LoopingLouieDocsStart looping

Design an agent loop.
Run it 24/7, anywhere.

Product / Live runImplement → review → approve
Looping Louierun_8F2A · live
Runningrun_8F2A18h 42m
Runs / Open Router release / #8F2A

Ship the Open Router integration

Implementation, security review, and human approval in one running pipeline.

Worker louie-eu-03Compute VM · eu-west-1Harness HermesCheckpoint saved 12s ago
Live workflow
Add OpenRouter as a model provider
↓
01 · Refinement loopShape idea & implement
DeepSeekDeveloper model

deepseek-v3.13 commits✓ Passed

→
02 · Security loopAudit auth & dependencies
FableSecurity reviewer · Anthropic

Claude Sonnet 4.5● ReviewingRound 2/3

→
03 · Human gateFinal approval
FDDJ4-eye reviewFD + DJ · 2 required

◷ Waiting2 approvals

✓ Implementation complete● Fable reviewing◷ Approval waiting

How it works

How Louie runs a task.

  1. 01

    Add the agents you need.

    Create one role for each job. Choose its coding harness, model, instructions, skills, and context.

  2. 02

    Define how work moves.

    Set who implements, who reviews, and where feedback goes. Repeat any step until the work is approved.

  3. 03

    Start the run and walk away.

    Run it on your laptop, server, or VM. Louie keeps it running and tells you when it finishes or needs you.

Features

What Louie does.

Keep the harness you use.

Use Codex, Claude Code, Hermes, OpenClaw, Pi, or Grok Build. Louie runs them inside the workflow; it does not replace them.

Pick the right expert.

Give each step its own model, role, instructions, skills, and context. Change one without rebuilding the workflow.

Run where your code lives.

Run a worker on your laptop, in Docker, on a private VM, or on a company runner. Your code and credentials stay there.

Resume instead of restarting.

Louie saves progress while the workflow runs. If a worker stops, another worker can continue from the last checkpoint.

Get notified when Louie needs you.

Louie tells you when a run finishes, fails, or waits for input. You do not need to keep the dashboard open.

Humans at key checkpoints.

Pause before a deploy or any critical step. A person can approve, reject, or send the work back with comments.

See every decision and change.

See each step, model, cost, retry, approval, artifact, and code diff in one run history.

Enforce policy across runs.

Set allowed models, spending limits, required reviews, and approval rules. Louie checks them on every run.

Looping Louie

Keep your coding agents.
Improve the workflow around them.

Louie coordinates the run from task to approval.

Back to the top