Skip to content

Development integration

Hugging Face MCP server for Claude, Codex and Cursor

Search models, datasets, Spaces and papers on Hugging Face. Connect the Hugging Face MCP server to Toolstile once, then choose which of its tools Claude Code, Codex, Cursor and Claude Desktop may use.

Free plan available. Server credentials stay in Toolstile, not in agent configs.

On this page

What your AI apps can do with Hugging Face

  • Search models, datasets and papers
  • Read repository details and files

Hugging Face MCP tools

4 tools and 158 resources, listed by Toolstile from the server itself on 29 October 2026. The server decides what it offers and can add or rename tools at any time; after you connect, Toolstile shows the current list.

ToolWhat it doesSuggested rule
hf_whoamiInspect the current Hugging Face authentication context, including the account, visible organization memberships, and credential access details. Read-only and never returns credential values.Allow (read-only)
hub_repo_searchSearch Hugging Face repositories with a shared query interface. You can target models, datasets, spaces, or aggregate across multiple repo types in one call. Include links to repositories in your response.Allow (read-only)
hub_repo_detailsGet details for one or more Hugging Face repos (model, dataset, or space). Auto-detects type unless specified. For datasets, use operations: overview, dataset_structure, dataset_preview. Use dataset_structure first to discover configs, splits, sizes, and schema. Use …Allow (read-only)
hf_fsWhen to use: Hugging Face Hub models, datasets, Spaces, collections, papers, daily papers, today's trending models, current paper leaderboard, docs, and repository files.Allow (read-only)

