Waveloom is an open-source Go-based terminal coding agent created by Menfre01. It is designed to autonomously inspect repositories, generate code, and execute development tasks directly from the command line while maintaining explicit control over file modifications and shell execution. During day-to-day use, Waveloom presents an interactive Terminal User Interface (TUI) featuring automatic terminal background theme detection and structured agent loops aligned with ACP-style execution. Developers submit natural language tasks, and Waveloom iterates through planning and execution steps, generating diff previews for file edits and writes inside approval dialogs prior to applying modifications. What sets Waveloom apart is its built-in operating-system-level sandbox isolation layer, leveraging Bubblewrap (`bwrap`) on Linux and `seatbelt` on macOS alongside binary decision permissions. It also supports cross-platform operations including Windows with Git Bash integration and includes modular agent skills and an internal evaluation framework (`eval/llmedit`). Waveloom is open source and hosted on GitHub under active community development. It uses environment-based configuration via `.env` files and an interactive setup wizard to configure underlying LLM endpoints and credentials.
Tags: ai agent, cli, coding agent, OpenSource, terminal
bwrap) on Linux and seatbelt on macOS. - Interactive TUI with Approval Previews — Displays real-time file diffs and command previews directly inside TUI approval modals before changes touch disk. - Turn-and-Step Agent Loop — Implements an iterative reasoning and tool execution loop modeled on modern agent architectures. - Configurable Agent Skills — Provides modular skills support for workflows like issue tracking, triage labeling, and domain-specific layouts. - Cross-Platform Compatibility — Supports Linux, macOS, and Windows environments, including native Git Bash integration for shell tooling. - Interactive Setup Wizard — Simplifies initial onboarding through interactive setup prompts and standard .env configuration.bwrap or seatbelt while providing clear visual diff previews in the TUI approval dialog before committing any writes to disk.eval/llmedit suite to test editing fidelity and execution loops across different task benchmarks.Waveloom is built for software engineers, DevOps practitioners, and Go developers who prefer a TUI-driven workflow and demand strong OS-level sandboxing for terminal-based AI agents. It is not intended for non-technical users or developers seeking IDE-embedded extension panels without terminal interaction.
bash git clone https://github.com/Menfre01/waveloom.git cd waveloom go build -o waveloom ./cmdbash cp .env.example .env # Add your LLM API keys and model configurations to .env ./waveloombash ./waveloom "Inspect pkg/tui, fix failing tests, and prompt for confirmation before modifying files"Part of CLI Coding Agents
Categories: AI Coding Agents, Code Assistants, Developer Tools