Tasks

tell resolve what to do, when to run, and where to report tasks can run on a schedule or in response to an event to monitor deployments, triage alerts, generate reports, and more, using the same investigation abilities resolve brings to a live incident, but on autopilot you create tasks in plain language ("every monday morning, summarize last week's alerts and dm me") in a chat with your agent, from a slack or ms teams thread, or in the resolve ui everything a task can do, your agent can also do when you ask in chat when tasks run on a schedule run at an interval or a cron expression, in your time zone hourly deploy checks, daily alert digests, weekly report generation on events ask your agent to watch a slack or ms teams channel for topics you care about ("watch #deploys for rollbacks, check the slo dashboard, and dm me if things are trending poorly") when a matching message appears, the agent wakes up and follows your instructions summarize, check dashboards, notify you, or kick off deeper work on demand one off tasks run immediately or at a due time ("tomorrow at noon, check whether checkout error rates recovered") running tasks can also file their own follow up tasks when they discover new work where tasks report each task has a reporting destination a slack channel or thread, a direct message, an ms teams conversation, or its task page in resolve reports, files, and links the task produces are attached to the task's run history so you can review any run later what tasks can read your integrations anything your configured integrations expose (metrics, logs, traces, dashboards, code, deploys, tickets), scoped to the teams the agent belongs to your mcp integrations any tools exposed by mcp servers you've connected investigations search and read past and in flight investigations, their findings, and their evidence your knowledge base and the agent's memory team runbooks and docs, plus learnings the agent has accumulated what tasks can do actions from your integrations any action included in your configured integrations query a dashboard, fetch logs, look up a deploy, comment on a ticket all actions are subject to the permissions of that integration send messages post slack or ms teams messages, reply in threads, send dms, add reactions personal agents deliver privately (dm) by default and only post publicly in a channel when you've explicitly allowed it actions from mcp integrations any action exposed by the mcp servers you've connected, alongside native integrations internal actions things resolve can change inside resolve these are the built in actions your agent can take on resolve itself each one is a named tool call, visible in the agent's activity feed, and scoped to the agent's own team (or, for personal agents, to you) create and update tasks the agent manages its own work queue ask it to create a task, change an existing task's instructions or schedule, pause one, or wind one down tasks are one shot (run once, now or at a due time), recurring (interval or cron, time zone aware), or projects (groupings of related tasks) lifecycle you can steer in plain language active, paused (waiting on you), done, or canceled only an explicit status change stops a recurring task; a single run finishing never silently kills the schedule running tasks can file follow up tasks for newly discovered work, including scheduled or recurring follow ups watch channels (event driven triggers) ask the agent to watch a slack channel for keywords and act when they appear each ask creates an independent watch rule (a channel can have several) with your interest recorded in your own words plus an instruction for what to do when it fires rules can be grouped under a project recommend knowledge updates the agent can propose new runbooks and edits to existing team docs, either from a conversation ("update my knowledge with what we’ve discussed here") or automatically by reviewing completed investigations for lessons create reports and files tasks and chats can produce durable outputs in the agent's workspace markdown reports digests, summaries, audit write ups rich html interactive reports with charts, visualizations, and buttons images generated diagrams and visuals, rendered inline files the agent writes into its workspace are attached to the task's outputs and viewable (and shareable) from the task page and run history in resolve run your skills skills are reusable instruction packages you define at the org, team, or personal level your encoded procedures the agent matches tasks and requests against its skill listing and loads the right skill's instructions before acting, so your procedures drive recurring workflows