Trigger.dev MCP. Manage job runs and monitor failure status instantly.

You don't gotta jump through a dozen dashboards just to check why some background job blew up. Trigger.dev gives your AI agent direct access to your entire job infrastructure, letting you manage everything with plain English commands. It’s like having the full admin console right in your chat window.

What You Can Control

Your agent can first scope your investigation by calling list_projects, which returns every project name and ID connected to Trigger.dev so you know exactly what context you're working in. It runs list_environments to give you a full list of deployment scopes—things like 'staging' or 'production'—letting you pinpoint where the failure happened.

You can also use list_schedules to retrieve all scheduled tasks, showing their specific cron schedule definitions so you know when things are supposed to run.

When you need to see what jobs have actually run, your agent calls list_runs. This tool lists multiple task runs based on criteria like a time window or status (say, 'failed' or 'completed'), giving you an immediate overview of execution times and statuses. If you find a specific job ID that needs deep inspection, the agent uses get_run to retrieve comprehensive details for that run.

That dump includes the full logs, the exact input payloads used, and any error stacks generated during the process.

If a job is running wild or doing something it shouldn't, you can instantly stop it by calling cancel_run, which kills a background task immediately using its unique ID. If you find a bug fix that needs testing without writing new code, the agent runs replay_run. This re-executes a specific past job run, using its original input payload so you can verify fixes safely.

Need to test something right now? You don't need an external webhook or manual setup. The agent calls trigger_task to manually start an asynchronous background process instantly, simulating exactly what an incoming event would do.

How Your Agent Handles It

You just tell your AI client what you want done—for example, 'Show me all failed inventory sync jobs in production that ran last week.' The agent handles the sequence: it first uses list_environments to confirm 'production' is a valid scope. Then, it combines list_runs with filtering logic for status and time range.

It presents you with the list of IDs, so you can then tell it, 'Get me the full logs for run ID 123.' The agent executes get_run, pulling back every log line and payload detail without you ever touching a dashboard. If you find something that looks good but needs to be validated against current data, you just ask it to replay the job using its previous settings via replay_run.