contextspy now available on PyPI for easy AI profiling

ContextSpy — a profiler that visualizes token usage for large‑language‑model agents — is now published on PyPI, allowing a one‑line installation via pip or the newer uv tool [2]. The move follows a multi‑option install strategy that already includes pre‑built binaries, Homebrew taps for Apple Silicon, and source builds, but the PyPI package offers the simplest path for most users, especially those on Windows or Linux where binary downloads can trigger antivirus warnings [1].

The package requires Python 3.11 or newer and installs a command‑line entry called contextspy that lands on the system PATH after the pip command completes [1]. For developers who run Python in an externally managed environment (e.g., Homebrew‑installed Python on macOS), the install may raise an “externally‑managed‑environment” error; the documentation advises creating a virtual environment first to avoid the block [1]. Once installed, users can start the proxy with contextspy start, generate a local CA certificate for cloud mode with sudo contextspy install-cert, and then launch supported coding agents (GitHub Copilot, Claude Code, Opencode, etc.) through the proxy using contextspy run <agent> <project> [1].

By exposing the full prompt composition—including system prompts, tool definitions, file contents, and conversation history—ContextSpy helps teams pinpoint why a session becomes slow, expensive, or hits context limits. The tool’s dashboard, reachable at http://127.0.0.1:5173, displays a live breakdown of token categories per request, turning opaque LLM traffic into actionable data [2]. This visibility is increasingly valuable as token costs rise and AI coding agents shift from subsidized subscriptions to per‑token billing, where input tokens can dominate expenses by a factor of 20‑50 [2].

The PyPI release lowers the barrier to entry, sidestepping the need for manual binary handling or Homebrew taps, and aligns ContextSpy with the broader Python ecosystem. As more developers adopt AI‑augmented workflows, the open question is whether the ease of installing via PyPI will accelerate adoption of profiling tools like ContextSpy, potentially reshaping how teams manage LLM costs and performance.