Zerostack 1.0.0: A Unix-Inspired Coding Agent in Pure Rust
Zerostack 1.0.0 has been published on crates.io. It's a coding agent built entirely in Rust, drawing inspiration from Unix design principles. Instead of a monolithic AI agent, Zerostack provides a set of composable CLI tools that communicate through stdin/stdout pipes, allowing developers to build custom workflows.
Key Features
- Unix philosophy: tools that do one thing and combine via pipes
- Pure Rust implementation — no runtime dependencies outside Rust ecosystem
- Small, focused commands for file editing, code search, shell execution, and LLM interaction
Usage Example
# Generate code with LLM and write to file
zerostack-ask "write a Rust function that sums a list" | zerostack-write src/sum.rsThe package is early-stage (v1.0.0) and aims to be a foundation for agentic workflows. Developers can extend it by chaining its tools with standard Unix utilities.
Who It's For
Rust developers who prefer modular, auditable AI tooling over black-box agents.
📖 Read the full source: HN AI Agents
👀 See Also
AutoSkillUpdate: A Claude Code Plugin That Detects Outdated Skills
AutoSkillUpdate is an open-source Claude Code plugin that scans your codebase, compares it against existing skills, and identifies drift. It provides drift reports with file paths and line references, then offers to rewrite outdated skills with user confirmation.
Alfred Beta Launches: Simplified OpenClaw Alternative for Non-Technical Users
Alfred is a new beta tool that provides approximately 70% of OpenClaw's functionality with significantly reduced complexity, featuring simple defaults for app connections, memory, usage modes, and infrastructure while allowing customization.
Relay: Open-Source Control Plane for OpenClaw AI Agents
Relay is an Electron desktop app that provides Claude Cowork-like workflow for OpenClaw, running on your infrastructure with your choice of LLM models and built-in governance features including approval gates and exportable audit trails.
AVP Protocol Enables LLM Agents to Share KV-Cache Instead of Text for Token Efficiency
AVP (Agent Vector Protocol) allows LLM agents to pass KV-cache directly between them instead of text, reducing token processing by 73-78% and achieving 2-4x speedups across Qwen, Llama, and DeepSeek models. The protocol works with HuggingFace and vLLM connectors and is available as a Python package.