Which integrations are available for this server?

Allows sending images to OpenAI vision models via the OpenAI Responses or Chat Completions API, returning text descriptions.

How do I use j-can-see?

1. Click on "Deploy Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@j-can-see Describe the screenshot in my clipboard." That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

j-can-see

by PichurChill

Overview Schema Source Related Servers Discussions Admin

TypeScript

Local

j-can-see

English | 中文文档

An MCP server that gives text-only AI coding agents a vision toolkit: describe/OCR images, locate elements by pixel coordinates, diff images, pick exact colors, and vectorize graphics.

Why

Text-only models cannot read image files. j-can-see exposes vision as normal MCP tools, so Claude Code, Codex, and other MCP clients can work with local files, URLs, clipboard images, and screenshots without multimodal input support.

Related MCP server: MCP Vision Server

Requirements

Node.js >= 20
No installation required; the commands below run the published npm package via npx

Claude Code

One command (recommended)

claude mcp add j-can-see -s user \
    -e J_SEE_TOKEN='your-key' \
    -e J_SEE_BASE_URL='https://your-vision-endpoint' \
    -e J_SEE_MODEL='grok-4.5' \
    -- npx -y j-can-see

-s user writes to the user-level config, outside any git repository.

Manual

Add this to the mcpServers section of ~/.claude.json (macOS/Linux) or %USERPROFILE%\.claude.json (Windows):

"j-can-see": {
  "command": "npx",
  "args": ["-y", "j-can-see"],
  "env": {
    "J_SEE_TOKEN": "your-key",
    "J_SEE_BASE_URL": "https://your-vision-endpoint",
    "J_SEE_MODEL": "grok-4.5"
  }
}

Codex

One command (recommended)

codex mcp add j-can-see \
    --env J_SEE_TOKEN='your-key' \
    --env J_SEE_BASE_URL='https://your-vision-endpoint' \
    --env J_SEE_MODEL='grok-4.5' \
    -- npx -y j-can-see

Manual

Add this to ~/.codex/config.toml (macOS/Linux) or %USERPROFILE%\.codex\config.toml (Windows):

[mcp_servers.j-can-see]
type = "stdio"
command = "npx"
args = ["-y", "j-can-see"]

[mcp_servers.j-can-see.env]
J_SEE_TOKEN = "your-key"
J_SEE_BASE_URL = "https://your-vision-endpoint"
J_SEE_MODEL = "grok-4.5"

Config file locations

Client	macOS / Linux	Windows
Claude Code MCP	`~/.claude.json`	`%USERPROFILE%\.claude.json`
Claude Code hooks/settings	`~/.claude/settings.json`	`%USERPROFILE%\.claude\settings.json`
Codex MCP	`~/.codex/config.toml`	`%USERPROFILE%\.codex\config.toml`

Environment variables

Variable	Required	Default	Description
`J_SEE_TOKEN`	Yes	-	Vision model API key
`J_SEE_BASE_URL`	Yes	-	Vision endpoint base URL; trailing slashes are stripped
`J_SEE_MODEL`	Yes	-	Vision model name supported by your endpoint
`J_SEE_API_SPEC`	No	`responses`	`responses` / `openai` / `anthropic`
`J_SEE_REASONING`	No	`none`	Reasoning effort; only used by the `openai` spec
`J_SEE_MAX_EDGE`	No	`1568`	Max long-edge pixels for image compression
`J_SEE_MAX_BYTES`	No	`52428800`	Max source file size in bytes
`J_SEE_MAX_PIXELS`	No	`40000000`	Max decoded pixels, checked from the header before decode
`J_SEE_TIMEOUT_MS`	No	`90000`	Total budget per vision tool call in ms (incl. queueing and retries); timeouts auto-retry at lower resolution with a note in the result
`J_SEE_OCR_TOTAL_TIMEOUT_MS`	No	`85000`	Total budget for multi-chunk `ocr_long` in ms; returns partial results when exhausted
`J_SEE_MAX_CONCURRENT`	No	`3`	Global vision concurrency cap (1-8); auto-demotes on 429/5xx/timeout, probes back up on new calls
`J_SEE_MAX_ATTEMPTS`	No	`3`	Max attempts per vision call (first try + retries)
`J_SEE_TASK_BUDGET_MS`	No	`85000`	Total budget for `see_image` each-batch mode in ms; returns partial results + resume params when exhausted
`J_SEE_SKILL_AUTO_INSTALL`	No	`1`	Set to `0` to disable automatic skill installation

The server still starts without J_SEE_TOKEN / J_SEE_BASE_URL / J_SEE_MODEL: local pixel tools keep working, and vision tools return a clear config error when called.

API spec

`J_SEE_API_SPEC`	Endpoint	Use case
`responses` (default)	`/v1/responses`	OpenAI Responses, aligned with GPT-5 / Codex ecosystem
`openai`	`/v1/chat/completions`	OpenAI Chat Completions and OpenAI-compatible proxies
`anthropic`	`/v1/messages`	Anthropic native API, no proxy needed

If your endpoint returns 404 for /v1/responses, set J_SEE_API_SPEC=openai. To call Anthropic directly:

claude mcp add j-can-see -s user \
    -e J_SEE_API_SPEC='anthropic' \
    -e J_SEE_TOKEN='sk-ant-...' \
    -e J_SEE_BASE_URL='https://api.anthropic.com' \
    -e J_SEE_MODEL='claude-sonnet-4-5' \
    -- npx -y j-can-see

Tools

Vision tools

Tool	Purpose
`see_image`	Describe or compare images, zoom into regions
`locate`	Find one target and return its pixel coordinates
`inspect`	Enumerate all elements of one type with text and coordinates
`ocr_long`	OCR tall screenshots in chunks, merged with a dedup audit

Local tools (no vision config needed)

Tool	Purpose
`crop`	Crop a region to a file
`image_diff`	Pixel diff percentage and densest changed grid cells
`colors`	Exact dominant or candidate colors
`trace`	Vectorize flat high-contrast graphics to SVG
`extract_fg`	Cut a foreground icon into a transparent PNG

Detailed usage methodology: SKILL.md.

