jasonkneen

jasonkneen / lazar

Public

A self-evolving agent in rust with one tool that only uses skills

14
2
100% credibility
Found Apr 30, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Lazar is a compact macOS tool that runs an AI agent capable of executing safe commands, storing memories, and developing its own skills through simple file-based evolution.

How It Works

1
🔍 Discover lazar

You stumble upon lazar, a simple tool that lets an AI helper grow and learn new abilities on its own.

2
💻 Set it up on your Mac

Download the files to a folder on your Mac and run the quick setup to get everything ready.

3
🧠 Connect a smart AI

Link it to a thinking service like Claude so your helper can understand and respond to you.

4
💬 Start your first chat

Type a fun question like 'what skills do you have?' and see it reply right away in your terminal.

5
✨ Watch it build abilities

Ask it to create new tricks or remember notes, and it saves them to grow smarter over time.

🚀 Enjoy your evolving buddy

Now you have a personal AI companion that learns from every conversation and handles tasks on its own.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is lazar?

Lazar is a Rust-built CLI agent for macOS that prompts Claude via Anthropic's API to run sandboxed bash commands through a single `execute(command)` tool. It persists state in on-disk folders for skills (its capabilities as markdown files), memory notes, workspace projects, and an append-only event log, letting the agent evolve itself by writing new skills on demand. Developers get a prompt-driven shell assistant—`lazar -p "build a skill for X"`—that starts minimal but grows emergent behaviors without touching its locked core.

Why is it gaining traction?

In the self-evolving agent GitHub space, lazar stands out for its radical minimalism: one immutable kernel handles prompting and sandboxing, while everything else—context loading, TUI building, log hygiene—emerges from skills the agent authors. Portable skills work across setups via env vars, and recursion caps at depth 5 with API key isolation by default. Devs hook on the "watch it self-improve" loop, like seeding a chat REPL via `lazar-chat.sh` or resetting with `--reset-all`.

Who should use this?

AI experimenters and Rust devs on macOS building personal agents for shell automation, like scripting workspace tasks or evolving custom tools from prompts. Ideal for those prototyping self-evolving agents inspired by OpenAI cookbooks but wanting disk-persisted, sandboxed execution over cloud-heavy alternatives. Skip if you're on Linux or need multi-model support.

Verdict

At 14 stars and 1.0% credibility, lazar is an early, rough prototype—docs are solid via README, but expect macOS-only quirks and manual setup via `setup.sh`. Worth a spin for self-evolving agent tinkerers; clone, set your Anthropic key, and prompt it to life if minimalism excites you.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.