Documentation / Getting started
What is Yuli
A complete framework for developing with AI without losing control: IDE, agent, metrics, methodology and quality.
Yuli Code is a desktop IDE where AI works with you without taking control away from you. It brings together in a single application the editor, code intelligence, Git, debugging, the database, containers, web design and an agent that plans, proposes and executes.
The agent proposes, the IDE controls and you authorize. This is the principle behind every Yuli feature.
What YULI stands for
YULI stands for Yield Unified Local Intelligent Framework for Orchestrated Generation & Execution.
It is not just an editor with a chat. It is a framework that unites six pieces:
| Piece | What it brings |
|---|---|
| IDE | Editor, navigation, refactoring, Git, debugging, database, containers, REST client and Web Studio. |
| Agent | Plans, proposes changes with their diff and operates the IDE through controlled tools. |
| Metrics | A–E rating, clean code practices, Six Sigma and comparison with the previous version (N−1). |
| Methodology | An 8-phase lifecycle, from Analysis to Delivery, with a mandatory Measurement phase. |
| Quality assurance | A versioned quality contract and a gate that applies the same rules to people and agents. |
| SDLC | Requirements, ADRs, threat model, supply chain, delivery and operations in the Governance Center. |
The three principles
Human control
No change from the agent reaches the disk without going through review. Actions with side effects (a commit, a push, a statement that writes to the database) are approved from a card.
Deterministic rules
Security boundaries are enforced by the engine on every operation. They do not depend on what the prompt says or on the model's goodwill.
Local first
The code, the index and the agent's state live on your machine and in your project. Only the AI you choose goes out to the internet. Details in Security.
How it is built
A native Rust engine does the heavy lifting (indexing, Git, processes, debugging, database, security and agent orchestration), and an Electron + TypeScript interface only renders. The language model reasons and operates the IDE's services through controlled tools. More in Architecture.
Two ways to use AI
| Subscription mode | Platform mode | |
|---|---|---|
| Who reasons | Your own Claude Code, Kiro, Codex or Antigravity | Yuli's engine |
| Who pays | Your current subscription | The API key of the provider you choose |
| How it uses the IDE | yuli_* tools over MCP |
Built-in tools |
In both modes the rules, the panels and the changes window are the same. Choose yours in Connect an AI.
Who it is for
- Teams that want the productivity of AI without giving up control, security or transparency over their code.
- JVM projects, including legacy systems (Ant, Eclipse, Struts, JSP, classic application servers).
- Front-ends built with React, Vue, Angular, HTML or JSP.
- Teams with quality or regulatory requirements, thanks to the quality contract and the gate.