Agent skill

The server auto-installs SKILL.md into ~/.claude/skills, ~/.codex/skills, ~/.agents/skills, and ~/.zcode/skills on every start, unless J_SEE_SKILL_AUTO_INSTALL=0 is set.

npx j-can-see --skill        # install the skill manually
npx j-can-see --print-skill  # print SKILL.md contents
npx j-can-see --hook         # print the Claude Code hook script

Optional: Claude Code hook

Without the hook, a text-only model may try to Read an image file and fail. The hook redirects Read calls on image files to see_image:

mkdir -p ~/.claude/hooks
npx j-can-see --hook > ~/.claude/hooks/block-image-read.mjs
chmod +x ~/.claude/hooks/block-image-read.mjs

Then add this to ~/.claude/settings.json (%USERPROFILE%\.claude\settings.json on Windows):

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Read",
        "hooks": [
          {
            "type": "command",
            "command": "node ~/.claude/hooks/block-image-read.mjs"
          }
        ]
      }
    ]
  }
}

Limitations

Linux clipboard is not supported; use a file path instead
Transparent PNGs are converted to JPEG
No retries or fallback: vision failures are reported as-is

Development

npm install
npm test
npm run build

Available Tools

1 tool

see_imageA

读取图片（本地文件 / URL / 剪贴板 / 最近截图）并通过视觉模型返回文字描述。用于主模型无多模态输入能力时的识图。

Inputs: source (required), prompt.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

Deploy Server

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A4.1/5.0

Scored 2026-08-13 11:30 across 1 tool

Disambiguation5/5

With only a single tool, there is no possibility of confusing it with any other tool. The tool's purpose is clearly defined around image understanding.

Naming Consistency5/5

The tool follows a clear verb_noun pattern (see_image), and since it is the only tool, the naming is internally consistent.

Tool Count3/5

A single tool feels thin for a server, but the scope is narrow (image description). It is borderline but not an extreme mismatch.

Completeness5/5

The tool covers all input methods mentioned (local file, URL, clipboard, recent screenshot) and fulfills the stated purpose of describing images for models without multimodal input.

Resources

Related MCP Connectors

FrenchieA
io.github.Lab94
OCR, transcription, file extraction, and image generation for AI agents via MCP.
AI & Machine Learning Image & Video Processing Speech Processing
72
https://mcp.getfrenchie.dev
mcp-qwen-image
io.github.AceDataCloud
MCP server for Qwen Image 3 AI image generation
AI & Machine Learning Image & Video Processing
1
https://qwen-image.mcp.acedata.cloud/mcp
mcp-openai
io.github.AceDataCloud
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
AI & Machine Learning Image & Video Processing
11
https://openai.mcp.acedata.cloud/mcp
yourimageshareA
io.github.yourimageshare
MCP server exposing the YourImageShare upload API (upload/list/delete) as tools for AI agents.
Cloud Storage Image & Video Processing
511
https://yourimageshare.com/mcp

Related MCP Servers

llm-vision-mcp
AI & Machine Learning Image & Video Processing
ghbalf
A
license
Not graded
quality
D
maintenance
An MCP server that enables any LLM to describe images from file paths, URLs, or base64 data by forwarding them to a supported vision provider such as OpenAI, Anthropic, or local Ollama models.
Updated 7 months ago (2026-04-21 22:53 UTC)
224 npm
10
MIT
MCP Vision Server
Image & Video Processing AI & Machine Learning
rosecat2359
F
license
Not graded
quality
D
maintenance
A versatile MCP server that adds vision capabilities (image analysis, OCR, image/video generation) to AI models lacking native vision, with support for multiple providers and automatic task routing.
Updated 5 months ago (2026-06-24 13:23 UTC)
-
ollama-vision-mcp
Image & Video Processing AI & Machine Learning
ducnm9
F
license
A
quality
D
maintenance
MCP server enabling LLM clients without vision capability to process images by delegating to local Ollama vision models. Supports describing images, OCR, asking questions, and processing clipboard images.
Updated 4 months ago (2026-07-06 09:07 UTC)
4
-
mcp-vision
Image & Video Processing AI & Machine Learning
MikeyBeez
A
license
A
quality
D
maintenance
MCP server for vision capabilities, enabling screenshot, camera, and image analysis using Ollama vision models.
Updated 10 months ago (2026-01-14 19:30 UTC)
4
1
MIT

j-can-see

Why

Requirements

Claude Code

One command (recommended)

Manual

Codex

One command (recommended)

Manual

Config file locations

Environment variables

API spec

Tools

Vision tools

Local tools (no vision config needed)

Agent skill

Optional: Claude Code hook

Limitations

Development

Available Tools

Tool Schema Changelog

TDQS

Resources

Related MCP Connectors

Related MCP Servers

llm-vision-mcp

MCP Vision Server

ollama-vision-mcp

mcp-vision