Browser Audio Capture

Browser Audio Capture

Give any AI 代理 ears for the browser. One Chrome 扩展 captures audio from any tab — meetings, YouTube, podcasts, 网页inars, courses, earnings calls — and 流s it to your AI 流水线.

Why Use This

Your AI 代理 can't hear anything h应用ening in your browser. This 技能 fixes that. Capture audio from any Chrome tab and 流 it to your 代理 — no API keys, no OAuth, no per-平台 integrations.

Use cases: meeting summaries, YouTube/podcast notes, competitive intel from earnings calls, auto-notes from online courses, customer call analysis — anything that plays audio in a browser tab.

Works with any AI 代理 — Claude, ChatGPT, OpenClaw, LangChAIn, CrewAI, or your own. If your 代理 can 运行 shell commands or 接收 HTTP, it 获取s browser audio.

Prerequisites

Chrome with remote 调试ging:

# macOS /应用s/Google\ Chrome.应用/Contents/MacOS/Google\ Chrome \ --remote-调试ging-port=9222 --user-data-dir=$HOME/.chrome-调试-性能分析 &

Python 3.9+ with AIohttp: pip 安装 AIohttp

Quick 启动 命令行工具 (any 代理 that can exec) # 列出 tabs — meetings flagged with 🎙️ python3 -m browser_capture.命令行工具 tabs

# Auto-检测 and capture meeting tab python3 -m browser_capture.命令行工具 capture

# Continuous watch mode python3 -m browser_capture.命令行工具 watch --interval 15

# 停止 python3 -m browser_capture.命令行工具 停止

Chrome 扩展 (one-命令行工具ck, persistent) chrome://扩展s/ → Developer mode → Load unpacked → scripts/扩展/ Join a meeting → 命令行工具ck Percept icon → 启动 Capturing Close popup — capture continues in background Supported 平台s

Google Meet • Zoom (网页) • Microsoft Teams • 网页ex • Whereby • Around • Cal.com • Riverside • 流Yard • Ping • DAIly.co • Jitsi • Discord — plus any future 平台 that 运行s in a browser.

Audio 输出

流s to http://127.0.0.1:8900/audio/browser as JSON:

{ "会话Id": "browser_1709234567890", "audio": "", "sampleRate": 16000, "格式化": "pcm16", "tabUrl": "https://meet.google.com/...", "tabTitle": "Weekly Standup" }

配置 端点 in scripts/扩展/offscreen.js (PERCEPT_URL). Point it at Whisper, Deepgram, NVIDIA Riva, or any transcription 服务.

Troubleshooting No tabs: Chrome needs --remote-调试ging-port=9222 Button won't 命令行工具ck: 移除 + re-添加 扩展 (MV3 缓存s aggressively) Audio not arriving: 检查 接收r on port 8900. 扩展 发送s to /audio/browser