parcela.dev

Log any LLM call. Re-run it exactly.

Capture a production completion (prompt, params, model, response) with one line. Press a button to replay it identically, or tweak the prompt or model and diff the output. No SDK lock-in, no heavyweight observability platform. Just the one thing you actually keep asking for.

Two lines to start

// After any LLM call, log it in one line:
await fetch("https://parcela.dev/api/v1/completions", {
  method: "POST",
  headers: { authorization: "Bearer " + PARCELA_KEY,
             "content-type": "application/json" },
  body: JSON.stringify({
    provider: "openai",
    endpoint: "https://api.openai.com/v1/chat/completions",
    request: requestBody,   // exactly what you sent
    response: responseBody, // exactly what you got
  }),
});

Then POST /api/v1/replay with the capture id and your provider key to run it again, byte-for-byte. Your provider key is used once and never stored.

Reproduce a regression

The exact request that misbehaved in prod, replayed on demand.

Diff across models

Re-run the same prompt on a new model and compare side by side.

No platform tax

One endpoint. Works with OpenAI, Anthropic, and compatibles.

Pricing

Free: 100 captures. Pro: $39/mo.