Promptcellar – capture every Claude Code prompt as JSONL in your repo
Git-native prompt logging for Claude Code—audit trails stay in your repo, not vendor clouds.
See what Claude Code and Codex actually send to the API — and what each part costs.
Mitmproxy capture shows why each turn costs without touching agent config.
Developers using Claude Code or Codex
Git-native prompt logging for Claude Code—audit trails stay in your repo, not vendor clouds.
No-proxy LLM cost tracking beats Helicone for teams avoiding traffic rerouting.
Real-time cost tracking for Claude Code sessions; saves money intuition before you ship.
Prompt clustering with cost attribution when LangSmith already does observability.
Cost attribution per event via OpenTelemetry, but Anthropic Workbench and Vercel do this already.
Drop-in proxy that cuts GPT token costs 40-60% without changing app code.