LangWatch

LangWatch is a powerful observability tool designed for developers and AI engineers who utilize AI coding assistants like Claude Code, Codex, Gemini CLI, and opencode. Its primary function is to provide granular insights into the usage and cost of these AI sessions, enabling users to make informed decisions about their AI tool consumption and budget. The problem LangWatch addresses stems from the often opaque nature of AI coding tool costs. While users might know the per-seat subscription price, understanding the actual expenditure per session, the specific models used, and the efficiency of caching mechanisms can be challenging. This lack of visibility can lead to unexpected budget overruns, especially as token prices evolve. LangWatch aims to demystify these costs and provide a clear picture of AI tool utilization. One of LangWatch's key features is detailed cost tracking. It breaks down the cost per session and per model, with cache reads and writes accounted for as separate token classes. This allows users to pinpoint exactly where their budget is being spent. For users on a Max plan, LangWatch offers a comparison between theoretical costs and billed amounts, providing a realistic view of their expenditure. Another significant capability is the tracking of every tool call. Bash commands, file edits, and MCP tool calls are logged as distinct spans, each with its duration. This provides a clear audit trail of actions taken during a session and their associated costs, aiding in debugging and performance analysis. LangWatch also offers a full terminal replay feature. This allows users to step through an entire session in the UI exactly as it happened. This visual replay is invaluable for understanding the flow of a session, identifying bottlenecks, and learning from the AI's interactions. The tool works by not sitting in the request path, thus introducing no additional latency to model calls. Instead, it reads session data that AI coding tools already write locally and ships it in batches in the background. This approach ensures that the tracking process itself does not impact the user's experience or the performance of the AI tools. The benefits for users include a clear understanding of AI session costs, enabling better budget management and optimization. By identifying which sessions and models are most expensive, users can refine their strategies and potentially reduce expenditure. The detailed logs and replays also aid in troubleshooting and improving the efficiency of AI-assisted coding workflows. Concrete use cases for LangWatch include individual developers tracking their personal AI coding expenses, teams monitoring shared AI tool usage to allocate costs accurately, and AI engineers analyzing the performance of different models and tools to optimize their development processes. It's also useful for understanding the cost-effectiveness of specific commands or sequences of actions within a session. LangWatch is free for individual use. It supports Claude Code, Codex, Gemini CLI, and opencode. The product is positioned for developers, AI engineers, and teams working with AI coding assistants who need visibility into usage and costs. In summary, LangWatch provides essential observability for AI coding sessions, offering detailed cost tracking, command logging, and terminal replays to help users understand, manage, and optimize their AI tool usage.