CSS to Image API

What the HTML/CSS to Image API MCP server does

The HTML/CSS to Image API MCP server gives compatible AI clients access to HCTI’s hosted rendering service. It supports several input styles: inline HTML and CSS, public web pages, selected elements on a page, and saved HCTI templates populated with variable values. Outputs can be images or PDFs, depending on the rendering request.

The package also includes guidance for image-generation workflows and dynamic Open Graph images. The Open Graph skill can create persistent social-preview configurations and use their image URLs in website metadata. That skill is distributed separately from the main skills directory for clients that should not discover duplicate skills.

How it works

The repository is an Agent Plugin package containing MCP and client-specific manifests. The hosted endpoint is https://mcp.hcti.io; clients connect to it as a remote Streamable HTTP MCP server rather than starting a local rendering process.

Authentication uses HCTI’s browser-based OAuth flow. The repository does not contain an API key or access token. Rendering requires an HCTI account and available credits, and usage or account limits can be checked through the MCP tools.

Once the client is connected and authorized, an agent can call the HCTI tools directly. Claude Code, Cursor, Cline, Gemini CLI, ChatGPT/Codex, GitHub Copilot, Grok Build, Kimi Code, and OpenClaw have repository-specific installation or configuration guidance, though availability may depend on the relevant marketplace or client integration.

Setup and configuration

For clients supporting remote MCP servers, add https://mcp.hcti.io as a Streamable HTTP server and complete the OAuth authorization in a browser. No local server installation is required. The repository includes portable manifests such as plugin.json, mcp.json, and .mcp.json, along with client-specific manifests and installation guidance.

Cursor users can install the package from its marketplace when published, then connect hcti in MCP settings. Cline can use the supplied installation guide or configure the remote server manually. Gemini CLI installs the extension from GitHub and uses /mcp auth hcti. Claude Code can install the community plugin or load a local checkout. ChatGPT and Codex use the published HTML/CSS to Image plugin. Local development checkouts are also documented for several clients.

Tools and capabilities

The HTML/CSS to Image API MCP server exposes these core tools:

• create_image renders supplied HTML and CSS as an image or PDF.
• create_url_image captures a public URL or a selected page element.
• create_templated_image renders a saved HCTI template with supplied variables.
• create_batch_images creates multiple image variations in one request.
• get_max_batch_size reports the account’s current batch limit.
• create_template saves a reusable HTML/CSS template.
• update_template changes an existing template.
• list_templates lists templates in the authorized account.
• list_template_versions shows a template’s version history.
• check_usage checks current organizational image usage and limits.

These tools cover one-off rendering, URL screenshots, repeatable template output, batch generation, and basic template and usage management.

Limitations and notes

The service is hosted, so this repository does not provide a local browser renderer. A browser OAuth authorization is required, and rendering depends on an HCTI account with available credits. The Open Graph skill and the main rendering skill have separate distribution paths; installing a standalone skill alone does not establish the MCP connection. The repository documents some integrations as pending marketplace publication, so installation steps may differ by client and release status.