Zerostack is a minimal coding agent written in Rust, focusing on low memory usage and high performance. It aims to provide an efficient local alternative for agentic coding tasks. The project highlights the use of Rust for building performant AI tools.
A starter prompt for Claude Code, what you'll need, and how to reach them.
You are an expert Rust developer. Build a minimal coding agent in Rust, optimized for memory footprint and performance, similar in concept to gi-dellav/zerostack but leveraging modern cloud-native practices. Use Axum for the web server, integrate with the Anthropic Claude API for LLM interactions (via `reqwest` or a dedicated Rust SDK if available and idiomatic), and structure it as a microservice. The agent's core loop should accept a coding task, generate a plan, execute code (simulated via file writes/reads and optional shell commands for a basic proof-of-concept), and refine its output. Focus on making the codebase clean, idiomatic Rust, with excellent error handling.
MVP scope: Implement a `/task` endpoint that accepts a JSON payload with `{"prompt": "Implement a simple `grep` like utility in Rust."}`. The agent should internally parse this, generate a `plan.md` file, and then generate a `main.rs` file for the requested utility, returning the contents of `main.rs` as the response. Do not implement actual compilation or execution initially, just file generation.
Build/verify gate: Provide a Dockerfile to build and run the application. The API should be callable via `curl` and successfully generate `plan.md` and `main.rs` files within the container's working directory or a mounted volume.Reach Rust developers and AI engineers who are building agents and AI tools by showing how Zerostack's performance benefits can be integrated or provide a blueprint for their own projects, especially those concerned with resource usage.
Minimal coding agent written in Rust, optimized for memory footprint and performance Topics: agent, agentic-ai, agentic-coding, agents, ai, claude-code, coding-agent, llm, rust, rust-lang.
Open a new issue on the gi-dellav/zerostack GitHub repository, or if they have an email in their profile, email them directly.
“I'm building a performant Rust-based coding agent inspired by Zerostack and wanted to share my approach focusing on cloud-native deployment and specific LLM integrations. Would love to get your thoughts or explore potential collaborations.”
Open the original ↗