Overview

The Gyazo Image Search, Upload, OCR MCP Server is a TypeScript implementation of a Model Context Protocol (MCP) server that exposes Gyazo images and metadata to AI assistants and other MCP-aware clients. It connects to your Gyazo account using an API access token and exposes image resources via gyazo-mcp:// URIs, along with tools to search, fetch, upload, and extract OCR/text from images.

This server is useful when you need fast programmatic access to screenshots and images stored on Gyazo from within an assistant workflow. Typical uses include full-text search across image titles and descriptions, retrieving image bytes and metadata for inline context, extracting OCR text, and uploading new captures directly from an automated pipeline.

Source code and issues: https://github.com/nota/gyazo-mcp-server

Features

Exposes Gyazo captures as MCP resources with gyazo-mcp:// URIs
Full-text search across user uploads (title, description, app, URL, date ranges)
Fetch raw image bytes and structured metadata (title, description, app, permalink)
Return OCR text when available for quick text extraction from images
Upload images to Gyazo with base64 payloads and optional metadata
Works with JPEG, PNG and other standard image formats
Easy integration via npm (npx) or Docker; suitable for local development and assistant integrations

Installation / Configuration

Prerequisites:

A Gyazo account
A Gyazo API access token (generate from https://gyazo.com/api)

Set the GYAZO_ACCESS_TOKEN environment variable with your token.

Install and run via npm (recommended):

# install globally (optional)
npm install -g @notainc/gyazo-mcp-server

# run directly with npx (no global install)
GYAZO_ACCESS_TOKEN="your-access-token" npx @notainc/gyazo-mcp-server

Run with Docker:

docker run -i --rm -e GYAZO_ACCESS_TOKEN="your-access-token" nota/gyazo-mcp-server

Add to Claude Desktop (example config):

{
  "mcpServers": {
    "gyazo-mcp-server": {
      "command": "npx",
      "args": ["@notainc/gyazo-mcp-server"],
      "env": {
        "GYAZO_ACCESS_TOKEN": "your-access-token-here"
      }
    }
  }
}

Development:

# install deps
npm ci

# build
npm run build

# development watch mode
npm run watch

# optional: build Docker image
npm run image:build

Available Resources

Each Gyazo capture is exposed as a resource with a gyazo-mcp:// URI. Resource contents include:

Raw image content (bytes / data URL)
Metadata: id, title, description, app name, created_at, original Gyazo URL/permalink
OCR text if available (plain text)
MIME type and file size

Typical resource URI format: gyazo-mcp:///

Clients can request the resource to obtain both the binary image and structured metadata useful for context-aware models.

Available Tools

Tool name	Purpose
gyazo_search	Full-text search across captures (keyword, title, app, URL, date range). Supports pagination and returns matching URIs + metadata.
gyazo_image	Fetch a specific image and its metadata by ID or URL. Returns bytes and metadata.
gyazo_latest_image	Retrieve the most recent image for the configured account (includes OCR if available).
gyazo_upload	Upload an image (base64) to Gyazo with optional title, description, referer, and app name. Returns permalink URL and image ID.

Example: search call (conceptual)

{
  "tool": "gyazo_search",
  "query": "error log screenshot",
  "limit": 10,
  "page": 1
}

Result includes a list of gyazo-mcp:// URIs and metadata for each match.

Use Cases

Assistants: Provide visual context for support tickets by searching a user’s Gyazo captures and including the latest screenshot with OCR text for diagnostics.
Automated workflows: Upload generated diagrams or screenshots from CI jobs, then return the permalink as part of a deployment report.
Knowledge extraction: Run OCR over a set of screenshots and index the extracted text into a searchable knowledge base.
Quick retrieval: Build a command that fetches the most recent capture and embeds it into a chat or document with metadata and a permalink.

Notes & Best Practices

Keep the GYAZO_ACCESS_TOKEN secret and restrict its usage to trusted environments.
For bulk operations, use pagination to avoid rate limits.
OCR availability depends on Gyazo’s processing of the image; not all images will include OCR text.
Use the MCP resource URIs to refer to images consistently from assistant contexts.

For source, issues and contribution details, see the project repo: https://github.com/nota/gyazo-mcp-server

Tags:media search