A terminal-based AI coding agent that applies hash-anchored file edits and integrates Language Server Protocol operations to reduce token waste and edit errors. It orchestrates tool calls for developers working from the command line and supports multiple model providers.
Project overview
It anchors file edits to content hashes and wires Language Server Protocol operations directly into those edits to address whitespace battles and string-not-found errors common in other AI coding agents.
Project type
AI Agent · AI Coding
Use cases
Coding & Development
Deployment
Refer to project documentation
License
MIT
Best for
Developers who work from the command line and need an AI coding agent that leverages Language Server Protocol operations and hash-anchored edits.
Key capabilities
Edits files by content hash anchors to prevent whitespace battles and string-not-found errors.
Wires Language Server Protocol operations directly into edits for rename, navigation, and diagnostics.
Fans out jobs across isolated worker subagents with typed, schema-validated results.
Drives a real browser via Puppeteer with stealth mode enabled by default.
Curates memory between sessions by writing and recalling durable facts project-scoped to the repository.
Limitations and risks
Requires Bun version 1.3.14 or higher.
Getting started
Install globally without manual build steps using Bun or Homebrew, then launch the agent via the command line.