Hugging Face MCP resources

ResourceAddressWhat it provides
hf-cliskill://hf-cli/SKILL.mdHugging Face Hub CLI (`hf`) for downloading, uploading, and managing models, datasets, spaces, buckets, repos, papers, jobs, and more on the Hugging Face Hub. Use when: handling authentication; managing local cache; managing Hugging Face Buckets; running or scheduling jobs on …
hf-cloud-aws-context-discoveryskill://hf-cloud-aws-context-discovery/SKILL.mdDiscover the user's local AWS context (active profile, region, account ID, caller identity) at the start of any AWS task. Use this skill before any other AWS work — deploying to SageMaker, creating resources, calling AWS APIs, or anything that touches an AWS account. Use it …
hf-cloud-python-env-setupskill://hf-cloud-python-env-setup/SKILL.mdSet up an isolated Python environment for SageMaker / AWS work, with the right Python version and current boto3. Use this skill whenever Python code will be executed for a SageMaker deployment, training job, or any AWS automation — including when about to run `pip install`, when …
requirements.txtskill://hf-cloud-python-env-setup/requirements.txt
check_versions.pyskill://hf-cloud-python-env-setup/scripts/check_versions.py
setup_env.pyskill://hf-cloud-python-env-setup/scripts/setup_env.py
hf-cloud-sagemaker-deployment-plannerskill://hf-cloud-sagemaker-deployment-planner/SKILL.mdPlan and coordinate the deployment of a model to Amazon SageMaker AI. Use this skill whenever the user wants to deploy, host, serve, or expose a model on SageMaker or AWS — including phrases like "deploy a model", "host this LLM on AWS", "serve this embedding model", "deploy a …
hf-cloud-sagemaker-iam-preflightskill://hf-cloud-sagemaker-iam-preflight/SKILL.mdEnsure a usable SageMaker execution role exists before deploying or training. Use this skill whenever about to create a SageMaker endpoint, model, training job, or any resource that requires an execution role. Use it especially when the user has not provided a role ARN …
minimum-permissions.jsonskill://hf-cloud-sagemaker-iam-preflight/references/minimum-permissions.json
trust-policy.jsonskill://hf-cloud-sagemaker-iam-preflight/references/trust-policy.json
check_role.pyskill://hf-cloud-sagemaker-iam-preflight/scripts/check_role.py
create_role.pyskill://hf-cloud-sagemaker-iam-preflight/scripts/create_role.py
hf-cloud-sagemaker-production-defaultsskill://hf-cloud-sagemaker-production-defaults/SKILL.mdCreate a SageMaker endpoint (real-time, real-time scale-to-zero, or async) with autoscaling, CloudWatch alarms, and tagging enabled by default. Use this skill whenever about to create a SageMaker endpoint, write deployment code that calls `create_endpoint`, or finalize a …
deployment-template.mdskill://hf-cloud-sagemaker-production-defaults/references/deployment-template.md
_common.pyskill://hf-cloud-sagemaker-production-defaults/scripts/_common.py
deploy.pyskill://hf-cloud-sagemaker-production-defaults/scripts/deploy.py
deploy_async.pyskill://hf-cloud-sagemaker-production-defaults/scripts/deploy_async.py
deploy_ic.pyskill://hf-cloud-sagemaker-production-defaults/scripts/deploy_ic.py
invoke_endpoint.pyskill://hf-cloud-sagemaker-production-defaults/scripts/invoke_endpoint.py
teardown.pyskill://hf-cloud-sagemaker-production-defaults/scripts/teardown.py
hf-cloud-serving-image-selectionskill://hf-cloud-serving-image-selection/SKILL.mdPick the right serving container for a SageMaker model deployment and find its current image URI. Use this skill whenever about to deploy a model to a SageMaker endpoint and an image URI needs to be chosen — including when the user says "deploy this LLM", "host this HuggingFace …
model-to-image.mdskill://hf-cloud-serving-image-selection/references/model-to-image.md
mirror_image.pyskill://hf-cloud-serving-image-selection/scripts/mirror_image.py
hf-memskill://hf-mem/SKILL.mdHugging Face CLI to estimate the required memory to load Safetensors or GGUF model weights for inference from the Hugging Face Hub
huggingface-bestskill://huggingface-best/SKILL.mdUse when the user asks about finding the best, top, or recommended model for a task, wants to know what AI model to use, or wants to compare models by benchmark scores. Triggers on: "best model for X", "what model should I use for", "top models for [task]", "which model runs on …
huggingface-community-evalsskill://huggingface-community-evals/SKILL.mdRun evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Use for backend selection, local GPU evals, and choosing between vLLM / Transformers / accelerate. Not for HF Jobs orchestration, model-card PRs, .eval_results publication, or …
.env.exampleskill://huggingface-community-evals/examples/.env.example
