Skip to content

Projects

The agent reads these, most general first:

  • AGENTS.md in your IXEL settings folder, and Claude Code’s ~/.claude/CLAUDE.md;
  • the project’s AGENTS.md, CLAUDE.md, .claude/CLAUDE.md and CLAUDE.local.md.

IXEL One keeps what it learns about a project in .ixel/MEMORY.md. /memory shows it.

/add-dir <path> lets the agent work in another folder without asking. /scope <dir> keeps it to one package of a monorepo; /scope off ends that.

Several repos can work as one system. Put an ixel.workspace.yaml in the folder above them. The agent then knows each repo, works across them, and searches them all.

  • /workspace lists the repos.
  • /workspace clone clones the ones that are missing.
  • /workspace sparse <folder> widens this repo’s sparse checkout.

Repos outside the workspace folder become working folders only after you trust the project.

.ixel/architecture.yaml lists your layers and which may import which. Every write and edit is checked for the imports it adds, in JavaScript and TypeScript, Python and Go. A change that breaks a rule is refused, or noted in warn mode. /arch checks the whole project.

IXEL One runs language servers for the project: TypeScript, Python, C and C++, Rust, Go, Java and many more. It starts them only once you trust the project. Installing one always asks first.

At the end of each turn, it reports the problems that turn’s changes left, and sends the agent back to fix them. /lsp shows the language servers; /lsp stop stops them.

/owners shows who owns your changed files, from CODEOWNERS, Backstage or git history.

Was this page helpful?