Show HN: Opperator – Build Claude Code–style local AI agents in your terminal (github.com)

Opperator is a new local-first framework and TUI for building, running and managing Claude-style AI agents from your terminal. It provides a background daemon that runs isolated Python agent processes (each with its own venv), a Builder agent that generates agent code from natural-language descriptions, hot reloading, lifecycle hooks, and a single terminal UI for debugging, logs and status. Secrets are stored in your system keyring, conversations and logs persist to SQLite, and TUI↔daemon communication uses Unix sockets/pipes. It ships as a macOS/Linux CLI (Python 3.8+), installs via a one-line curl script, and currently requires an Opper account (new users get $10 inference credits and Gemini Flash 2.5 is the default), though agents can call any LLM provider or local models. For the AI/ML community this lowers the heavy infra burden around agent runtimes—scaffolding, secret management, restart/debug cycles and multi-daemon deployment—so practitioners can focus on workflow logic and tool integration (file ops, webhooks, content generation, dev automation, etc.). Technical conveniences like per-agent isolation, IPC, message persistence, LLM-callable commands, async task monitoring, and cloud deploy/migration support make it practical to iterate on reproducible, private automations that access local files and models. Note trade-offs: Opperator currently ties into an Opper account and defaults to a cloud LLM, so teams prioritizing entirely offline, provider-free setups should watch planned standalone work.