Gyazo MCP Server
Gyazo MCP server: the agent finds screenshots by keyword or app, reads text on captures through OCR and uploads new ones
- 28 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The Gyazo token opens all your captures for reading, and screenshots often hold passwords, emails and customer data
- The `gyazo_upload` tool publishes images to your account and returns links
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/gyazo-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The Gyazo token opens all your captures for reading, and screenshots often hold passwords, emails and customer data
- The `gyazo_upload` tool publishes images to your account and returns links
This is third-party code. Review the repository files before installing.
What it does
The server opens the agent to a Gyazo library of screenshots. Each capture is available as a resource at a `gyazo-mcp://` address: the agent gets the image itself plus metadata, meaning title, description, app and page URL. If a capture has an OCR result, the text from the picture comes with it. There are four tools. `gyazo_search` finds captures by keyword, title, app, URL or date range and pages through results. `gyazo_image` fetches a capture by id or link. `gyazo_latest_image` returns the newest capture together with its OCR text. `gyazo_upload` uploads a base64 image, takes a title, description, source URL and app name, and returns a permanent link and an id. The server is written in TypeScript and runs over stdio. Access goes through the `GYAZO_ACCESS_TOKEN` token created in the API section on gyazo.com.
Who it is for. Designers, testers and documentation writers who keep screenshots in Gyazo and want to search them and hand them to an agent without hunting for links by hand.
Good fit when
- A bug is shown on a screenshot taken an hour ago: the agent takes the latest capture and reads the error text through OCR
- The library holds hundreds of captures and you need the one taken in a specific app last week
- The agent produced an image for a report or ticket and it has to go to Gyazo with a permanent link returned
- A guide needs captures selected by a keyword from the window title or page URL
Not a fit when
- The captures live in another service: the server works only with the Gyazo API
- You need to capture the screen right now: the server searches and uploads ready images and has no screen capture
- Captures contain passwords or customer data that must not go to an external agent
Example request
Take the latest capture from Gyazo, read the error text on it and suggest where to look for the causeMCP
- Transport
- stdio
- Authentication
- API key
Environment variables
GYAZO_ACCESS_TOKENrequired, secret- Gyazo API access token from the Your Access Token section at gyazo.com/api
Limitations
- A Gyazo account and a token from the API section are required: you register an application with a name and callback URL, and `http://localhost` works.
- Installation goes through npm or Docker, so the machine needs Node.js or Docker.
- Not every capture has an OCR result; the README treats it as optional.
- The server hands the agent full capture contents and has no read-only mode.
How to disable. Remove the `gyazo-mcp-server` block from `claude_desktop_config.json` or another MCP configuration and restart the client. The token can be revoked on the API page at gyazo.com.
FAQ
How do I get a token for Gyazo MCP?
Sign in at gyazo.com/api, click Register applications and New Application, then fill in a name and callback URL (`http://localhost` works). On the application page scroll to Your Access Token and click Generate.
How do I connect the server to Claude Desktop?
Add a `gyazo-mcp-server` block to `claude_desktop_config.json` with the `npx` command, the `@notainc/gyazo-mcp-server` argument and the `GYAZO_ACCESS_TOKEN` variable under `env`. The file is in `~/Library/Application Support/Claude/` on macOS and `%APPDATA%/Claude/` on Windows.
Which tools does the server have?
Four: `gyazo_search`, `gyazo_image`, `gyazo_latest_image` and `gyazo_upload`. The first three read captures and the last uploads a new image and returns a permanent link.
Can I run the server in Docker?
Yes. Build the image with `npm run image:build` and put `docker run -i --rm -e GYAZO_ACCESS_TOKEN gyazo-mcp-server` in the client configuration. The token goes in as an environment variable.
Does the server read text on captures?
Yes, when a capture has OCR data: `gyazo_latest_image` and the resources pass the recognized text along with the image. For other captures the agent sees only the picture and metadata.
Discussion
No questions or reviews yet.
Related
MCP serversA local MCP server for editing and analyzing images: the agent crops frames, finds objects and reads text from photos
The MCP server gives agents GIMP image editing with visual previews, layers, filters and file export operations.
Cloudglue MCP: video analysis for agents
Cloudglue MCP Server
An agent in Cursor or Claude Desktop describes video, splits it into chapters and shots and finds moments by speech or picture
Hosted Glif MCP server and plugin for creating images, video and audio from Claude Code, Codex, Cursor and ChatGPT
The official remote MCP server of Kling AI: an agent makes videos and images from text and from a picture right in the work chat
RunAPI MCP server: a catalog of 160 models for images, video, music and speech, pricing, task runs and balance
README in short
The page describes the server resources and four tools, then walks through package installation and getting a token. It ends with Claude Desktop configurations through `npx` and Docker, plus build and development commands.