# gi-dellav/zerostack

Lightweight coding agent written in Rust, optimized for memory footprint and performance

Repository: https://github.com/gi-dellav/zerostack
Canonical: https://ross.abutalabs.com/products/zerostack
Homepage: https://gi-dellav.github.io/zerostack/
Language: Rust
License: GPL-3.0
License Family: copyleft
Topics: agent, agentic-ai, agentic-coding, agents, ai, claude-code, coding-agent, llm, rust, rust-lang
Last push: 2026-08-24T06:42:45+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 94, longevity 8
- inputs: {"age_days": 113, "days_push": 9, "days_rel": 40, "gap_med": 0, "n_releases_24m": 36}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1600, forks 125 (observed 2026-08-28T04:05:09.504278+00:00)

## What it is
zerostack is a minimal coding agent written in Rust, optimized for a tiny memory footprint (~16 MB RAM) and fast performance. It provides a terminal UI, multi-provider LLM support (OpenRouter, OpenAI, Anthropic, Gemini, Ollama), standard coding-agent tools, subagents, MCP support, persistent Markdown memory, and a configurable permission system.

## Use cases
- run an AI coding agent in the terminal with low memory usage
- refactor and edit codebases with an LLM agent from the CLI
- use Claude or GPT models to autonomously implement features
- run parallel coding agents on the same repo with git worktrees
- connect a coding agent to editors like Zed via ACP
- give an agent persistent memory across coding sessions
- explore a codebase quickly with parallel subagents
- replace heavier JS-based coding agents like Claude Code or opencode

## When to choose
- you want a lightweight, fast coding agent with minimal RAM and CPU usage
- you prefer a Rust-based terminal tool over Node.js-based agents
- you need multi-provider LLM support including local models via Ollama
- you want built-in subagents, worktrees, memory, and MCP extensibility
- you need a configurable permission system for agent tool use

## When to avoid
- you need a GUI or IDE-native experience rather than a terminal UI
- you rely on a rich plugin/skill ecosystem built around Claude Code or opencode
- you require GPL-3.0-incompatible licensing for your project
- you need a very mature, battle-tested agent with a large community

## Facets
- artifact type: cli-tool
- maturity: active
- function: agent-framework, llm-inference, cli, terminal-ui, mcp, chatbot
- domain: developer-tools, large-language-models
- platform: windows, cli, rust
- tags: coding-agent, agentic-coding, multi-provider-llm, claude-code-alternative, opencode-inspired, subagents, persistent-memory, git-worktrees, sandboxing, acp, ai-agents, command-line, automation, linux, macos

## Member repositories
- gi-dellav/zerostack (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:09.504278+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:53:11.621946+00:00, confidence not recorded.
  - readme: https://github.com/gi-dellav/zerostack (fetched 2026-08-28T04:05:09.504278+00:00, sha df3103b84ec5)
  - homepage: https://gi-dellav.github.io/zerostack/ (fetched 2026-08-29T11:24:23.953115+00:00, sha f28e4b63ae03)
  - registry_crates: https://crates.io/api/v1/crates/zerostack (fetched 2026-08-29T11:24:23.962700+00:00, sha 4562f078efa9)
- Data as of 2026-08-30T08:39:29.467469+00:00.