USAGE_EXAMPLES.mdskill://huggingface-community-evals/examples/USAGE_EXAMPLES.md
inspect_eval_uv.pyskill://huggingface-community-evals/scripts/inspect_eval_uv.py
inspect_vllm_uv.pyskill://huggingface-community-evals/scripts/inspect_vllm_uv.py
lighteval_vllm_uv.pyskill://huggingface-community-evals/scripts/lighteval_vllm_uv.py
huggingface-datasetsskill://huggingface-datasets/SKILL.mdUse this skill for Hugging Face Dataset Viewer API workflows that fetch subset/split metadata, paginate rows, search text, apply filters, download parquet URLs, and read size or statistics.
huggingface-gradioskill://huggingface-gradio/SKILL.mdBuild Gradio web UIs and demos in Python. Use when creating or editing Gradio apps, components, event listeners, layouts, or chatbots.
examples.mdskill://huggingface-gradio/examples.md
huggingface-llm-trainerskill://huggingface-llm-trainer/SKILL.mdTrain or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs …
gguf_conversion.mdskill://huggingface-llm-trainer/references/gguf_conversion.md
hardware_guide.mdskill://huggingface-llm-trainer/references/hardware_guide.md
hub_saving.mdskill://huggingface-llm-trainer/references/hub_saving.md
local_training_macos.mdskill://huggingface-llm-trainer/references/local_training_macos.md
reliability_principles.mdskill://huggingface-llm-trainer/references/reliability_principles.md
trackio_guide.mdskill://huggingface-llm-trainer/references/trackio_guide.md
training_methods.mdskill://huggingface-llm-trainer/references/training_methods.md
training_patterns.mdskill://huggingface-llm-trainer/references/training_patterns.md
troubleshooting.mdskill://huggingface-llm-trainer/references/troubleshooting.md
unsloth.mdskill://huggingface-llm-trainer/references/unsloth.md
convert_to_gguf.pyskill://huggingface-llm-trainer/scripts/convert_to_gguf.py
dataset_inspector.pyskill://huggingface-llm-trainer/scripts/dataset_inspector.py
estimate_cost.pyskill://huggingface-llm-trainer/scripts/estimate_cost.py
hf_benchmarks.pyskill://huggingface-llm-trainer/scripts/hf_benchmarks.py
train_dpo_example.pyskill://huggingface-llm-trainer/scripts/train_dpo_example.py
train_grpo_example.pyskill://huggingface-llm-trainer/scripts/train_grpo_example.py
train_sft_example.pyskill://huggingface-llm-trainer/scripts/train_sft_example.py
unsloth_sft_example.pyskill://huggingface-llm-trainer/scripts/unsloth_sft_example.py
huggingface-local-modelsskill://huggingface-local-models/SKILL.mdUse to select models to run locally with llama.cpp and GGUF on CPU, Mac Metal, CUDA, or ROCm. Covers finding GGUFs, quant selection, running servers, exact GGUF file lookup, conversion, and OpenAI-compatible local serving.
hardware.mdskill://huggingface-local-models/references/hardware.md
hub-discovery.mdskill://huggingface-local-models/references/hub-discovery.md
quantization.mdskill://huggingface-local-models/references/quantization.md
huggingface-lora-space-builderskill://huggingface-lora-space-builder/SKILL.mdBuild and publish a Gradio demo on Hugging Face Spaces for a user-provided LoRA. Use when someone asks to create, generate, ship, or publish a Space, demo, Gradio app, or playground for a LoRA — including LoRAs for Qwen-Image, Qwen-Image-Edit, LTX-Video, Wan, FLUX, SDXL, or …
adapting-to-the-lora.mdskill://huggingface-lora-space-builder/references/adapting-to-the-lora.md
krea-2.mdskill://huggingface-lora-space-builder/references/base-models/krea-2.md
ltx.mdskill://huggingface-lora-space-builder/references/base-models/ltx.md
qwen-image.mdskill://huggingface-lora-space-builder/references/base-models/qwen-image.md
creative-mode.mdskill://huggingface-lora-space-builder/references/creative-mode.md
tasks.mdskill://huggingface-lora-space-builder/references/tasks.md
zerogpu-and-publishing.mdskill://huggingface-lora-space-builder/references/zerogpu-and-publishing.md
huggingface-paper-publisherskill://huggingface-paper-publisher/SKILL.mdPublish and manage research papers on Hugging Face Hub. Supports creating paper pages, linking papers to models/datasets, claiming authorship, and generating professional markdown-based research articles.
example_usage.mdskill://huggingface-paper-publisher/examples/example_usage.md
quick_reference.mdskill://huggingface-paper-publisher/references/quick_reference.md
paper_manager.pyskill://huggingface-paper-publisher/scripts/paper_manager.py
arxiv.mdskill://huggingface-paper-publisher/templates/arxiv.md
ml-report.mdskill://huggingface-paper-publisher/templates/ml-report.md
modern.mdskill://huggingface-paper-publisher/templates/modern.md
standard.mdskill://huggingface-paper-publisher/templates/standard.md
