The latest articles on DEV Community by Serik Ospanov (@serik_ospanov_c84c8ee2e54). https://dev.to/serik_ospanov_c84c8ee2e54 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3859323%2F50e4054d-fcad-4111-ad7b-7c998c9bb198.png https://dev.to/serik_ospanov_c84c8ee2e54 en Serik Ospanov Fri, 03 Apr 2026 10:58:21 +0000 https://dev.to/serik_ospanov_c84c8ee2e54/i-got-tired-of-agent-frameworks-that-just-chat-so-i-built-an-os-for-agents-that-actually-do-stuff-1k9m https://dev.to/serik_ospanov_c84c8ee2e54/i-got-tired-of-agent-frameworks-that-just-chat-so-i-built-an-os-for-agents-that-actually-do-stuff-1k9m <p>Hey dev.to 👋</p> <p>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.</p> <p>So I went a different route.</p> <p><strong>Meet <a href="https://github.com/ginkida/rustyhand" rel="noopener noreferrer">RustyHand</a></strong> — an open-source <strong>Agent Operating System</strong> 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.</p> <h3> Why it's different: </h3> <ul> <li>🦀 <strong>Single ~32MB binary</strong>. One install. One command. Done.</li> <li>⚡ <strong>Autonomous by default</strong>. Cron jobs, event triggers, recurring workflows — agents that work 24/7.</li> <li>🔌 <strong>37 channel adapters</strong>. Telegram, Discord, Slack, WhatsApp, Matrix… your agents meet users where they are.</li> <li>🤖 <strong>37 pre-built templates</strong>. Researcher, coder, analyst, social media manager — <code>rustyhand agent new researcher</code> and go.</li> <li>🧠 <strong>26 LLM providers, 130+ models</strong>. Auto-fallback, cost tracking, per-agent budgets.</li> <li>🔐 <strong>16 security layers</strong>. WASM sandboxing, taint tracking, signed manifests, zeroized secrets — because agents shouldn't become attack vectors.</li> <li>🧩 <strong>MCP server built-in</strong>. Let Claude, Cursor, or any MCP-compatible AI <em>control</em> your agent OS.</li> </ul> <h3> 60 seconds to running: </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>curl <span class="nt">-fsSL</span> https://raw.githubusercontent.com/ginkida/rustyhand/main/scripts/install.sh | sh rustyhand init rustyhand start <span class="c"># Dashboard at http://localhost:4200</span> </code></pre> </div> <h3> Real talk: </h3> <p>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.</p> <p>If you've ever wanted agents that <em>do</em> instead of just <em>talk</em>, I'd love for you to kick the tires:</p> <p>👉 <a href="https://github.com/ginkida/rustyhand" rel="noopener noreferrer">GitHub: ginkida/rustyhand</a></p> <p>Drop a comment if you try it out — what's the first autonomous agent <em>you'd</em> build? 🛠️</p> <p><em>(P.S. It's MIT licensed. Use it, break it, rebuild it. Just maybe send a PR if you fix something.)</em></p> ai rust opensource agents