Ordewell is a tool that converts a single goal into an ordered plan of coding-agent tasks, each with configurable runner, model, and mode, then executes and verifies them. It provides a transparent, editable plan before execution, supports multiple AI providers and runners, and includes both a CLI TUI and VS Code extension.
Bough is a terminal coding agent where the model writes JavaScript programs that execute tool calls directly, enabling batch operations and conditional branching in a single step rather than sequential round trips. The tool supports multiple LLM providers, operates locally with append-only session history, and is highly configurable through YAML plugins.
Codex is OpenAI's coding agent that reads, edits, and tests code within projects, available across all ChatGPT plans. Unlike ChatGPT, it performs autonomous work by inspecting repositories, modifying files, running commands, and verifying results in a structured loop. The desktop app integrates chat, files, terminal, and browser in one window for streamlined development workflows.
The author describes running a local coding agent using Oh My Pi with Qwen3.8-27B on two RTX 3090s. Key optimizations include adjusting thinking budgets, token limits, and subagent concurrency to achieve practical inference speeds. Local setups offer privacy and cost predictability but require careful tuning and accept slower inference compared to hosted frontier models like Claude or GPT.
Hazzel is a lightweight terminal coding agent that reads code, edits files, runs commands, and manages git repositories with user approval at each step. It supports multiple AI providers (OpenAI, Anthropic, Mistral, Groq, Gemini, DeepSeek, Ollama) and emphasizes transparency and control over feature bloat.
A1ex is a simple LLM coding agent written in Lua that integrates with OpenAI-compatible API endpoints. The project emphasizes security risks inherent to LLM agents, recommending isolation in containers or VMs and cautioning against exposing API keys to untrusted prompts.
Frank is a local plugin for AI coding agents that enforces evidence-based replies by requiring verification receipts after code changes and preventing agents from claiming completion without concrete evidence. The tool adds hooks to Claude Code and other platforms to push back on disputed answers with specific verdicts and checks, demonstrating through benchmarks that it reduces false apologies and improves agent accuracy in technical disputes.
MiniDSH is a minimal, local-first coding-agent harness that provides tools, sessions, and permissions around a language model. It serves as both useful software for editing files and running commands with user-controlled authority, and a reference architecture demonstrating how to maintain global architectural integrity in AI-assisted systems built through agent-driven development.
Metis is a coding agent that performs software engineering tasks through search, memory, execution, and verification across terminal and desktop interfaces. It supports multiple LLM providers, features recursive multi-agent workflows, durable session memory, and benchmark-validated reliability for real-world development tasks.
A developer built Tetris using a local coding agent running open-weight models offline on a laptop with plain HTML, CSS, and JavaScript. The project succeeded technically but revealed a critical failure: the model repeatedly crashed when generating the game logic file due to context window misconfiguration (32768 tokens set but insufficient for the actual task), causing 70 minutes of retries with zero progress despite correct reasoning.