huggingface-papersskill://huggingface-papers/SKILL.mdLook up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata such as authors, linked models/datasets/spaces, Github repo and project page. Use when the user shares a Hugging Face paper page URL, an arXiv URL or ID, or asks to summarize, …
README.mdskill://huggingface-spaces/README.md
huggingface-spacesskill://huggingface-spaces/SKILL.mdBuild, deploy, and maintain applications on Hugging Face Spaces — Gradio / Docker / Static SDKs, ZeroGPU and dedicated hardware, model loading, debugging, buckets, inference providers, community grants. Use whenever the user asks to create or host an app on Hugging Face, port …
3d-cuda-extensions.mdskill://huggingface-spaces/references/3d-cuda-extensions.md
3d-generation.mdskill://huggingface-spaces/references/3d-generation.md
3d-gsplat.mdskill://huggingface-spaces/references/3d-gsplat.md
3d-models.mdskill://huggingface-spaces/references/3d-models.md
3d-outputs.mdskill://huggingface-spaces/references/3d-outputs.md
buckets.mdskill://huggingface-spaces/references/buckets.md
debugging.mdskill://huggingface-spaces/references/debugging.md
gradio.mdskill://huggingface-spaces/references/gradio.md
grants.mdskill://huggingface-spaces/references/grants.md
inference-providers.mdskill://huggingface-spaces/references/inference-providers.md
known-errors.mdskill://huggingface-spaces/references/known-errors.md
requirements.mdskill://huggingface-spaces/references/requirements.md
zerogpu.mdskill://huggingface-spaces/references/zerogpu.md
huggingface-tool-builderskill://huggingface-tool-builder/SKILL.mdUse this skill when the user wants to build tool/scripts or achieve a task where using data from the Hugging Face API would help. This is especially useful when chaining or combining API calls or the task will be repeated/automated. This Skill creates a reusable script to fetch, …
baseline_hf_api.pyskill://huggingface-tool-builder/references/baseline_hf_api.py
baseline_hf_api.shskill://huggingface-tool-builder/references/baseline_hf_api.sh
baseline_hf_api.tsxskill://huggingface-tool-builder/references/baseline_hf_api.tsx
find_models_by_paper.shskill://huggingface-tool-builder/references/find_models_by_paper.sh
hf_enrich_models.shskill://huggingface-tool-builder/references/hf_enrich_models.sh
hf_model_card_frontmatter.shskill://huggingface-tool-builder/references/hf_model_card_frontmatter.sh
hf_model_papers_auth.shskill://huggingface-tool-builder/references/hf_model_papers_auth.sh
huggingface-trackioskill://huggingface-trackio/SKILL.mdTrack and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API), firing alerts for training diagnostics, or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, alerts with webhooks, HF Space …
alerts.mdskill://huggingface-trackio/references/alerts.md
logging_metrics.mdskill://huggingface-trackio/references/logging_metrics.md
retrieving_metrics.mdskill://huggingface-trackio/references/retrieving_metrics.md
huggingface-vision-trainerskill://huggingface-vision-trainer/SKILL.mdTrains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging …
finetune_sam2_trainer.mdskill://huggingface-vision-trainer/references/finetune_sam2_trainer.md
hub_saving.mdskill://huggingface-vision-trainer/references/hub_saving.md
image_classification_training_notebook.mdskill://huggingface-vision-trainer/references/image_classification_training_notebook.md
object_detection_training_notebook.mdskill://huggingface-vision-trainer/references/object_detection_training_notebook.md
reliability_principles.mdskill://huggingface-vision-trainer/references/reliability_principles.md
timm_trainer.mdskill://huggingface-vision-trainer/references/timm_trainer.md
dataset_inspector.pyskill://huggingface-vision-trainer/scripts/dataset_inspector.py
estimate_cost.pyskill://huggingface-vision-trainer/scripts/estimate_cost.py
image_classification_training.pyskill://huggingface-vision-trainer/scripts/image_classification_training.py
object_detection_training.pyskill://huggingface-vision-trainer/scripts/object_detection_training.py
sam_segmentation_training.pyskill://huggingface-vision-trainer/scripts/sam_segmentation_training.py
huggingface-zerogpuskill://huggingface-zerogpu/SKILL.mdAI demos and GPU compute with Gradio Spaces and Hugging Face Spaces ZeroGPU. Use when writing or reviewing code that uses `@spaces.GPU`, configuring `python_version` or `requirements.txt` for a ZeroGPU Space, or handling ZeroGPU-specific code constraints — pickle-based process …
concurrency.mdskill://huggingface-zerogpu/references/concurrency.md
cuda-and-deps.mdskill://huggingface-zerogpu/references/cuda-and-deps.md
how-quota-works.mdskill://huggingface-zerogpu/references/how-quota-works.md
