Coding-agent observability for session costs, tokens, cache usage, tool calls, file edits, and terminal replays across Claude Code, Codex, and more.
Features
- Per-session token and cost tracking
- Separate cache read and write accounting
- Bash, MCP, and file-edit tracing
- Full terminal session replay
- Support for multiple coding agents
Use Cases
- Coding-agent cost governance
- Team budget and usage analysis
- Tool-call auditing
- Session troubleshooting and review
- Model and cache strategy comparison
FAQ
LangWatch connects to Claude Code with one command and also supports Codex, Gemini CLI, and OpenCode. It reports tokens, cache reads and writes, model costs, Bash and MCP calls, file edits, and reconstructed terminal replays per session, alongside its broader LLM observability and evaluation platform. Core capabilities include: Per-session token and cost tracking, Separate cache read and write accounting, Bash, MCP, and file-edit tracing.
Common scenarios include: Coding-agent cost governance, Team budget and usage analysis, Tool-call auditing.
