RestFlow

AI agents work. You rest.

A high-performance AI agent runtime built in Rust — run 10+ agents in parallel

What is RestFlow?

RestFlow is a high-performance AI agent runtime built in Rust. It runs multiple autonomous agents in parallel — each agent thinks, decides, acts, and observes in a ReAct loop until your task is complete.

Key Features

Category	Features
AI Agent Core	ReAct loop, 10+ parallel background agents, subagent spawning, working + long-term memory
Built-in Tools	HTTP, Bash, File, Python (Monty sandbox), Email, Telegram, Web Search, Vision, Transcribe
LLM Providers	Anthropic Claude, OpenAI GPT, DeepSeek, Claude Code CLI, OpenAI Codex, Gemini
Security	Approval gate for bash/file ops, command chain detection, header sanitization
Integration	MCP server (port 8787), Telegram bot, scheduled cron tasks, skill system
Platform	CLI daemon, Tauri desktop app, Docker, self-hosted

Quick Start

CLI (Recommended)

# Homebrew (macOS/Linux)
brew install lhwzds/tap/restflow

# npm (Cross-platform)
npm install -g restflow-cli

# Build from source
cargo install --git https://github.com/lhwzds/restflow --package restflow-cli

Docker

docker compose up -d --build

MCP HTTP server available at http://localhost:8787/mcp

First Run

# Start RestFlow daemon
restflow start

# Configure API key
restflow secret set ANTHROPIC_API_KEY sk-ant-xxx

# Configure Telegram bot for AI chat
restflow secret set TELEGRAM_BOT_TOKEN <your-bot-token>

Tech Stack

Backend

Rust (async Tokio runtime)
RMCP (MCP HTTP transport)
Redb (Embedded database)
Monty (Sandboxed Python runtime)

Frontend

Vue 3 + TypeScript
Pinia (State Management)

AI Integration

Multi-provider LLM support (Anthropic, OpenAI, DeepSeek, Gemini)
MCP (Model Context Protocol) server
ReAct agent loop with tool execution

Documentation

CLI Reference - Command-line interface guide
Authentication - Setting up AI service credentials
GitHub Repository - Source code

License

Apache License 2.0