What harvis.dev MCP server does
The harvis.dev MCP server gives an AI agent a way to publish a static site from files. Its stated purpose is narrow: an agent supplies the files for a static site, and the service deploys them. After the operation completes, the tool returns a live URL that can be used to access the deployed site.
This fits workflows where an agent creates or updates static web content and needs to make that content available without a separate manual publishing step. The available description does not identify a framework, hosting provider, repository workflow, build system, domain configuration, or database connection. It supports describing the service as a static-site deployment endpoint exposed through MCP.
Tools and capabilities
The documented tool is deploy_site. It publishes static-site files and returns a live URL. The supplied material does not specify the tool's argument schema, accepted file representation, response fields beyond the live URL, authentication method, or error behavior.
The capability can be summarized as:
- Accept files for a static site through an AI-agent tool call.
- Publish those files as a deployed site.
- Return the URL for the resulting live site.
No additional tools are documented. In particular, the available information does not establish support for editing an existing deployment, deleting a site, configuring custom domains, viewing deployment history, or managing environment variables.
Limitations and notes
The harvis.dev MCP server's published description does not include a README, installation commands, transport details, client examples, environment variables, license information, or operational limits. Those details should be confirmed from the service before implementation.
The description also does not explain whether deployment performs a build step or expects already-built static assets. It only states that deploy_site publishes files. Treat framework compilation, asset preprocessing, authentication, domain management, and rollback behavior as unspecified unless the service documents them elsewhere.
For evaluation, the clearest supported use case is direct publication of static-site files from an AI agent. The returned live URL is the documented result that lets the agent or a user locate the deployment. No claim can be made from the available material about persistence, access control, uptime, pricing, file-size limits, or supported site types beyond static sites.