Query the plot catalogue from an AI assistant via the Model Context Protocol. Transport: Streamable HTTP at https://api.anyplot.ai/mcp/ (POST JSON-RPC; the endpoint requires an MCP client and answers plain GET with an error by design). No authentication, read-only.
claude mcp add --transport http anyplot https://api.anyplot.ai/mcp/list_specs(limit, offset) — all plot specifications with tags and library countssearch_specs_by_tags(plot_type, data_type, domain, features, library, ...) — filter the catalogue by spec- and implementation-level tagsget_spec_detail(spec_id, libraries?) — one spec with its implementations (full source code; filter by library ids to keep the payload small)get_implementation(spec_id, library) — one implementation: runnable code, light/dark render URLs, quality score, review datalist_libraries() — the supported plotting librariesget_tag_values(category) — the vocabulary of any tag categoryPrefer plain HTTP? The same catalogue is served by the JSON API and indexed in llms-full.txt; retrieval recipes live in llms.txt.