Note

This repo is now public archive. The project is now called openwalrus/walrus.

Walrus

Run autonomous agents with built-in LLM inference. No API keys. No cloud. Just one binary.

curl -fsSL https://raw.githubusercontent.com/openwalrus/walrus/main/install.sh | sh

Or install with Cargo:

cargo install openwalrus

What It Does

• Local inference — runs LLMs on your machine (Metal on macOS, CUDA on Linux)
• Persistent memory — agents remember across sessions (SQLite + FTS5)
• Built-in tools — file I/O, shell, MCP servers, cron scheduling
• Multi-channel — talk to your agents from the terminal, Telegram, or Discord
• Skills — extend agents with Markdown prompt files, no code needed

Quick Start

# 1. Start the daemon
walrus daemon

# 2. Chat with your agent
walrus attach

Models are configured in ~/.walrus/config.toml. Point it at a local model or any OpenAI-compatible API:

[model]
model = "deepseek-r1:8b"

Cloud Providers

walrus works with OpenAI, Anthropic, DeepSeek, and other OpenAI-compatible APIs. Local inference is the default — cloud is opt-in.

License

GPL-3.0