Docs/Workspace

Project tools

Git, memory, packages, dependency and context optimizers, custom endpoints, System One and reports.

Everything around the code lives next to your agents, in the same window.

Changes and Git

  • Diff viewer for every file the crew touched.
  • AI-written commit messages, then commit & push in one step.
  • A branch menu and Pull Request creation, signed in with GitHub (githubAuth.ts).

Memory and skills

  • Memory store (memoryStore.ts) — project rules that are given to agents on every call.
  • Shared skills (sharedSkills.ts) — skills written once and shared between agents.

Package manager

A UniGetUI-style package manager for the project: see and update dependencies from package.json, pyproject.toml and more.

Dependency optimizer

An AI-assisted optimizer that finds unused packages and removes them. It tells stdlib, local and aliased (mapped) imports apart so it doesn’t flag what is actually used. It is also offered to agents as a tool.

Context optimizer

Long context is compressed intelligently instead of being cut off blindly (optimizer/), so agents keep what matters and spend fewer tokens.

Custom endpoints

Connect any OpenAI-compatible provider with live model discovery. The OpenRouter key is stored encrypted — see Configuration and security.

System One

A “compiler” mode that produces a strict JSON contract validated by a Zod schema: the target file, create / modify, the code and a memory patch (systemOneCompiler.ts). See Agent chat and System One.

Reports, history and usage

  • Export a session report as Markdown, HTML or JSON.
  • Browse the run history.
  • See usage across agents (usageLedger.ts).

Edit this page on GitHub ↗