ │ │ tmux │ │ │ tmux │ │
│ │session│ │ │session│ │ │session│ │ │session│ │
│ └───┬───┘ │ └───┬───┘ │ └───┬───┘ │ └───┬───┘ │
│ │ │ │ │ │ │ │ │
│ ┌───┴───┐ │ ┌───┴───┐ │ ┌───┴───┐ │ ┌───┴───┐ │
│ │ git │ │ │ git │ │ │ git │ │ │ git │ │
│ │worktree│ │ │worktree│ │ │worktree│ │ │worktree│ │
│ └───────┘ │ └───────┘ │ └───────┘ │ └───────┘ │
├─────────────┴─────────────┴─────────────┴──────────────┤
│ 🧠 Self-Improving Intelligence Layer │
│ (outcome tracking · skill graphs · routing) │
├─────────────────────────────────────────────────────────┤
│ 🔌 ACP/MCP Integration (Quick · Kiro · IDE) │
└─────────────────────────────────────────────────────────┘
✨ Features
| Feature | Description | |
|---|---|---|
| 🧠 | Self-Improving Intelligence | Learns from task outcomes, builds skill graphs, improves routing over time |
| 🪟 | Tmux Session Management | Each agent runs in an isolated tmux session with full terminal access |
| 🌳 | Git Worktree Isolation | Parallel agents work on separate branches without conflicts |
| 🎯 | Intelligent Task Routing | Automatically assigns tasks to the best-suited agent based on history |
| 🔄 | Hot-Swap Agents | Switch between agents mid-task without losing context |
| 📡 | ACP/MCP Native | First-class integration with Amazon Quick Desktop, Kiro IDE, and MCP servers |
| 🎨 | Rich TUI | Beautiful terminal interface with real-time agent status and output |
| 📊 | Performance Analytics | Track agent success rates, completion times, and cost metrics |
| 🔒 | Sandboxed Execution | Agents run in isolated environments with configurable permissions |
| 🧩 | Plugin Architecture | Extend with custom agents, routers, and integrations |
| ⚡ | Parallel Execution | Run multiple agents simultaneously on different tasks |
| 🔁 | Auto-Recovery | Detects failures and automatically retries or re-routes to another agent |
🚀 Quick Start
One-Line Install
curl -fsSL https://hermes-squad.dev/install.sh | bash
Homebrew
brew tap barnsl/hermes-squad
brew install hermes-squad
Manual Install
git clone https://github.com/barnsl/hermes-squad.git
cd hermes-squad
make install
Verify Installation
hermes-squad --version
# Hermes Squad v0.1.0
First Run
# Start the TUI
hermes-squad
# Or launch with a specific task
hermes-squad run "Refactor the auth module to use JWT tokens"
# Launch multiple agents in parallel
hermes-squad parallel \
--agent claude-code "Write unit tests for auth" \
--agent kiro "Update API documentation" \
--agent gemini "Optimize database queries"
📸 Demo
<p align="center"> <img src="docs/assets/demo.gif" alt="Hermes Squad Demo" width="800" /> </p>
Screenshot: Three agents working in parallel — Claude Code writing tests, Kiro updating docs, and Gemini optimizing queries.
<details> <summary>📹 More Screenshots</summary>
| TUI Overview | Agent Detail | Performance Dashboard |
|---|---|---|
![]() |
![]() |
![]() |
</details>
⚙️ Configuration
Hermes Squad uses a layered configuration system:
~/.config/hermes-squad/config.toml # Global config
.hermes-squad/config.toml # Project-level config
.hermes-squad/agents.toml # Agent definitions
.hermes-squad/skills.toml # Learned skill graphs
Minimal Configuration
# ~/.config/hermes-squad/config.toml
[general]
default_agent = "claude-code"
parallel_limit = 4
auto_commit = true
[intelligence]
learning_enabled = true
skill_graph_path = "~/.config/hermes-squad/skills.db"
routing_strategy = "performance" # "performance" | "cost" | "round-robin" | "manual"
[integrations.quick]
enabled = true
acp_endpoint = "localhost:7862"
[integrations.kiro]
enabled = true
acp_endpoint = "localhost:7863"
[tui]
theme = "dark"
show_metrics = true
split_view = "horizontal"
Agent Configuration
# .hermes-squad/agents.toml
[[agent]]
name = "claude-code"
binary = "claude"
args = ["--dangerously-skip-permissions"]
max_concurrent = 2
cost_weight = 0.8
specialties = ["refactoring", "testing", "architecture"]
[[agent]]
name = "kiro"
binary = "kiro-cli"
args = ["--agent-mode"]
max_concurrent = 1
cost_weight = 0.3
specialties = ["documentation", "specs", "planning"]
[[agent]]
name = "hermes"
binary = "hermes-agent"
args = ["--self-improve"]
max_concurrent = 1
