Other agents

Any coding agent that runs shell commands, including local models, can run the same workflow through the qualitylayer command. Supported agents also get the App's conversation view and session controls.

Open full size ↗ Open full size ↗

Start a task

1. Check the command

   Have your agent run qualitylayer help. Nothing listed? Run qualitylayer doctor.

2. Ask in plain words

   "Using QualityLayer, plan a CSV export without the App."

3. Continue in your agent

   It follows the same workflow, asks with numbered choices and saves the documents on disk.

Note

Results depend on the model: it has to call tools reliably over many turns.

Plan with one agent, build with another

Open full size ↗ Open full size ↗

On Start the build, choose In my own agent, then ask your agent in the project to continue the approved task, or paste the prepared request. It follows Implement and Verify until your final approval.

What changes

	Supported agents	Any other agent
Setup	Done when you install	Ask in your own words
Model and effort	Set in QualityLayer	Set in your agent
Building	One session builds the slices; it can use its agent's own subagents	One session builds the slices with its own tools
Sessions	The App's Conversation and session controls	Your agent's own window and controls
Cost panel	Per step, when the agent's records hold token counts	Not shown

For the agent's author

The prompt asks the agent to run qualitylayer next --json, follow the reply and read the files it names, until the reply says done or stop.

To keep working after a question, run qualitylayer wait --until-event in the background. It returns on an answer, review result or finished check, or after 30 minutes.

Checks run on a clean copy of the commit, without .venv or node_modules. Use $QL_REPO_ROOT to reach your folder, for example "$QL_REPO_ROOT/.venv/bin/python" -m pytest.