VoxClaw runs a small HTTP listener on a user’s Mac and speaks text sent by remote agents. It supports Apple’s built-in TTS or OpenAI neural voices (when the user provides an API key) and exposes endpoints for single reads, agent notifications (summaries/progress/failures), and a status check.
Use VoxClaw when your agent runs headless or on a remote server but you want spoken output played locally on the user’s Mac — for final task summaries, failure alerts, or optional live progress narration. Ideal for agents that need audible confirmations without granting remote audio access.
GET /status), speak text (POST /read), and send agent notifications (POST /agent-notify) including body params like voice, rate, and instructions. The SKILL.md also documents discovery via Bonjour (_voxclaw._tcp) and a recommended connect order.Compatible with agents that can make HTTP requests from a host (curl/requests). Works well with remote server agents such as Copilot/Codex/Gemini CLI that can perform HTTP POSTs or shell curl commands.
VoxClaw is a macOS menu-bar app that provides HTTP-based TTS output for agents, supporting Apple TTS and OpenAI neural voices. The SKILL.md is well-written with clear API documentation, connection procedures, and error handling guidance. No bundled scripts to test — it's purely a reference/integration skill that relies on curl to a local HTTP endpoint.
Clean skill — no security red flags. No hardcoded credentials, no remote code execution patterns, no destructive commands. The skill is purely instructional (how to POST to a local HTTP service). Minor deduction for unauthenticated local HTTP API (acceptable for LAN use but worth noting) and the niche audience (requires a Mac with the app installed). Well-structured SKILL.md with clear frontmatter, API reference, and usage examples.