raw-cli: Local Coding Agent with Composable Tools

What changed

The raw-cli project, developed in TypeScript, has recently released version v0.1.3 on October 9, 2026. This release marks a fresh update for the local coding agent, which is built to support Large Language Models (LLMs) with limited context windows. The agent is designed with a modular architecture, featuring composable tools, skills, hooks, and agents. It provides multiple interfaces for interaction, including a terminal REPL (Read-Eval-Print Loop) for command-line use and a browser-based dashboard for a visual interface. The project also supports the Agent Communication Protocol (ACP).

Why it matters for builders

For developers working with LLMs that have strict context length limitations, raw-cli offers a practical solution. The ability to compose tools, skills, and agents allows builders to create specialized coding assistants tailored to specific tasks or project requirements. The dual interface options (terminal and browser) cater to different developer preferences and workflows, enhancing accessibility and usability.

Practical impact

Builders can leverage raw-cli to develop and deploy local AI coding agents without relying on cloud-based services, potentially improving privacy and reducing latency. The composable nature encourages experimentation with different agent configurations and tool integrations. The recent release suggests ongoing development and support for the project, making it a viable option for those seeking to enhance their local development environment with AI capabilities. Developers can explore integrating custom tools or skills to extend the agent's functionality for their specific coding needs.

Caveats and source limits

The provided metadata indicates a recent release (v0.1.3, October 9, 2026) and a moderate number of stars (102) and forks (11). However, details regarding specific supported LLM models, performance benchmarks, advanced configuration options, or comprehensive documentation are not available in the source. The project's maturity and the full extent of its capabilities are not fully detailed, and the absence of explicit examples or extensive documentation might present a learning curve for new users.