Automations

Automations

ClawDE automations are trigger-action rules that fire automatically when events happen in the daemon. Use them to run tests after sessions, extract TODOs as tasks, or notify you when a long session completes.

Trigger types

Trigger	When it fires
session_complete	An AI session ends (completed or errored)
task_done	A task's status changes to "done"
file_saved	A file is written by a tool call
cron	A cron expression fires

Action types

Action	What it does
run_tests	Runs a shell command (e.g. cargo test) and pushes results as automation.testResults
send_notification	Pushes automation.notification to all connected clients
create_task	Creates a new task in the task store
run_script	Runs an arbitrary shell script

Config — .claw/config.toml

[[automations]]
name        = "run-tests-after-session"
description = "Run tests when a session completes"
trigger     = "session_complete"
enabled     = true
action      = "run_tests"

[automations.action_config]
command = "cargo test --quiet"

Multiple automations can share the same trigger type. They all fire independently.

Built-in automations

Three automations are built-in and always registered. They can be disabled in the UI or config but not deleted.

Name	Trigger	Default
run-tests-on-complete	session_complete	Disabled
todo-extractor	session_complete	Enabled
long-session-notifier	session_complete + session_duration_secs>300	Enabled

Condition expressions

Simple conditions filter when an automation fires:

Expression	Meaning
session_duration_secs>300	Only fires when session ran >5 minutes
file_ext=.rs	Only fires for Rust file changes

Leave condition empty (or omit it) to fire on every matching trigger.

RPC methods

Method	Description
automation.list	List all automations with status
automation.trigger	Manually trigger an automation (useful for testing)
automation.disable	Enable or disable an automation by name

Push events

• automation.testResults — fired after run_tests action completes
• automation.notification — fired by send_notification action

Example: TODO extractor

When a session completes, the todo-extractor automation scans the session's AI output for lines containing TODO: and creates a task for each one. This turns in-conversation TODOs into tracked tasks automatically.