Loop Agent
local · checking Source on GitHub ↗

A SMALL AGENT, OPEN TO INSPECTION

Loop Agent

An AI agent with tools, memory, and a step-by-step execution trace.

Follow a task from instruction to tool calls to saved memory. Every result has a record you can open.

Recorded walkthrough. These are real Python runs, recorded at build time; it does not run an AI model or save your data. Select an example below.

TASK

Choose a recorded task

Deterministic planner · no API key

EXECUTION TRACE

Inside the turn

0 ms

Recorded calls and results from the selected example. This is not the model’s private reasoning.

No turn yet.

Choose an example to inspect its recorded calls and results.

SAVED STATE · SQLITE

Memory after this recorded turn

0 recent memories 0 recent turns 4 local tools
EMPTY BY DEFAULT

Ask the agent to remember something. It will appear here and remain after restart.