Claurst — Agentic Coding Harness Built in Rust

Introduction

Claurst is a coding agent harness written in Rust that provides a streamlined, fast interface for AI-assisted development. It focuses on being a practical tool for builders who want to ship code quickly with AI help, without the overhead of heavier frameworks or complex setup.

What Claurst Does

• Provides an interactive TUI for conversing with AI models about code
• Reads and writes files in your project directory based on model instructions
• Supports multiple LLM providers including Claude, OpenAI, and Gemini
• Offers a session system for organizing work across tasks
• Ships as a single binary with no runtime dependencies

Architecture Overview

Claurst is a monolithic Rust binary containing a TUI renderer, an LLM client layer, and a file operations module. The TUI is built with the ink-inspired rendering approach for responsive terminal output. The LLM client supports tool use for file read, write, search, and shell command execution. Session state is persisted to disk as JSON.

Self-Hosting & Configuration

• Install via cargo install from source or download a release binary
• Configure API keys for your preferred LLM provider in ~/.claurst/config.toml
• Set default model, temperature, and context window preferences
• Project-specific settings can be placed in .claurst.toml at the repo root
• Supports environment variables for all configuration options

Key Features

• Single Rust binary with instant startup and low memory footprint
• Multi-provider support for Claude, OpenAI, Gemini, and DeepSeek
• Built-in TUI with syntax highlighting and markdown rendering
• Session management for organizing multi-step coding tasks
• File and shell tool use for autonomous code modifications

Comparison with Similar Tools

• Claude Code — feature-rich but requires Node.js runtime; Claurst is a single Rust binary
• CodeWhale — focuses on open-weight models; Claurst supports both cloud and local providers
• Aider — Python-based with git integration; Claurst is Rust-native with session management
• OpenCode — Go-based agent; Claurst provides a similar experience in Rust

FAQ

Q: Why Rust instead of TypeScript or Python? A: Rust provides instant startup, low memory usage, and a single self-contained binary without runtime dependencies.

Q: Can I use local models? A: Yes, point the configuration at any OpenAI-compatible endpoint such as Ollama or llama.cpp.

Q: Does it support git integration? A: Claurst can execute git commands via its shell tool, but does not have built-in git commit workflows.

Q: What platforms are supported? A: Linux, macOS, and Windows. Pre-built binaries are available for all three.

Sources

• https://github.com/Kuberwastaken/claurst