DocsGetting StartedQuick Start

Quick Start

Three commands to a running AI assistant.

1. Initialize

ryvos init

The interactive wizard asks for your LLM provider and API key. For non-interactive setup:

ryvos init -y                                    # Defaults (Anthropic)
ryvos init --provider openai --model gpt-4o      # Specific provider
ryvos init --provider ollama --model llama3.1:8b  # Local LLM

This creates ~/.ryvos/config.toml (or ryvos.toml in the current directory).

2. Start Talking

ryvos

This opens an interactive REPL. Type natural language and Ryvos will plan, use tools, and respond:

You: list all rust files in this project
Ryvos: I'll use the glob tool to find Rust files.

[Tool: glob] pattern="**/*.rs"

Found 47 Rust source files:
  src/main.rs
  crates/ryvos-core/src/lib.rs
  ...

3. Try the Other Modes

ryvos run "summarize this project"   # Single prompt, then exit
ryvos tui                             # Full terminal UI (ratatui)
ryvos serve                           # Web UI at http://localhost:18789
ryvos daemon                          # Always-on with Telegram/Discord/Slack
ryvos daemon --gateway                # Always-on + Web UI

Watch Security in Action

Try asking Ryvos to do something risky:

You: delete all log files

Ryvos classifies this as a T3 action (shell command with destructive pattern) and asks for approval:

⚠ Approval required for bash (tier: T3)
  Command: rm *.log
  [approve / deny]

Type approve or deny. If you don't respond within 60 seconds, the request is automatically denied.

Try something truly dangerous:

You: run rm -rf /

✕ Tool blocked: bash (tier: T4)
  Dangerous pattern detected: recursive delete
  This action is always denied.

T4 actions are never executed, regardless of approval.

System Health Check

ryvos doctor

Checks your LLM connection, Docker availability, MCP servers, and configuration validity.

Next Steps

InstallationConfiguration