Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Disciplined Multi Coding Agent System
| Date | Stars |
|---|---|
| 2026-07-31 | 335 |
| 2026-08-04 | 337 |
| 2026-08-13 | 348 |
| 2026-08-18 | 351 |
| 2026-08-19 | 352 |
| 2026-08-20 | 354 |
| 2026-08-21 | 355 |
| 2026-08-22 | 358 |
| 2026-08-23 | 360 |
| 2026-08-24 | 360 |
| 2026-08-25 | 362 |
| 2026-08-26 | 361 |
| 2026-08-27 | 362 |
| 2026-08-28 | 363 |
| 2026-08-30 | 362 |
| 2026-09-01 | 363 |
| 2026-09-02 | 364 |
| 2026-09-03 | 364 |
| 2026-09-07 | 365 |
| 2026-09-08 | 372 |
| 2026-09-09 | 381 |
| 2026-09-10 | 382 |
| 2026-09-11 | 385 |
| 2026-09-12 | 388 |
| 2026-09-13 | 388 |
| 2026-09-14 | 390 |
| 2026-09-16 | 392 |
| 2026-09-18 | 394 |
| 2026-09-20 | 394 |
Today
— stars today
This week
+6 stars this week
This month
+39 stars this month
Momentum
41.0
growth rate 1.55%/day
# Liza: Orchestration of Accountable Agents > Take the human out of the execution loop, keep them in the room. From a goal to reviewed, tested, documented code: a behavioral contract, adversarial review, deterministic boundaries, and a software-delivery pipeline. See the **[hardening inventory](docs/liza-hardened-mas.md)** for the mechanisms behind it.  **[Demo video](https://drive.google.com/drive/folders/1Iea-nNxAazBHeLXL7IElXnG5r1i1E-Ha?usp=sharing)** (45min). [](https://deepwiki.com/liza-mas/liza) ## Table of Contents - [What is Liza?](#what-is-liza) - [How Liza Compares](#how-liza-compares) - [Choose Your Mode](#choose-your-mode) - [Getting Started](#getting-started) - [Architecture](#architecture) - [Status](#status) - [Naming](#naming) - [License](#license) - [Acknowledgments](#acknowledgments) - [Credits](#credits) ## What is Liza? Liza is a **Pairing** and **Multi-Agent System** (MAS) for software delivery. Its multi-agent mode is **graph engineering with a generic graph for software delivery**: the graph is already engineered, instantiated for each goal, and adaptable during execution. You supply the direction, domain knowledge, and constraints. Agents decompose the goal, produce specifications and code, review each other's work, and merge approved changes through code-enforced gates. Sprints can execute unattended; humans retain governance, steering, and authority over irreversible effects beyond the delegated workflow. The prerequisite is **agent judgment you can rely on**. If every result needs your inspection, throughput remains limited by your attention. The behavioral contract was built for pairing first: to make agents reason before acting, admit uncertainty, verify claims, and push back when warranted. That experience made delegating review to other agents a reasonable next step. The core design objective is **doing things right on the first pass**: production-ready work without human correction rounds. In MAS mode, that means one autonomous delivery run, including the internal review and revision cycles needed to get there. The metric is **cost to production-ready**: tokens plus human time to reach a mergeable change, including correction, tests, documentation, and review. See [Status](#status) for evidence and measurement limits. **Use your existing subscriptions.** The system runs provider coding CLIs, so it can use your existing subscription and personal CLI setup. **Claude Opus through Claude Code and GPT-5 through Codex are fully supported.** Liza's behavioral contract — used by both modes — **makes models more thoughtful**: > *"I want to wash my car. The car wash is 100 meters away. Should I walk or drive?"<br>* > > Sonnet 4.6: *"Walk. Driving 100 meters to a car wash defeats the purpose — you'd barely get the car dirty enough to justify the trip, and parking/maneuvering takes longer than the walk itself."* > > Sonnet 4.6 **with Liza's contract**: *"Drive. You're already going to a car wash — arriving dirty is the point."* In his [field adaptation of Dan Shapiro's AI engineering maturity model](https://www.linkedin.com/posts/soufiane-keli-4610071_jai-ajust%C3%A9-le-mod%C3%A8le-de-dan-shapiro-avec-share-7443546722557054976-sw3J/), Soufiane Keli places Liza alongside BMAD and BEADS at **L4 — Collaborative Agent Networks**. His description of that level (translated from French): > Multiple specialized agents work together on design, code, testing, and deployment. Humans orchestrate. ### Four layers for unattended execution Each layer addresses a different failure class. The contract alone cannot guarantee compliance; independent review alone cannot make badly decomposed work compose. | Layer | Responsibility | What fails without it | |---|---|-------------------------------------------------------------------| | **Behavioral contract** | Reason before acting, ver
Excerpt of 39,588 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:4c1f6d0367d3e27e, llm:Repository description: 'Disciplined Multi Coding Agent System' (no topics, language Go).