Mcp Yolo

io.github.rjn32sPublished by GitHub account rjn32sv0.1.2Updated Oct 24, 2026

An MCP server providing zero-shot object detection and segmentation using Ultralytics YOLOE.

VerifiedSTDIODesktop onlyMedia & DesignAI & ML

Overview

AI-generated overview

Local MCP server that runs Ultralytics YOLOE zero-shot object detection and instance segmentation on images from text prompts.

What you can do
  • Detect objects in an image using arbitrary natural-language class prompts, without retraining.
  • Segment detected objects and return polygon masks for them.
  • Accept images as local file paths, remote URLs, or Base64-encoded strings.
When to use it
Useful when an assistant needs to find or outline objects in images by describing them in words, such as locating a specific item in a photo or producing masks for downstream image work. It suits local, desktop-based image analysis rather than hosted or web use.
Requirements
Runs as a local process on the user's machine (desktop only, not web-executable), installed from the PyPI package mcp-yolo and started with uv. It downloads and runs the YOLOE model weights (yoloe-26l-seg.pt by default), so expect model downloads and local compute; no accounts, API keys or environment variables are declared.
Before you install
Images given as remote URLs are fetched over the network, so image content leaves the machine to that host. Detection quality and speed depend on hardware and input resolution. No credentials are requested, and no write, send or delete actions are described.

Installation

In SourceWeft

  1. Open Mcp Yolo in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

MCP-YOLO

[PyPI] [Downloads] mcp-name: io.github.rjn32s/mcp-yolo

MCP-YOLO is an agent-first development platform that provides Zero-Shot Object Detection and Segmentation as a Model Context Protocol (MCP) server. Powered by Ultralytics YOLOE, it enables developers and AI agents to detect and segment objects using arbitrary text prompts without retraining.

Key Features

  • Zero-Shot Detection: Detect any object using natural language (e.g., "the blue coffee cup next to the spoon").
  • Instance Segmentation: Precise polygon masks for discovered objects.
  • Flexible Image Inputs: Supports local file paths, remote URLs, and Base64 encoded strings.
  • Agent Optimized: Includes custom "Skills" for autonomous deployment and benchmarking.

YOLOE Performance Reference

YOLOE builds upon the latest YOLO architectures (like YOLO11 and YOLO26) to provide state-of-the-art open-vocabulary performance.

ModelBased OnmAP (COCO)Speed (T4/ms)Params (M)
YOLOE26-NYOLO26-N40.91.7~3.0
YOLOE26-SYOLO26-S48.62.5~10.0
YOLOE26-LYOLO26-L55.06.2~40.0
YOLOE-LYOLO11-L~52.0~5.0~26.0

Note: Performance varies depending on the hardware and input resolution. mcp-yolo uses yoloe-26l-seg.pt by default for high precision.

Quick Start

Installation

bash
uv pip install mcp-yolo

Running the Server

bash
uv run mcp-yolo

MCP Tools

detect_objects

Performs zero-shot detection.

  • Arguments:
    • image_source (str): Path, URL, or Base64.
    • classes (list[str], optional): Custom text prompts to detect.

segment_objects

Performs zero-shot instance segmentation.

  • Arguments:
    • image_source (str): Path, URL, or Base64.
    • classes (list[str], optional): Custom text prompts to segment.

Publishing

This project is configured for automated PyPI publishing. See the pypi_setup_guide.md for details.

Source: README.md at commit 0df8eb7

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.2LatestSep 16, 2026

Score

40% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.