saigontechnology/AgentCrew

View on GitHub ↗

Chat application with multi-agents system supports multi-models and MCP

215 ★41 forksPythonUpdated 20d ago

What you need to know

Build a crew of focused AI specialists that code, research, architect, review and automate — used from a desktop app, terminal, one-shot jobs, A2A server or HTTP API, with an @AgentName handoff syntax and per-agent tools.

Install

curl -LsSf https://agentcrew.dev/install.sh | bash   # macOS/Linux
powershell -ExecutionPolicy ByPass -c "irm https://agentcrew.dev/install.ps1 | iex"   # Windows
pip install agentcrew-ai   # any platform

Usage

  • agentcrew chat # desktop GUI
  • agentcrew chat --console # terminal mode
  • agentcrew job --agent 'CodeAssistant' 'Review for security issues' ./src/**/*.py # CI/batch
  • @Architect / @Coding / @Reviewer hand off subtasks between agents in one conversation

Key features

  • Multi-agent orchestration with the @AgentName handoff syntax and full context carry-over
  • 11+ provider options including subscription-based (OpenCode Go, GitHub Copilot, ChatGPT Codex) and fully offline custom models (Ollama)
  • Per-agent tool enablement: code analysis, file editing, web search, browser automation, command execution, memory and MCP tools
  • Speaks Console/GUI, A2A (HTTP+JSON-RPC) and ACP (WebSocket) protocols; multi-instance agent networks

Best for

Teams that want a multi-specialist agent crew for feature development, research and automation from a single interface.

Platforms: macOS · Linux · WindowsClients: Desktop GUI · Terminal · HTTP API

Reviewed 2026-08-11

Topics

ai-agentsai-assistantanthropic-claudemcp-clientmodel-context-protocolmulti-agent-systemsterminal-apptool-use
Stars
215★
Forks
41
Language
Python
License
Apache-2.0
Created
2025-02-26
Last push
2026-08-15