LIQ CODE1. Install where you work
One release covers the desktop app, the console client, and the CLI, so you can start in a window and continue in a terminal on the same machine.
How it works
From intent to reviewed code.
A LIQ CODE session is a loop you stay in charge of: describe the outcome, agree on the plan, let the agent do the work, then review what it produced.
The loop
2. Point it at a project
The agent indexes the code it is given, which keeps navigation cheap on large repositories instead of re-reading files on every question.
3. Agree on a plan
Planning runs as its own phase: you see the intended approach and adjust it before a single file is touched.
4. Let it execute
The agent edits files, runs commands, and splits larger work across parallel sub-agents, inside an isolated environment when you ask for one.
5. Review every change
Nothing is accepted on your behalf. You read the diff, run the tests, and decide what stays.
6. Automate what repeats
Once a task has a shape, the same agent runs non-interactively from a script or a CI step and returns output you can parse.
Coding, research, and scaffolding a project from an empty folder all run through this same loop, and every run draws on the AI credits included in your plan.