how-zerogpu-works.mdskill://huggingface-zerogpu/references/how-zerogpu-works.md
train-sentence-transformersskill://train-sentence-transformers/SKILL.mdTrain or fine-tune sentence-transformers models across `SentenceTransformer` (bi-encoder, dense or static embedding model for retrieval, similarity, clustering, classification, paraphrase mining, dedup, multimodal), `CrossEncoder` (reranker, pair scoring for two-stage retrieval …
base_model_selection.mdskill://train-sentence-transformers/references/base_model_selection.md
dataset_formats.mdskill://train-sentence-transformers/references/dataset_formats.md
evaluators_cross_encoder.mdskill://train-sentence-transformers/references/evaluators_cross_encoder.md
evaluators_multi_vector_encoder.mdskill://train-sentence-transformers/references/evaluators_multi_vector_encoder.md
evaluators_sentence_transformer.mdskill://train-sentence-transformers/references/evaluators_sentence_transformer.md
evaluators_sparse_encoder.mdskill://train-sentence-transformers/references/evaluators_sparse_encoder.md
hardware_guide.mdskill://train-sentence-transformers/references/hardware_guide.md
hf_jobs_execution.mdskill://train-sentence-transformers/references/hf_jobs_execution.md
losses_cross_encoder.mdskill://train-sentence-transformers/references/losses_cross_encoder.md
losses_multi_vector_encoder.mdskill://train-sentence-transformers/references/losses_multi_vector_encoder.md
losses_sentence_transformer.mdskill://train-sentence-transformers/references/losses_sentence_transformer.md
losses_sparse_encoder.mdskill://train-sentence-transformers/references/losses_sparse_encoder.md
model_architectures.mdskill://train-sentence-transformers/references/model_architectures.md
prompts_and_instructions.mdskill://train-sentence-transformers/references/prompts_and_instructions.md
training_args.mdskill://train-sentence-transformers/references/training_args.md
troubleshooting.mdskill://train-sentence-transformers/references/troubleshooting.md
mine_hard_negatives.pyskill://train-sentence-transformers/scripts/mine_hard_negatives.py
train_cross_encoder_distillation_example.pyskill://train-sentence-transformers/scripts/train_cross_encoder_distillation_example.py
train_cross_encoder_example.pyskill://train-sentence-transformers/scripts/train_cross_encoder_example.py
train_cross_encoder_listwise_example.pyskill://train-sentence-transformers/scripts/train_cross_encoder_listwise_example.py
train_multi_vector_encoder_example.pyskill://train-sentence-transformers/scripts/train_multi_vector_encoder_example.py
train_sentence_transformer_distillation_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_distillation_example.py
train_sentence_transformer_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_example.py
train_sentence_transformer_make_multilingual_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_make_multilingual_example.py
train_sentence_transformer_matryoshka_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_matryoshka_example.py
train_sentence_transformer_multi_dataset_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_multi_dataset_example.py
train_sentence_transformer_static_embedding_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_static_embedding_example.py
train_sentence_transformer_with_lora_example.pyskill://train-sentence-transformers/scripts/train_sentence_transformer_with_lora_example.py
train_sparse_encoder_distillation_example.pyskill://train-sentence-transformers/scripts/train_sparse_encoder_distillation_example.py
train_sparse_encoder_example.pyskill://train-sentence-transformers/scripts/train_sparse_encoder_example.py
transformers-jsskill://transformers-js/SKILL.mdUse Transformers.js to run state-of-the-art machine learning models directly in JavaScript/TypeScript. Supports NLP (text classification, translation, summarization), computer vision (image classification, object detection), audio (speech recognition, audio classification), and …
CACHE.mdskill://transformers-js/references/CACHE.md
CONFIGURATION.mdskill://transformers-js/references/CONFIGURATION.md
EXAMPLES.mdskill://transformers-js/references/EXAMPLES.md
MODEL_ARCHITECTURES.mdskill://transformers-js/references/MODEL_ARCHITECTURES.md
MODEL_REGISTRY.mdskill://transformers-js/references/MODEL_REGISTRY.md
PIPELINE_OPTIONS.mdskill://transformers-js/references/PIPELINE_OPTIONS.md
TEXT_GENERATION.mdskill://transformers-js/references/TEXT_GENERATION.md
trl-trainingskill://trl-training/SKILL.mdPost-train LLMs with TRL (Transformers Reinforcement Learning) — SFT, DPO, GRPO, KTO, and reward-model training. Use when writing or debugging training code with the TRL Python API or the trl CLI.

