What Ever Works is, and how it relates to Ever Gauzy
Ever Gauzy is a business-management platform for ERP, CRM, HR and operational records. Ever Works is a separate agentic runtime for researching, generating and maintaining digital work. They sit in the same Ever Platform family and share some account and integration infrastructure, without a documented path for Works to operate on Gauzy's business records.
Someone encountering Ever's current product pages could reasonably assume Ever Works is the new AI layer for Ever Gauzy. That is not how the products are currently documented.
Gauzy is software for running ordinary business operations and storing the records those operations produce. Works is a separate agentic runtime for researching, generating and maintaining digital work. They belong to the same Ever Platform family and increasingly share account and integration infrastructure. Current documentation does not show Works operating on Gauzy's employees, invoices, CRM records or project data.
Ever Gauzy's current README now promotes Works as a newly released product. It does not give a dated launch announcement, and the Works repository has been public since January 2025. The reason to look at the pair now is that Works is being actively promoted as part of the Ever product family, and the documented relationship between the two is changing quickly.
The operational system
Official Gauzy material describes an open business-management platform covering ERP, CRM, human resources, applicant tracking, project and work management, time and activity tracking, sales, billing and inventory. Those functions sit inside one role: Gauzy is the system in which an organisation would keep operational records and run the processes those records support.
It is the older product, and the hosted SaaS is still labelled Alpha and testing in current official material. Gauzy should be read as a developing operational platform, not as a mature commercial ERP in the sense of SAP, Dynamics or NetSuite.
The runtime Ever is promoting
Current Works documentation calls it an open agentic runtime, or a workshop for AI. The product is organised around Missions, Ideas, Works and persistent Agents. Official pages describe a system that can research a topic, generate code and content, keep that output in Git, deploy it, and continue the work on a schedule. They say it can create and maintain websites, blogs, directories, landing pages and Git repositories. The same runtime can be reached through a dashboard, an API, a command-line interface or MCP, and it can use more than one AI provider.
Ever also says its Agents can operate continuously, and that the longer-term direction includes running business activity. Those are Ever's claims. What the current documentation actually specifies is more concrete: a Work is a persistent project, Agents can be given permissions, budgets and a heartbeat, and generated code and content can live in Git repositories the organisation controls. Owning those repositories does not remove every dependency on the platform. It does mean Works is treating output as ordinary version-controlled artefacts rather than as pages locked inside a proprietary builder.
Works' current repository and documentation offer an AGPLv3 Community Edition alongside commercial licensing, and they describe hosted and self-hosted operation. The hosted service is labelled pre-release. Ever's GitHub organisation profile still contains older wording, including a commercial-not-OSS line for Works and the earlier directory-builder description. The current product sites are the better guide to what Works is trying to become.
How Ever currently connects the family
Ever currently lists Gauzy and Works as separate products inside Ever Platform. The pricing page lists them as specialised platforms and as parts of the wider Ever Platform bundle. That is enough to establish a commercial and family relationship.
Ever Connect documents a shared integration catalogue that can span several Ever products. Currently active integrations include the installation address, usage-statistics linking, usage reporting, webhooks, a Gauzy cloud-billing link, and Works application and deployment status. Several further integrations are marked coming soon, including Ever ID sign-in, counterparty lookup and discoverability, public-profile import, managed operations, provider access and marketplace installs. Some of those future items name both Gauzy and Works.
Those flows are genuine platform integration, limited to the fields each definition allows. The same documentation states that the shared scopes do not include content, documents, task or project data, repositories, budgets, agents or memory. The Works status integration is explicit that only identifiers and status move.
Current Gauzy documentation gives a clearer example of a direct product relationship: Ever Teams is described as connecting to Gauzy's headless APIs. That is a documented application-and-data relationship. Works does not have an equivalent documented path into Gauzy's operational records.
Ever also lists Ever Gauzy AI as a separate stealth product. The public site and repository were not available at the time of writing. It is not Ever Works.
What organisations should actually examine
Ever appears to be assembling two layers: conventional operational systems that hold business process and data, and an agentic runtime designed to research, create, maintain and execute digital work. Those layers are moving under a shared identity, consent and control plane. The more ambitious picture, in which autonomous agents directly operate the ERP, CRM or HR system as one product, is not what the current documentation describes.
If vendors keep building those layers separately, the evaluation questions are ordinary ones. Which product owns the underlying business record? What can agents see, and what can they change? Which integrations exist today, and which are still marked coming soon? What audit and permission boundaries apply?
Both products are still developing. Ever says it is building the wider platform in public, and that some things may be incomplete, missing or broken. For organisations looking at this kind of stack, the current documentation is useful because it is more specific than the marketing. A shared account and consent layer can exist without giving agents access to the operational records underneath. That is the kind of boundary that sits inside ordinary AI integration work: what an agent is allowed to see, and what it is allowed to change.