We’re building Ever Works in public. Some things may be incomplete, missing, or broken while we continue improving the platform.We’re building Ever Works in public — expect a few rough edges.
Feature · Always on
A site that only changes when someone remembers to change it falls behind its topic within weeks. Ever Works keeps track of that topic for you: once a Work is live, scheduled runs go back to the sources, re-check what the Work lists, refresh what has gone stale and redeploy — while Missions and Agents wake on their own clocks to decide what to do next. Every timer in the account is listed in one place, and a digest tells you what came of it.
Each loop is owned by the thing it belongs to and set where you already work. None of them needs a person to press a button once it is armed.
Open a Work → Schedule and pick one of seven cadences, from hourly to monthly. Each run revisits the lifecycle — re-researching the topic, refreshing stale content, regenerating affected pages and redeploying. Run now skips the wait, and a schedule that keeps failing pauses itself: three consecutive failures by default, configurable from one to ten.
Research happens when a Work starts and again on its schedule, so the sources behind each page are re-read as the topic moves rather than captured once and forgotten.
Some checks need no schedule from you at all. Item source validation, community pull-request processing and website-template auto-updates each run hourly, and comparison pages regenerate every six hours.
Put a Mission on a five-field cron and each tick proposes new Ideas toward its goal, held back by an outstanding-Ideas cap — 20 unless you set your own. Turn on Auto-build and every Idea a tick spawns is queued straight into its own Work, inside the Mission’s and the account’s spend caps.
Give an Agent a cron cadence and it wakes even with nothing assigned, works out the single most useful next action and takes it. Tune it per Agent to propose work, only observe, or do nothing on an idle tick.
A Goal reads one metric — Stripe gross volume, GA4 sessions, a PostHog event count or a JSON endpoint of your own — on the cadence you set, never more often than every 15 minutes, and records every observation against your target.
Four settings, once. After that the loops run on their own and report back.
New accounts start with the Work agent gate switched off. Enable it at Settings → Work agent; until you do, no build path runs, and a Mission on a cron queues Ideas that never become Works.
Give each Work a cadence once its first generation has completed, give Agents a heartbeat, and put the Missions that should keep proposing on a cron.
One list gathers every timer in the account — Work schedules, Mission ticks, Agent heartbeats, recurring Tasks, source validation, data sync and inbound triggers — each with its owner, cadence, next run and status, and a link to the thing that owns it.
Switch on a daily or weekly digest at Settings → Digest. It counts the agent runs that completed or failed, the tasks that moved, the pull requests opened, events from connected sources, decisions waiting on you and goal progress — from real rows, and it is not sent at all when nothing happened.
Every loop on this page is opt-in, capped, and recorded in the Activity log and in your own Git history. A few limits are worth knowing before you build a routine around them.