Give your agents custom APIs.
Turn anything you can do on a website into a reusable task, then let ChatGPT, Claude, Copilot, Codex, Cursor, Claude Code, or any MCP client create and call it.
Turn anything you can do on a website into a reusable task, then let ChatGPT, Claude, Copilot, Codex, Cursor, Claude Code, or any MCP client create and call it.
An MCP server exposes tools and context to compatible AI clients through the Model Context Protocol. Instead of building a separate integration for every agent, one MCP server can make the same capabilities available across supported clients.
It lets AI clients work with Figranium: discover templates and your tasks, create tasks, and run them. That means an agent can find reusable templates or use a Figranium task as the missing API for a website or workflow.
No. The website does not need an API or an MCP server. Figranium handles the website task, while this MCP server exposes that task to your AI client.
No. Cloud clients can connect to the remote OAuth MCP endpoint, while local clients can run the Figranium MCP package over STDIO. The connection dialog above shows the appropriate setup for each client.
Yes. Both Figranium and the Figranium MCP server are open source, and you can inspect their source on GitHub.