Yang1999code

Yang1999code / controllable-agent

Public

20-interface AI agent framework with 3-layer architecture, multi-agent delegation, and self-evolution.

14
0
100% credibility
Found May 05, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

An open-source framework for multi-agent AI systems that collaborate on coding tasks with planning, execution, review, memory, and a terminal interface.

How It Works

1
🔍 Discover the Helper

You hear about a smart AI team that can build apps and code for you, like having a group of expert helpers at your desk.

2
📥 Get It Ready

Download the program and follow easy steps to set it up on your computer, just like installing any app.

3
🧠 Connect the AI Brain

Link it to a smart AI service so your helpers can think and create, taking just a moment to enter your access details.

4
🚀 Launch Your Assistant

Open the chat window in your terminal and say hello – it greets you with its model name and ready tools.

5
Pick Your Adventure
💬
Quick Chat

Ask one helper for advice or small tasks, like explaining code.

👥
Team Mode

Tell the team a big goal, like 'build a login system,' and watch them plan, code, check, and learn together.

6
✨ Watch the Magic

See the agents divide work – one plans, others code and review in real-time, with colorful updates and a live panel.

🎉 Task Mastered

Your project is built, tested, and saved with lessons learned for next time – complex work done effortlessly!

Sign up to see the full architecture

5 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 controllable-agent?

Controllable-agent is a Python framework for building controllable multi-agent systems with a 3-layer architecture, enabling AI agents to handle complex tasks like coding full apps through delegation and self-evolution. You give it a goal like "build a user auth system," and it spins up specialist agents—planner, coder, reviewer, coordinator, memorizer—that collaborate, review code in pairs, run tests, and extract reusable skills into a wiki-style memory store. Users interact via a real-time TUI CLI with commands like /多智能体 for multi-agent mode, interruptions for steering, and tools for file ops, bash, web browsing.

Why is it gaining traction?

It stands out with controllable delegation where agents self-organize without hard limits, plus a structured memory that evolves from task digests to wiki pages, reducing hallucinations on repeats. The 20-interface design and built-in TUI with flowcharts make debugging agent flows intuitive, unlike single-model agents that forget mid-task. Developers hook on the parallel coder-reviewer pairs and automatic skill crystallization for faster iterations.

Who should use this?

Backend devs prototyping full-stack features like APIs or auth modules, where manual coding-review cycles slow you down. AI tinkerers experimenting with controllable multi-agent setups for motion prediction or RAG agents. Solo coders automating tests and refactoring via bash/web tools in a shared memory space.

Verdict

Promising early framework for self-evolving multi-agent work, with strong docs, 449 passing tests, and editable YAML configs—but at 14 stars and 1.0% credibility, it's alpha-stage; fork and contribute if you need production polish. Try the CLI for Python agent delegation today.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.