FAQ
Does ReevesAgents store my access keys?
No. Each coding tool uses its existing login and keeps its own keys. ReevesAgents stores task records under ~/.reeves and redacts text that looks like an access key before writing it.
Does my code leave my machine?
Yes. Each coding tool sends your code to its own AI provider. ReevesAgents adds no intermediary. Its browser interface is accessible only from your computer at 127.0.0.1. For remote access, use an encrypted connection that forwards the port (SSH).
Which coding tools does it support?
Claude Code, Codex, OpenCode, Hermes, Kimi, DeepSeek, Pi, Qwen, Aider, Gemini CLI, Cline, Crush, Goose, Kilo Code, Letta Code, Mistral Vibe, Grok Build, Kiro, Factory Droid, Qoder, Warp, Junie, GitHub Copilot CLI, Cursor CLI and Rovo Dev. ReevesAgents can start tools that are installed and signed in. reevesagents doctor lists the tools it finds. It also lists four helper tools agents can call from their shell: Perplexity, Google Colab, Hugging Face and Ollama.
What happens to my assistants when I close the interface?
They keep working. The terminal tool tmux keeps each assistant running in its own window after the interface closes.
Do I have to let one assistant control the others?
No. Assistants can work side by side. reevesagents attach <cli> gives one coding tool permission to start, read and direct the others through a standard connection between AI tools (MCP). detach removes the connection. The max_agents setting limits assistants per run. This limit applies to assistants started through MCP.
Does it run on Windows?
Yes, in Windows Subsystem for Linux (WSL), which supports the terminal tool tmux. The separate reevesagents-win package provides Agent Control directly on Windows through MCP, the standard connection between AI tools.
Facts you can check
- GitHub stars
- 107
Downloads cover 2025-04-14 to 2026-10-14. GitHub and npm figures checked on 2026-10-15; downloads count package fetches.
Last updated: