Frequently asked questions

General

Is Yuli free?

Yes. The IDE is a free download and doesn't require an account. You pay for the AI: with the subscription you already have (Claude Code, Kiro, Codex or Antigravity) or with your provider's API key.

What does YULI stand for?

Yield Unified Local Intelligent Framework for Orchestrated Generation & Execution. See What is Yuli.

Which systems does it run on?

Windows, macOS and Linux, 64-bit. See Install Yuli.

Is it finished?

Yuli is in BETA. It works for real-world projects, but it may have bugs. Report them with Help → Report a problem….

Which languages does it support?

It has highlighting for dozens of languages and language servers for Java, TypeScript, JavaScript, CSS, Python and Rust. Its strongest areas are Java (including legacy systems with Ant, Eclipse, Struts and JSP) and the web front end. See Editor and languages.

The AI

Do I need an API key?

No. By default Yuli works with your Claude Code, Kiro, Codex or Antigravity subscription. An API key is only needed in Platform mode. See Connect an AI.

Can I use a local model?

Yes. In Platform mode, the Local provider accepts servers compatible with the OpenAI API: Ollama, LM Studio, vLLM or llama.cpp. See Models and providers.

Does Yuli modify Claude Code or use my account?

No. Yuli opens the claude you installed, with documented options. You sign in yourself, using the official flow. Yuli never touches your credentials.

Can the agent break my project?

Every change goes through the AI changes window before it reaches the disk and leaves a checkpoint. You can undo an entire task with Ctrl+Alt+Shift+Z. See Review changes.

Can the agent push or delete my branch?

Not unless you ask. Git, CI and DevOps are blocked unless explicitly requested, and even then each action asks for approval. Creating branches requires its own request. In unattended mode, git push is not available.

Which language does it reply in?

In the language you write to it in.

Why doesn't the agent finish until everything is green?

That's the definition of done: if the turn changed code, it must build, pass the tests, start and pass the Measurement phase. That way you never get half-finished work presented as done.

Security and privacy

Does my code leave my machine?

Only what the AI you chose reads, sent to that provider. Before anything leaves, secrets are replaced with a placeholder. With a local model, nothing leaves your network. See Security.

Does Yuli read my .env files?

No. Neither Yuli's agent nor Claude Code, Kiro, Codex or Antigravity can read .env, private keys, certificates or credentials. See Secrets and privacy.

Where are my API keys stored?

In the operating system's keychain. Never in the project.

Can the agent delete data from my database?

Only if you grant Super permissions on that connection, with that kind of statement checked, and you approve each statement. By default it can only query. See Secure database.

Does Yuli send telemetry?

Problem reports are only sent if you choose to, and they never include your code or your conversations. Agent metrics are calculated locally.

Project and team

Does Yuli change my repository?

By default it doesn't write agent configuration inside the project. What it stores in .ai/ is split between what is shared with the team (contract, memory, rules) and what is local, which it excludes from Git with .git/info/exclude. It never edits your .gitignore. See The .ai folder.

Can I enforce quality rules across the whole team?

Yes, with the quality contract in .ai/contract.json and the gate in Git hooks and CI. They apply equally to people and agents.

Do commits say an AI wrote them?

No. Yuli never adds an AI Co-Authored-By or “Generated with”. The author is always the person responsible.

Can I use Yuli with Cursor, Copilot or Gemini CLI?

They don't run inside Yuli, but you can export the same project rules to them from the Governance Center. See Kiro, Codex and Antigravity.