Suggested access for each AI app

  • 4 tools marked read-only (hf_whoami, hub_repo_search, hub_repo_details, hf_fs). These are usually safe to allow.

These are hints the server reports about its own tools, not guarantees.

Coding agent in Claude Code or Cursor. Allow the read tools so the agent can look things up without stopping, and require approval for tools that change shared data.

Chat or research app in Claude Desktop. Allow reading and searching. Allow writes only where you want the app to act for you.

Unattended agent, such as a scheduled Codex job. No one is watching the session, so keep it narrow: allow what the job needs, require approval for changes and block the rest.

Blocked tools are removed from the tool list an app sees, so it won't try them. Because rules apply per app, the same Hugging Face connection can be broad in one app and read-only in another, and the credential stays in Toolstile instead of each app's config file.

How it works with Toolstile

  1. Connect once. Add Hugging Face to a workspace (no sign-in or access token). Toolstile keeps the credential, so no agent config holds it.
  2. Choose access per app. Give Claude Code every Hugging Face tool, give Cursor only the read tools, or keep Hugging Face away from an app entirely.
  3. Share approved accounts. Teammates use the connection without copying tokens, and you can remove access in one place.
  4. Review activity. See which app called which tool and when, so you can tighten access with evidence.

Add the Hugging Face MCP server to an AI app

Through Toolstile, every AI app uses one gateway entry, whichever servers you connect. In Claude Code:

Shell
claude mcp add --transport http toolstile 'https://toolstile.com/mcp'

Then connect Hugging Face in your workspace and choose which of its tools that app may use. Client setup covers Codex, Cursor and Claude Desktop.

To connect directly instead, add the Hugging Face endpoint to each app separately. In Claude Code:

Shell
claude mcp add --transport http huggingface 'https://huggingface.co/mcp'

In Cursor's mcp.json:

JSON
{
  "mcpServers": {
    "huggingface": {
      "url": "https://huggingface.co/mcp"
    }
  }
}

Some servers accept only clients they have approved, or need a token header. Check the Hugging Face MCP documentation for what your client needs.

Before you connect

No account needed for public search. Use a Hugging Face access token to reach your own and gated content.

Choose no sign-in for public tools, or enter a Hugging Face access token.

Server details

DetailValue
PublisherHugging Face
Endpointhttps://huggingface.co/mcp
AuthenticationNo sign-in or access token
CategoriesDevelopment, Marketing & research
Tool list4 tools, observed 29 October 2026 (discovery)

Hugging Face MCP documentation