MCP.Market

Pixlint MCP server

by amitsingh-24·io.github.amitsingh-24/pixlint·v1.1.0
No reviews yet

Lint, curate & prepare computer-vision datasets from your AI assistant — MCP server, 67 tools

B80/100grade B
Write a reviewInstall
What users say
No reviews yet
Be the first
Safety scan
B80/100

full report

Adoption
New

0 stars

Reviews

Write one

Nobody has reviewed Pixlint yet.

If you have run it, two minutes of your experience saves the next person an afternoon.

Pixlint tools (67, 5 write)

write = sends, deletes, buys or posts

Read from the package source without running it. The installed server may list more.

  • analyze_distribution_tool

    Analyze class distribution, spatial heatmap, label co-occurrence, complexity

  • analyze_quality_tool

    Analyze image quality (blur, exposure, noise, contrast, resolution, color)

  • augment_dataset_tool

    Apply augmentation pipeline to a dataset

  • auto_label_tool

    Auto-label images with a pretrained detector (COCO-80), producing a NEW pre-annotated dataset. Requires the [torch] extra.

  • auto_tag_dataset_tool

    Auto-tag dataset images with ResNet ImageNet classes

  • batch_video_to_datasets_tool

    Convert all videos in a directory to datasets

  • check_integrity_tool

    Check dataset integrity (corrupt images, missing labels, annotation bounds)

  • clean_dataset_toolwrite action

    Create a NEW cleaned dataset: drop corrupt images, clip out-of-bounds boxes, drop degenerate boxes, optionally remove duplicates.

  • cluster_dataset_tool

    Cluster dataset images using embeddings

  • compute_channel_stats_tool

    Compute dataset channel-wise mean and std

  • compute_embeddings_tool

    Compute image embeddings using ResNet50 or CLIP

  • compute_statistics_tool

    Compute dataset statistics (image sizes, class counts, annotation density)

  • convert_format_tool

    Convert dataset between formats (COCO, VOC, YOLO, KITTI, CSV)

  • dataset_diff_tool

    Compare two datasets and show differences

  • dataset_health_score_tool

    Compute overall dataset health score

  • dataset_info

    Get information about a loaded dataset

  • dataset_readiness_report_tool

    The 'Dataset Doctor': run the full diagnostic battery and return a prioritized, EXECUTABLE remediation plan plus a runnable pipeline JSON.

  • detect_format_tool

    Auto-detect the format of a dataset directory

  • detect_leakage_tool

    Detect data leakage between train/test splits

  • detect_outliers_tool

    Detect outlier images in dataset

  • discover_slices_tool

    Discover weak (under-represented or low-quality) data slices across class/brightness/size/aspect so you know what to collect or augment next.

  • diversity_sampling_tool

    Select diverse images for labeling using embedding distances

  • enrich_metadata_tool

    Enrich dataset metadata with captions and tags

  • execute_pipeline_toolwrite action

    Execute a pipeline on a dataset

  • execute_template_toolwrite action

    Execute a pipeline template on a dataset

  • export_cvat_xml_tool

    Export dataset to CVAT XML format

  • export_dataset_tool

    Export dataset in various formats (pytorch, tensorflow, ultralytics, hdf5)

  • export_fiftyone_tool

    Export dataset to FiftyOne format

  • export_huggingface_tool

    Export dataset to Hugging Face `datasets` format and optionally publish it to the Hub. Requires the [huggingface] extra; HF_TOKEN env for pushing.

  • export_labelme_json_tool

    Export dataset to LabelMe JSON format

  • export_webdataset_tool

    Export dataset to WebDataset tar format

  • extract_frames_tool

    Extract frames from a video file

  • filter_dataset_toolwrite action

    Create a NEW filtered subset dataset (by class, box area, presence of annotations, or explicit image IDs).

  • find_duplicates_tool

    Find duplicate images (exact, perceptual hash, SSIM, semantic)

  • find_label_errors_tool

    Find likely-MISLABELED annotations via embedding neighbor-consistency (a 'data linter'). Requires the [torch] extra.

  • generate_captions_tool

    Generate captions for dataset images using BLIP or tags using ResNet

  • generate_kfold_tool

    Generate K-fold cross-validation splits

  • get_template_tool

    Get a pipeline template by ID

  • list_cloud_objects_tool

    List objects in a cloud storage bucket

  • list_datasets

    List all loaded datasets

  • list_pipelines_tool

    List all registered pipelines

  • list_templates_tool

    List pre-built pipeline templates

  • load_cloud_dataset_tool

    Load a dataset from cloud storage (S3, GCS, Azure)

  • load_dataset

    Load a computer vision dataset from a directory

  • load_pipeline_tool

    Load a pipeline definition from a JSON file

  • merge_datasets_tool

    Merge multiple datasets into one with label unification and deduplication

  • normalize_dataset_tool

    Normalize dataset images using channel mean/std

  • plot_distribution_tool

    Generate class distribution chart (bar, pie, treemap, radar)

  • plot_duplicate_groups_tool

    Generate duplicate group analysis charts

  • plot_quality_scores_tool

    Generate quality score distribution chart

  • plot_spatial_heatmap_tool

    Generate spatial object distribution heatmap

  • preview_augmentation_tool

    Preview augmentation results for sample images

  • preview_images_tool

    Preview dataset images in a grid with annotation overlays

  • preview_single_image_tool

    Preview a single image with annotation details

  • query_dataset_tool

    Query a dataset with structured predicates (class/size/quality/position/object-count/text-similarity) and return matching image IDs. The agent translates a natural-language request into `filters`. Keys: classes, exclude_classes, min_annotations, max_annotations, min_width, max_width, min_height, max_height, min_area, max_area, quality_issues, min_quality, max_quality, position (left|right|top|bott

  • query_strategy_tool

    Combined active learning query strategy

  • register_pipeline_tool

    Register a pipeline definition for reuse

  • remap_classes_toolwrite action

    Create a NEW dataset with classes renamed/merged/dropped per the given old->new label mapping.

