Local tool for Claude Code users

See what your agent loop actually did, iteration by iteration.

Set a goal, a continue prompt and a stop condition. LoopLens runs the loop on your own machine and records each pass: the prompt it sent, the model that really answered, tool calls, tokens, cost, and why it kept going or stopped.

Request early access

LoopLens timeline showing three iterations of a loop, each with the prompt sent, the model served, output and the stop-condition check. — A three-iteration run that stops when the output contains DONE.

Runs locally

A single Node server with no dependencies, bound to 127.0.0.1. Your prompts and transcripts stay on your machine as plain JSONL files.

Shows the model that served

Each iteration records the model you asked for and the one that answered, so silent fallbacks are visible.

Replay without spending quota

Every run is recorded. Replay it at any speed to study where a loop drifted, at no cost.

Guarded by default

The server refuses requests from any other website, and permission settings come only from your own environment, never from a request.