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.
Control & safety · Audit trail
Autonomy is only comfortable when you can reconstruct what happened while you were not watching. Ever Works writes two records of everything it does: rows in the platform — generations, deployments, imports, plugin and schedule changes, member invitations, sign-ins — and commits in the repositories you own. The first tells you what ran, when, and how it ended. The second tells you exactly what changed, line by line.
Each surface answers a different question. They read the same history from different angles — account-wide, one Work, one run, one line of code.
The Log view at /activity lists everything across your account, newest first, 25 to a page. Filter by action type, by status — pending, in progress, completed, failed, cancelled — and by a search that matches the summary or the Work name. Read it as a table or as a board grouped into status columns; a generation still running carries a red Stop button.
Each Work has its own Activity tab with a merged timeline: platform log rows, that Work’s generation history, and events from its deployed site. Category chips narrow it to Generation, Items, Deployment, Settings, Comparisons, Community PR, Users, Submissions, Reports or Sync, and the feed refreshes itself while you watch.
Below the feed, the External activity panel shows what your connectors brought in for that Work — commits and pull requests, tracker issues, docs pages, chat messages, meeting recordings. Every row names the plugin that produced it and, where the connector captured one, links back to the original in the source system.
Open a run at /agents/sessions and you get its timeline: message, tool-call and file counts, the list of files it touched, and expandable tool-call rows with the arguments and results — secrets redacted. Terminal output is stored too, redacted at ingest and kept for the retention window your plan sets.
A Work’s History tab records what actually changed, not just that a run happened: items added, updated and removed, comparison pages created or deleted, category, tag and collection edits, and items merged from community pull requests — filterable by activity type and expandable per entry.
Settings → Usage & Credits answers "where did the credits go?" over one period — a calendar month or a rolling 7 or 30 days — broken down four ways: per day, by model, by Agent, and by Work. Export CSV downloads every usage event in the period you are looking at.
The path from "something looks wrong" to the exact diff that caused it is always the same, whatever went wrong.
The dashboard leads with what is actually broken: an Agent auto-paused after repeated failures, a generation that could not be built, a Task blocked on input, an account past its budget cap. Click the one you care about. A failed or paused schedule does not surface here yet — find it in the Schedules view at /activity?view=schedules, or on the Work’s own Schedule tab.
Go to that Work’s Activity tab for its own timeline, or to /activity and filter the global Log by type and status. Filters and page number live in the URL, so the filtered view you found is a link you can paste to a colleague.
Open the run from Sessions (/agents/sessions) to read its detail at /agents/sessions/:runId — the timeline of tool calls and the list of files it touched — then Open terminal to replay the run’s stored stdout: the same stream the machine printed, with secrets redacted before it was ever written down.
The change itself is a commit on the task’s branch in your own repository, usually arriving as a pull request. Diff it, comment on it, revert it — with your normal Git tools, on infrastructure you control.
The platform log is the operational record: what ran, when it ran, whether it succeeded, and what it cost. Git is the substantive record: what the runtime actually wrote. Neither is a summary written after the fact — both are produced as the work happens, which is why they still agree six months later. Two honest limits worth knowing before you rely on them: the Log’s Type dropdown offers fewer types than the log records, and there is no date-range picker yet — narrow by type or status, or export and filter in a spreadsheet.
See what your Works are actually doing — traffic, generation runs, refresh history, agent activity.
Learn more →Run many agents at once from one cockpit — live status, current activity, tokens and cost — and steer, interrupt, or resume any of them.
Learn more →Watch an agent work in real time. Attach, detach, reconnect — and read back the whole session afterwards.
Learn more →Work refreshes, Mission ticks, Agent heartbeats, recurring Tasks and webhook triggers — every clock and every hook in one Schedules view.
Learn more →Every run is metered in credits, recorded in a ledger you can read — with daily free credits, plan entitlements, and free runs on your own keys.
Learn more →Code and content both live in your own Git, and the platform is open source under AGPLv3 — nothing is locked in.
Learn more →Describe what you want to build. Ever Works researches it, ships it, and keeps it improving — content and code, owned in your own Git.