  • resize_dataset_tool

    Resize dataset images (letterbox, stretch, or crop)

  • sample_dataset_tool

    Sample image IDs from a dataset (random or stratified)

  • save_pipeline_tool

    Save a pipeline definition to a JSON file

  • semantic_search_tool

    Search images by semantic similarity to a text query

  • split_dataset_tool

    Split dataset into train/val/test sets

  • temporal_split_tool

    Temporal split of video frames (train/val/test)

  • uncertainty_sampling_tool

    Select images for labeling using uncertainty sampling

  • unload_dataset_tool

    Unload a dataset from memory and remove it from the registry

  • video_to_dataset_tool

    Convert video to CVDataset of frames

Public scan report

scanner v0.1.9 · 2026-09-24 · same rubric, same numbers if you re-run it

1 medium
  • Code scan61 source files scanned20/25
  • –Live reliabilityno gateway calls yet and no remote to proben/a
  • –Tool poisoningtools not inspected (local package is not executed); not countedn/a
  • Auth qualitylocal package, no credentials required12/15
  • Maintenancelast push 46 days ago12/15
  • Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Findings (1)

  • mediumeval / new Function usedexec.eval
    pixlint-1.1.0/src/pixlint/analysis/autolabel.py: …esnet50_fpn(weights=weights).to(device).eval() class_counts: dict[str, int] = {}…
Overall 80/100. Components that don't apply are left out of the denominator. Any critical finding is an F.RubricAppeal a findingJSON

How Pixlint compares

Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.

Safety score
80/100
higher than 60% of 48,255 servers
Tools
67
the typical server has 7
Tools that change things
7%
typical: 0%. Fewer write tools means less an agent can break

What its tools let an agent do: 57 read · 4 create · 4 change · 2 run or deploy

Install Pixlint in Claude Code, Cursor or VS Code

claude mcp add pixlint -- uvx pixlint
Add to Cursor

What the publisher says

From the Pixlint repository's README, as published. We do not edit it. Read it on GitHub

PixLint

Lint, curate, and prepare computer-vision datasets — right from your AI assistant.

PixLint is an MCP server that gives AI assistants — Claude, Cursor, VS Code, and any MCP client — direct, conversational access to a complete computer-vision dataset toolkit: analyze quality, find duplicates and label errors, clean and curate, split, augment, convert formats, and export to every major training framework.

It runs locally over stdio, or self-hosted on the internet over authenticated HTTP.

Why PixLint

Most dataset tooling is either a paid SaaS or a heavy GUI app. PixLint is a single, open-source, self-hostable server an AI agent can drive end to end — and it does things others keep behind paid tiers:

  • 🩺 Dataset Doctor — one call runs a full diagnostic and returns a prioritized, executable fix plan.
  • Label-error detection — automatically surface images that are probably mislabeled.
  • Natural-language query — "find blurry images with a person on the left", answered over your data.
  • Weak-slice discovery — find under-represented or low-quality slices to collect or augment next.
  • Curation that writes a new dataset — clean / filter / remap, not just report.
  • Auto-labeling with a pretrained detector, and one-command Hugging Face publishing.

Features

103 operations — 67 tools, 23 resources, 13 prompts.

Quick Start

1. Install

pip install pixlint

Optional extras add heavier capabilities:

pip install "pixlint[torch]"        # embeddings, auto-labeling, label-error detection
pip install "pixlint[huggingface]"  # Hugging Face export + publishing
pip install "pixlint[all]"          # everything

2. Connect your AI assistant

Claude Desktop — claudedesktopconfig.json:

{
  "mcpServers": {
    "pixlint": {
      "command": "pixlint",
      "env": { "CV_DATA_DIR": "/path/to/your/datasets" }
    }
  }
}

Cursor / VS Code — .cursor/mcp.json or .vscode/mcp.json:

{
  "mcpServers": {
    "pixlint": {
      "command": "pixlint",
      "env": { "CV_DATA_DIR": "/path/to/your/datasets" }
    }
  }
}

CVDATADIR is the directory PixLint is allowed to read datasets from.

3. Just ask

"Load my dataset at /data/cocoperson, give it a readiness report, then clean it and export for YOLO."*

Shortened. The full README is on GitHub.

Nothing above is checked by us. What we check is on the safety report.

Pixlint: common questions

Is Pixlint MCP server safe?
Mostly: it is graded B (80/100). Read the Pixlint safety report
How do I install Pixlint?
It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
Does Pixlint need an API key?
No secret keys are declared. It reads 2 settings from the environment.
Is Pixlint maintained?
The last commit was 87 days ago (2026-08-09). The latest release is v1.1.0.
How does Pixlint compare with other MCP servers?
safety score 80/100 (higher than 60% of 48,255 servers); tools 67 (the typical server has 7); tools that change things 7% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.