Shipped
Missions
Set an ongoing goal in plain language and the platform keeps proposing Ideas that ladder up to it — one-shot or on a schedule, with budgets and guardrails so it never runs away.
Learn more →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.
Roadmap
Ever Works is the open-source AI runtime that researches, ships, and maintains business websites — content and code — and keeps working 24/7. This is the honest view of what is live today, what we are building right now, and what comes after. It lives in the website code next to the platform pages, so it moves when the product does.
Now
Everything here is live in the product today. We keep improving it — the polish work is continuous, and bugs against any of these are welcome on GitHub.
Set an ongoing goal in plain language and the platform keeps proposing Ideas that ladder up to it — one-shot or on a schedule, with budgets and guardrails so it never runs away.
Learn more →Lightweight proposals you can research and refine before committing. Suggested at sign-up, generated on a cadence, or added by you — build, dismiss, or retry them.
Learn more →The buildable unit. Create a website, landing page, blog, or directory from a prompt, a template, or an imported Awesome-list README — real code and researched content in your own Git.
Learn more →Named, persistent AI employees with scopes, heartbeats, budgets, and permissions. Start from the public agent catalog — PM, Coder, Researcher, and more — or define your own.
Learn more →Reusable, versioned Markdown recipes an agent pulls in when relevant — a curated public catalog plus your own custom skills, attached per agent.
Learn more →Group agents and people into nested teams inside an organization, assign the Works each team owns, and see the whole structure drawn out on an org chart.
Per-Work, typed, Git-backed institutional context — brand voice, SEO rules, research, decisions — that every run reads from, with a review queue for agent-authored memory.
Learn more →Scheduled updates, continuous refresh, and background workers keep adding content, improving code, and researching long after the first build — 24/7, not just on prompt.
Learn more →One trackable unit of work per Task, a private branch and checkout per run, acceptance checks that decide when it is done, and a merge policy that says who may land the pull request.
Learn more →Watch every agent run from one cockpit — status, activity, tokens, cost — stream its terminal, and steer, interrupt, or resume it while it is still in flight.
Learn more →Per-Mission, per-Idea, per-Work, per-Agent, and account-wide caps that gate spend before a call runs, on top of a readable credits ledger with daily free credits and free runs on your own keys.
Learn more →Slack, GitHub pull-request review, meetings, and native connectors land as one normalized event stream — and agents get real inbound and outbound mailboxes of their own.
Learn more →Manage Works from any MCP-compatible AI assistant, or from the terminal with the public ever-works CLI — both drive the same API you can call directly with an API key.
A capability-driven plugin system with 102 built-in plugins across 19 capability categories: AI providers and gateways, search, deployment, screenshots, content extraction, pipelines, and prompt management.
Run the whole stack from official Docker images or on a Kubernetes cluster, and deploy Works to Vercel, Kubernetes, or your own infrastructure with custom domains.
Learn more →Enroll any machine as an execution node with a one-time token — outbound-only, nothing to expose — and see it beside your own cluster nodes in one Fleet view.
Learn more →Next
Work that is under way in the platform repository. Some of it is already visible in the product behind a "coming soon" badge; none of it should be relied on until it ships.
The full stack — API, dashboard, jobs runtime, and database — as one installable app with a guided installer, able to connect outward to an external database, jobs backend, or provider. Today you run in the cloud or self-host with Docker.
Learn more →A Store kind of Work: pick a storefront template and a commerce backend, then let agents research the catalog, write product copy, and keep the storefront optimized. The Store chip is on the create surface today; the generator and commerce integrations are being built.
Learn more →Register a company as a Work through formation providers, staff it with organization-wide agents, and run its Works under one shared knowledge base. Organizations, the Company chip and v1 manual registration are live — registering records a real Organization, but nothing is filed with a registrar. Incorporation through a formation provider and the wider company-as-a-business surface are still ahead.
Learn more →One campaign brief already provisions a campaign Work, its Goal, the prebuilt go-to-market agents and the first pipeline Tasks in a single call. Still to come: the pipeline plugin does not auto-enable, there is no draft-review screen, and the campaign Work does not yet carry its own budget and schedule.
Learn more →Install plugins on demand at runtime instead of bundling every plugin into every build — the groundwork a marketplace needs.
Expanding beyond the current provider set with more model providers, gateways, and routing options, with discovery and configuration done from the dashboard rather than a config file.
Multi-step generation with validation and quality scoring before anything is published, so scheduled runs raise the bar instead of just adding volume.
Learn more →Real-time pipeline status in the dashboard and richer analytics on what each run produced and cost.
Learn more →Later
Accepted directions that are not yet scheduled. Priority is weighed by user demand, strategic fit, effort, and breaking-change risk — your issues and pull requests move these up.
A marketplace for community-contributed plugins, with better discovery, documentation, and more extension points in the plugin SDK.
Analyze screenshots and extract visual information during research, so directory and comparison pages can reason about what a site looks like, not only what it says.
Process only changed or new items on a scheduled run, making large Works cheaper to keep fresh.
Role-based access across multi-user organizations, beyond the per-Work Manager, Editor, and Viewer roles available today.
Faceted search, reviews and ratings, more payment providers and subscription options, and more built-in themes across the template catalog.
More built-in languages, stronger right-to-left support, per-Work language settings, and automated translation workflows.
The roadmap is shaped in the open. Open an issue with the problem you are trying to solve, your use case, and how you would expect it to work — check existing issues first and add your case to one that matches. Pull requests that implement a requested feature are the fastest path from idea to reality. For partnership or enterprise needs, contact us directly.
Describe what you want to build. Ever Works researches it, ships it, and keeps it improving — content and code, owned in your own Git.