Automations

Create Neroli Workspace Commands and agent automations for repeatable setup, run, build, verification, review, cleanup, and scheduled work.

Automations teach your workspace to operate itself.

They cover setup, run, checks, cleanup, and review.

Humans and agents can run the same reusable actions.

Workspace Commands handle exact shell work, such as copying env files or starting Docker.

Agent automations launch sessions for work that needs judgment.

Use them to review recent runs or update documentation.

Claude Code and Codex can create or update both from inside Neroli.

Agents launched inside Neroli know about the current session, visible panes, attached browser, simulator, known services, and available automations, so they can use the same workspace controls you do.

Automations can be toolbar buttons, lifecycle hooks, scheduled agent sessions, or cleanup scripts with stop behavior.

The real Neroli workspace with its Automations entry visible — Automations can be toolbar buttons, lifecycle hooks, scheduled agent sessions, or cleanup scripts with stop behavior.

Two kinds of workspace action

Type	What it does	Example
Workspace Command	Runs deterministic shell work in the selected session.	Copy `.env.local`, assign ports, start Docker, run Build, stop services.
Agent automation	Starts an ordinary agent session with a prompt.	Review recent sessions, improve commands, update docs from shipped features.

Both are workspace-native. Manual Workspace Commands can appear in the session top bar, either pinned or in its overflow menu. They can also run on a schedule or when a session is created, archived, restored, or removed. Agent automations run manually or on a schedule, either on this Mac or in Cloud.

Automations worth creating

Automation	Why it is useful
Prepare session	Copies env files, installs deps, assigns unique ports, and creates isolated Docker or database state.
Run dev	Starts the dev server or stack from the current session, often as a pinned toolbar button.
Stop / teardown	Stops the dev server, shuts down Docker, frees ports, and removes generated state.
Build / Verify	Gives humans and agents the same one-click checks before work lands.
Improve my workspace	Agent automation that inspects recent sessions and proposes easier setup/run/verify flows.
Docs from shipped work	Agent automation that drafts docs updates from new features and merged sessions.
Review before merge	Agent automation that checks diffs, migrations, env changes, tests, docs gaps, and risky files.
Browser verification	Opens the app, drives the attached browser, and captures evidence for the review.
Verify Merge	Runs checks or starts review agents for sessions waiting in Merge.
Cleanup Done	Archives or summarizes completed sessions after validation.

The isolated dev environment pattern

The most useful automation is often Prepare session.

Copy local config

Copy .env.local, certificates, or generated files into the session without committing secrets.

Assign unique ports

Give each session its own backend, database, or frontend ports so parallel runs do not collide.

Provision services

Start or prepare an isolated Docker database, queue, cache, or local stack for that session.

Expose buttons

Pin Run, Build, and Verify to the toolbar so humans and agents use the same workflow.

Add stop behavior

Give services a Stop script so Neroli can tear them down cleanly when the run ends.

Let agents improve the workspace

Because Neroli sessions can inspect their own workspace and recent runs, you can ask an agent to set this up for you:

Create Neroli automations for this repo:
- Prepare session: copy env, assign unique ports, install deps, prepare Docker
- Run dev: start the app from the current session
- Stop: tear down services cleanly
- Verify: run the checks we should use before review

It can inspect package scripts, Docker config, previous commands, and failed sessions, then create automations that make the next session smoother.

Make automations visible

Good automations should be easy to launch from where the work happens:

pin Run, Build, and Verify in the session toolbar
keep Stop paired with anything that starts a server or service
use lifecycle setup when every new session needs the same preparation
use scheduled agent automations for ongoing maintenance, docs, and improvements
keep personal machine-specific actions local, and share repo-wide actions with the team

What happens to an agent run

Each occurrence is an ordinary session. It keeps the provider’s normal startup, approval, question, failure, suspend, resume, and follow-up behavior. Pausing an automation prevents future triggers; it does not stop a turn already running. Deleting an automation removes future scheduling, while sessions it already created remain available.

Cloud occurrences always start a new chat and continue while your Mac is offline. Local occurrences may start a new chat or target one eligible existing chat. If Neroli was closed at a local schedule boundary, it catches up once on the next launch instead of replaying every missed tick. An overlapping tick is skipped while the previous initial turn is still active.

Team automations are visible to the whole Team by default. The owner or a Team Owner/Admin can set Visible to to Only me. See Teams for run privacy, takeover, ownership, and move-to-Mac behavior.

Next: use Recipes and gotchas to turn automations into isolated dev sessions, cloud handoff, and review workflows.