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.
Compare · Blog builders
AI SEO content platform & technical-SEO agent
Arvow — formerly Journalist AI — researches keywords, writes SEO articles and publishes them into a CMS you already run, then fixes technical SEO on the live site. Ever Works researches the topic, generates the site itself into repositories you own, and keeps it improving on a schedule — content and code in your Git.
Arvow — the product that was Journalist AI until its 2025 rebrand — is a hosted AI SEO content platform, and the first thing to understand is that it builds nothing: you connect it to a website you already run on WordPress, Shopify, Ghost, Webflow, Wix, Framer, Squarespace, Blogger or GoHighLevel, and it fills that site with content. Inside that scope it is thorough: automatic keyword research on every plan, long-form SEO articles with generated images and internal and external links, scheduled auto-publishing, an Auto Indexer that pushes new URLs to search engines, an AI SEO Editor that rewrites articles you have already published, and an AI SEO Agent that crawls the connected site and applies technical fixes such as schema, canonicals, redirects and alt text. Ever Works starts from the opposite end: it researches the subject, generates the site itself from a Next.js or Astro template into three Git repositories you own, deploys it, and re-runs that pipeline on a schedule with named Agents, an approval queue and quality gates doing the follow-up work — while being honest that its blog kind is still early, that the pipeline is item-shaped today, and that long-form posts are commissioned as Agent Tasks until a dedicated post writer lands. It also has no path into a CMS it did not generate. Arvow wins on SEO content volume into an existing site; Ever Works wins on shipping the site, owning it in Git, and running it as an open-source, self-hosted operation.
Arvow’s own material is explicit that it does not build or host websites — its output is articles and technical fixes pushed into a CMS you operate. Ever Works produces the site as the deliverable: every Work gets three Git repositories in your own GitHub account — data, markdown and website — and the website repository is a Next.js or Astro codebase cloned from a maintained template into that account, deployed to Vercel, a Kubernetes cluster you control or managed hosting. A Blog Work defaults to the web template, a Next.js marketing site with a Markdown blog, and the platform’s directory template shows where the SEO scaffolding goes: 21 locales with hreflang, six JSON-LD types, canonicals, a per-locale sitemap, RSS, Atom and JSON feeds and an llms.txt manifest, all generated as code rather than injected into someone else’s CMS. Two honest caveats. The blog kind is early: the generation pipeline that runs for it is the same items pipeline every Work gets, so a Blog Work’s Posts tab is fed by item-shaped research today, and a long-form post is commissioned as a Task for a Work-scoped Agent until the dedicated post writer lands. And Ever Works has no publishing path into a site it did not generate, so if you will not migrate, Arvow serves you and Ever Works does not yet.
Arvow’s autoblog runs on daily, weekly or custom schedules, and its pages describe editing an article before or after it is published; no approvals queue or staging step is documented. Its keyword research and clustering are real, and Ever Works has no equivalent yet — the search providers that ship do web discovery, not keyword metrics. What Ever Works runs instead is a research pipeline as the build itself: it analyzes the domain, searches the web, extracts and deduplicates entries, validates their sources and organizes them into categories and tags, with the brand, style, personas and glossary documents from the Work’s Git-backed Knowledge Base injected into every run. That pipeline re-runs on seven cadences from hourly to monthly (available cadences can depend on your cloud plan), retries a failed run after fifteen minutes, pauses the schedule after repeated failures and recovers stuck runs on its own, and named Agents wake on their own heartbeats to take the next useful action. The autonomy has gates around it: a new Agent queues every action with side effects for a human until you loosen its guardrails, quality gates run the real build, test, lint and type-check commands and send red work back to the Agent instead of to you, a Task can run in its own branch and private checkout, and a merge policy you set per Work decides whether an Agent may land its own pull request. Ever Works does not yet score a published post for decay and rewrite it in place the way Arvow’s SEO Editor does; its refresh is a scheduled re-run of the whole pipeline.
Arvow is proprietary and hosted only: articles live in its system and are pushed into your CMS, there is no repository, no code output and no self-hosting, and the REST API and webhooks arrive with the Business plan. To its credit, whatever it has already published stays in your CMS if you leave; what stops is the automation. Ever Works commits both the content and the generated code to repositories in your own GitHub account, and even the Knowledge Base your Agents read from lives in the Work’s data repository. The runtime is open source under AGPLv3 with no paid feature tier, runs on Docker or Kubernetes, uses the model provider you choose — including a local one — and the same Works, Missions, Ideas and Knowledge Base are drivable from a REST API with your own keys, a CLI, the chat rail inside the dashboard and an MCP server your AI client can call.
Arvow is single-purpose by design: it produces SEO content and technical fixes for one kind of asset. Ever Works organizes around a goal — a Mission proposes Ideas, accepted Ideas become Works, and Agents with a scope, a budget and a heartbeat keep each one moving, with spend caps enforced before every AI call. A blog can be one Work beside a directory, a marketing site or a landing page, all researched and maintained by the same standing team, and one intent can grow into a portfolio of self-maintaining sites (and, on the roadmap, stores and whole companies).
Ever Works vs Journalist AI (now Arvow), head to head.
| Feature | Ever Works | Journalist AI (now Arvow) |
|---|---|---|
| Prompt → working app/site | ✓ | – |
| Researches the topic for you | Web research; item-shaped today | Keyword-led |
| Writes the content (copy, posts, SEO) | Items pipeline today; post writer rolling out | Long-form SEO articles |
| Generated code from a template base | ✓ | – |
| Ongoing autonomous maintenance | Scheduled | Autoblog + SEO Agent |
| Content version-controlled in Git | ✓ | – |
| Code in your Git | ✓ | No code output |
| Open source | ✓ | – |
| Self-hostable runtime | ✓ | – |
| No vendor lock-in | ✓ | Published posts stay in your CMS |
| Publishes into a CMS you already run | – | 9 platforms + Zapier |
| Existing-post refresh | Scheduled re-run | SEO fixes |
| Keyword research & content calendar | – | ✓ |
| Image generation for posts | – | ✓ |
Choose Journalist AI (now Arvow) when you already run a site on WordPress, Shopify, Ghost, Webflow, Wix, Framer, Squarespace, Blogger or GoHighLevel, you are not going to migrate it, and the job is SEO content at volume: automatic keyword research, long-form articles with images and internal links in any of 150+ languages, scheduled publishing straight into that CMS, indexing submission, technical-SEO fixes applied to the live pages, rewrites of posts you have already published, and a tracker for whether AI answer engines cite you. It is also the better pick for agencies that need white-label reports, sub-accounts and a managed human-plus-AI option, for anyone who wants to try three articles without a card or a repository, and — plainly — for anyone whose requirement is “publish into the site I have”, which Ever Works cannot do today.
Visit Journalist AI (now Arvow) ↗Give Ever Works a topic and its agents research it, generate the site and its content, deploy it, and keep re-running on the cadence you set — all version-controlled in your own Git. A dedicated long-form post writer is rolling out.