IIIF MCP Server

Overview

The IIIF MCP Server v1.1.0 bridges the gap between AI assistants and the rich ecosystem of IIIF (International Image Interoperability Framework) resources. By exposing a unified API surface, it allows Claude and other AI clients to seamlessly query, navigate, and retrieve high‑resolution images, metadata, annotations, and collection structures from any IIIF‑compliant repository. This eliminates the need for custom integration code, enabling developers to focus on higher‑level AI logic rather than low‑level IIIF protocol handling.

At its core, the server implements full search and metadata retrieval capabilities. It supports the IIIF Content Search API across versions 0, 1, and 2, allowing AI assistants to perform full‑text queries within manifests or collections. Metadata extraction pulls from the IIIF Presentation API (v2 and v3), offering flexible property selection, multilingual labels, and robust handling of nested or HTML‑rich metadata fields. These features empower AI agents to answer contextual questions about artifacts, provenance, and collection hierarchies without additional parsing logic.

The image operations layer is equally comprehensive. Clients can request images with precise region, size, rotation, quality, and format parameters, thanks to fully validated URL construction. The server also provides image info retrieval for both v2 and v3, ensuring compatibility with a wide range of IIIF image servers. A standout addition in v1.1.0 is the ability to fetch raw image data as Base64, with automatic size constraints (max 1500 px or 1 M pixels) and percentage‑based region extraction—ideal for embedding images directly into AI responses or generating thumbnails on the fly.

Beyond basic retrieval, the server offers advanced collection management and annotation operations. Developers can list nested collections, navigate parent‑child relationships, and retrieve manifests within a collection—all with support for both v2 and v3 structures. Annotation search and extraction allow AI assistants to pull text, comments, or transcriptions from resources, with multilingual support that detects and groups annotations by language. This is particularly valuable for scholarly workflows where contextual commentary or user‑generated notes must be surfaced.

Integration into AI pipelines is straightforward: the MCP server exposes standard tool definitions that Claude can invoke with natural language prompts. For example, a user might ask for the “first 10 images of the Catalogue of Ancient Manuscripts,” and the assistant will translate that into a sequence of MCP calls—searching for the collection, retrieving its manifests, and fetching the requested image tiles. The server’s single‑file bundling option (via esbuild) further simplifies deployment in environments without npm, making it a lightweight yet powerful addition to any AI‑driven digital humanities or cultural heritage platform.