LangWatch

LangWatch is an observability tool for AI coding assistant sessions, with a recent Product Hunt launch focused on Claude Code usage tracking. It helps developers see what their Claude Code, Codex, Gemini CLI, and opencode sessions actually cost, which models did the work, how cache reads and writes affected spend, and what commands or tool calls happened during a run. The product is most useful for builders who already rely on coding agents and need clearer budget and debugging data. The problem LangWatch solves is visibility. A monthly subscription or API invoice tells a team what it paid, but not always which session burned the most budget or whether a specific workflow was worth the cost. LangWatch records session-level cost data, token usage, cache reads, cache writes, model activity, bash commands, file edits, skills, and MCP calls. It turns those details into spans and session views so developers can inspect the work instead of guessing from the final diff. One practical feature is the comparison between theoretical API pricing and billed subscription usage. For Claude Max users, this can help answer whether a subscription plan is cheaper than API usage for a real workload. The Product Hunt discussion also notes that LangWatch stamps rates when sessions arrive, so historical reports are not silently repriced when model rates change later. That matters for teams using the numbers in monthly planning. LangWatch also provides a terminal-style replay. The makers clarified that it does not record the real TTY directly; it captures telemetry and reconstructs the session on the frontend. That approach can work for headless or CI-like runs where a raw terminal recording would be awkward. For debugging, the replay pairs with span data for commands, file edits, skills, and MCP calls. The strongest users are solo AI builders, staff engineers, and engineering managers who want to control coding-agent spend without banning useful tools. It is also relevant for teams adopting MCP because it exposes tool calls as part of the session trace. Product Hunt listed the Claude Code tracking launch as free for individuals, but teams should verify current LangWatch pricing and retention limits on the official site before standardizing on it. Before rolling it out broadly, teams should test LangWatch on a few representative coding sessions and compare the reported cost data with their own billing records. That gives a baseline for cache savings, heavy users, and which agent workflows deserve closer review.