Hey dev.to 👋
Quick story: I've spent way too many hours wiring up "autonomous" agents in Python, only to realize they're basically fancy chatbots that wait for me to type something. Cool for demos. Not great for actually getting work done while I sleep.
So I went a different route.
Meet RustyHand — an open-source Agent Operating System built in Rust. Not a framework. Not a wrapper. A full OS for agents that run on schedules, monitor things, generate leads, manage socials, and report back — all without you babysitting them.
Why it's different:
- 🦀 Single ~32MB binary. One install. One command. Done.
- ⚡ Autonomous by default. Cron jobs, event triggers, recurring workflows — agents that work 24/7.
- 🔌 37 channel adapters. Telegram, Discord, Slack, WhatsApp, Matrix… your agents meet users where they are.
- 🤖 37 pre-built templates. Researcher, coder, analyst, social media manager —
rustyhand agent new researcherand go. - 🧠 26 LLM providers, 130+ models. Auto-fallback, cost tracking, per-agent budgets.
- 🔐 16 security layers. WASM sandboxing, taint tracking, signed manifests, zeroized secrets — because agents shouldn't become attack vectors.
- 🧩 MCP server built-in. Let Claude, Cursor, or any MCP-compatible AI control your agent OS.
60 seconds to running:
curl -fsSL https://raw.githubusercontent.com/ginkida/rustyhand/main/scripts/install.sh | sh
rustyhand init
rustyhand start
# Dashboard at http://localhost:4200
Real talk:
This is v0.1.0. It's feature-complete, battle-tested internally (3,000+ tests, zero clippy warnings), but yes — you might hit a rough edge. Pin a commit for prod, report issues, help shape v1.0.
If you've ever wanted agents that do instead of just talk, I'd love for you to kick the tires:
Drop a comment if you try it out — what's the first autonomous agent you'd build? 🛠️
(P.S. It's MIT licensed. Use it, break it, rebuild it. Just maybe send a PR if you fix something.)
Top comments (0)