cost_weight = 0.5
specialties = ["reasoning", "multi-step", "research"]
🤖 Supported Agents
| Agent | Status | Specialties | Notes |
|---|---|---|---|
| Claude Code | ✅ Stable | Refactoring, Testing, Architecture | Primary agent, best all-rounder |
| Kiro CLI | ✅ Stable | Documentation, Specs, Planning | Spec-driven development |
| Codex | ✅ Stable | Code generation, Completion | Fast for targeted edits |
| Gemini CLI | ✅ Stable | Analysis, Optimization, Research | Strong on large codebases |
| Hermes Agent | ✅ Stable | Reasoning, Multi-step, Self-improvement | Meta-cognition layer |
| Aider | ✅ Stable | Pair programming, Git integration | Great for iterative changes |
| OpenCode | 🧪 Beta | Terminal-native coding | Lightweight alternative |
| Amp | 🧪 Beta | Codebase search, Navigation | Excellent for exploration |
Adding Custom Agents
[[agent]]
name = "my-custom-agent"
binary = "/path/to/agent"
args = ["--flag"]
prompt_file = ".hermes-squad/prompts/custom.md"
specialties = ["domain-specific"]
🔌 Integrations
Amazon Quick Desktop (ACP)
Hermes Squad exposes an ACP server that Amazon Quick can connect to directly:
[integrations.quick]
enabled = true
acp_endpoint = "localhost:7862"
expose_tools = ["run_agent", "list_sessions", "get_status", "parallel_run"]
# Register with Quick Desktop
hermes-squad register --quick
Kiro IDE (ACP)
Native integration with Kiro's agent protocol for IDE-embedded orchestration:
[integrations.kiro]
enabled = true
acp_endpoint = "localhost:7863"
workspace_sync = true
# Register with Kiro
hermes-squad register --kiro
MCP Server
Expose Hermes Squad capabilities as an MCP server for any compatible client:
# Start MCP server
hermes-squad mcp serve --port 8080
# Or add to MCP config
{
"mcpServers": {
"hermes-squad": {
"command": "hermes-squad",
"args": ["mcp", "serve"],
"env": {
"HERMES_SQUAD_CONFIG": "~/.config/hermes-squad/config.toml"
}
}
}
}
MCP Tools Exposed
| Tool | Description |
|---|---|
hermes_run |
Execute a task with intelligent agent routing |
hermes_parallel |
Run multiple tasks in parallel across agents |
hermes_status |
Get status of all active sessions |
hermes_history |
Query task history and outcomes |
hermes_learn |
Trigger learning from recent outcomes |
hermes_config |
View/update configuration |
📚 Documentation
| Document | Description |
|---|---|
| Getting Started | Installation and first steps |
| Architecture | System design and internals |
| Agent Guide | Configuring and extending agents |
| Intelligence Layer | How self-improvement works |
| ACP Integration | Quick & Kiro integration guide |
| MCP Server | MCP server reference |
| CLI Reference | Complete CLI documentation |
| FAQ | Frequently asked questions |
| Troubleshooting | Common issues and fixes |
🏗️ Project Structure
hermes-squad/
├── src/
│ ├── core/ # Core orchestration engine
│ ├── agents/ # Agent adapters and lifecycle
│ ├── intelligence/ # Self-improving ML layer
│ ├── session/ # Tmux session management
│ ├── git/ # Git worktree operations
│ ├── tui/ # Terminal UI (Ink/React)
│ ├── integrations/ # ACP/MCP connectors
│ └── cli/ # CLI entry point
├── tests/ # Test suites
├── docs/ # Documentation
├── scripts/ # Build and release scripts
├── .hermes-squad/ # Default project config
└── config/ # Default configuration templates
🤝 Contributing
We love contributions! Please see CONTRIBUTING.md for guidelines.
# Development setup
git clone https://github.com/barnsl/hermes-squad.git
cd hermes-squad
make dev-setup
make dev # Start in development mode
make test # Run tests
make lint # Check code style
📄 License
MIT © 2026 BarnsL
🙏 Credits
Hermes Squad stands on the shoulders of giants:
- Claude Squad — Session management architecture and TUI inspiration
- Hermes Agent — Self-improving intelligence and skill graph concepts
- Amazon Quick — ACP protocol and desktop integration
- Kiro — IDE-native agent protocol and spec-driven development
<p align="center"> <sub>Built with 🧠 by <a href="https://github.com/barnsl">BarnsL</a> — orchestrating the future of AI-assisted development</sub> </p>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.


