Automations

Automations are the mechanism that lets Agents work without a human trigger.

Traditional AI tools require you to manually start every conversation. Automations break that limitation — you define a trigger condition and a set of instructions, and when the condition is met, XClaw automatically creates a session and has the Agent execute.

Trigger types are diverse: scheduled (cron expressions), external events (webhooks), system events (session labeled, tool invoked). This means you can build complete event-driven workflows.

Some real examples: auto-generate a daily code change summary and post it to Slack every morning; run a security scan automatically after each PR submission; when a monitoring metric goes abnormal, analyze logs and generate a report.

XClaw's automation engine supports 35+ event types internally. Each rule can be configured with retry strategies and timeouts. Execution history is fully recorded so you can audit the Agent's automated work at any time.

The core idea behind automations is pushing the Agent from "passive responder" to "proactive worker." You no longer need to remember when to ask the Agent to do something — define the rules, and it does the right thing at the right time.

How to

Open Settings > Automations and click "New Rule" to get started. First, pick a trigger type — the most common are "Scheduled" (enter a cron expression like 0 9 * * 1-5 for weekday mornings at 9am) and "Webhook" (generates a URL that external services can call to trigger the rule).

Next, write the instructions for the Agent. Just describe the task in natural language, the same way you'd talk to the Agent in a session. For example: "Check commits on the main branch from the last 24 hours, generate a change summary, and post it to the #dev-updates Slack channel."

Each rule also supports retry count and timeout configuration, handy for network tasks that might occasionally fail. Hit "Save & Enable" and the rule starts working.

Want to check the run history? Click any rule in the list to see when it was triggered, the execution result, and detailed logs.