Independent by design · no pay-to-rank placements  · How we evaluate tools →
Anclum Submit
W

WardLu/shadow-vision

Independent directory — not the official site. Details below come from the project's public repository.

1 ★ measured · checked 2026-11-19

Before you install: no copy-paste command · license MIT

First step: nothing runnable here yet - start from the repository: https://github.com/WardLu/shadow-vision

★ 1·updated 2026-10-17
View on GitHub ↗Direct link · no redirect page

We have neither an install command nor a public remote endpoint from this repository, so we do not print a config. A made-up one is worse than none. How each client wires MCP up →

Star count
1 ★ · WardLu/shadow-vision · read from the GitHub search API · (checked 2026-11-19 20:00:48)
Source
collected from the punkpeye/awesome-mcp-servers list · type: MCP server
Harvest history
first seen 2026-11-18 · last seen 2026-11-20 · seen 3× · listed by 3 sources
Score
score 300 · 1 ★ · 3 sources · mcp
FAQ (answers come from our records)
Who is it for?
We filed it under these tags: MCP servers. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository WardLu/shadow-vision;first catalogued 2026-11-18;last seen 2026-11-20。There is no official certification or quality rating here — we record sources and dates only.
README excerpt (English, untranslated) · Read the full README on GitHub →

# Shadow Vision

Excerpted from the repository README (abridged)

English | 简体中文 | Shadow Lab
Give text-only LLMs a pair of eyes. Shadow Vision is an open-source MCP vision service that enables AI Agents to see, understand, and analyze real-world information through vision_ocr, vision_inspect, vision_annotate, vision_layout, vision_reconstruct, and vision_compare — without switching host models.
## Why It's Different
MCP-Native: Compatible with Codex, Claude Desktop, Cursor, and other MCP clients
Pluggable Backends: Ollama, OpenAI-compatible, Anthropic, Gemini
Local-First: Keep images and inference entirely on your machine when using Ollama
Versatile Input: Supports local file paths, base64 image data, or remote HTTP(S) URLs
## How It Works
The text model calls Shadow Vision tools via MCP. Shadow Vision forwards the image and prompt to the configured vision backend and returns the text result back to the model.
## Quick Start
### 0. One-Click Run (No Clone Needed)
No need to clone the repository. Run directly:
bash
uvx shadow-vision # Python / uv users (recommended)
# Or for Node users
npx shadow-vision # Requires uv installed locally
MCP configuration example:
toml
[mcp_servers.vision]
command = "uvx"
args = ["shadow-vision"]
env = { VISION_BACKEND = "ollama", VISION_MODEL = "qwen3-vl:2b-instruct" }
> npx shadow-vision is a thin wrapper that invokes uvx shadow-vision internally and requires uv installed on your machine. Both entry points behave identically.
### 1. Install from Source (Development / Self-Hosting)
Requires Python 3.11+ and uv:
bash
git clone https://github.com/WardLu/shadow-vision.git
cd shadow-vision
uv sync
### 2. Use Local Ollama (Recommended for Beginners)
First install Ollama. If not running the Ollama desktop app, start the service manually:
bash
ollama serve
ollama pull qwen3-vl:2b-instruct
ollama list
qwen3-vl:2b-instruct is the default vision model (non-thinking version, faster response). If you need stronger reasoning/thinking capability, switch to qwen3-vl:2b (thinking version); or change VISION_MODEL to any other vision model listed in ollama list.
### 3. Register as an MCP Service
Codex can run directly:
bash
codex mcp add vision -- uv run shadow-vision
Or add to ~/.codex/config.toml:
toml
[mcp_servers.vision]
type = "stdio"
command = "uv"
args = ["run", "shadow-vision"]
cwd = "/path/to/shadow-vision"
env = { VISION_BACKEND = "ollama", VISION_MODEL = "qwen3-vl:2b-instruct", OLLAMA_URL = "http://127.0.0.1:11434/api/chat" }
After restarting your MCP client, simply ask the model to "take a look at this image".
## Switching Models and Backends
VISION_BACKEND determines how requests are routed, and VISION_MODEL specifies the vision model. Update the environment variables in your MCP configuration and restart your client.
Switch to a local model:
toml
env = { VISION_BACKEND = "ollama", VISION_MODEL = "your-downloaded-vision-model", OLLAMA_URL = "http://127.0.0.1:11434/api/chat" }
Switch to an OpenAI-compatible service:
toml
env = { VISION_BACKEND = "openai_compatible", VISION_MODEL = "provider-vision-model-name", OPENAI_API_BASE = "https://api.example.com/v1", OPENAI_API_KEY = "sk-...", OPENAI_MAX_TOKENS = "1024", OPENAI_MAX_TOKENS_FIELD = "max_tokens" }
OPENAI_ denotes the OpenAI Chat Completions compatible protocol, which also applies to LM Studio, vLLM, and any other service offering /v1/chat/completions.
Listed badge (optional)
Copy this line into the README if you want the listing mark. Optional, and it does not affect being listed.
[![Listed on Anclum](https://anclum.com/badge/listed.svg)](https://anclum.com/en/s/wardlu-shadow-vision/)
Source
awesome · https://github.com/WardLu/shadow-vision
Browse by field
MCP servers
Traction
1 ★
Listed by
3 independent sources
added
today

More MCP servers

All 6,555 →

Filed under

Every tag opens the full list for that filter. Counts are the live catalogue figures.

Browse all fields →

← Back to index