license	cc-by-4.0
task_categories	text-generation text-classification
language	en
size_categories	1K<n<10K
tags	mcp model-context-protocol ai-agents llm-tooling agent-tooling structured-data claude llm-training
pretty_name	MCP Servers Catalog

MCP Servers Catalog

A structured, machine-readable catalog of 2,223 MCP (Model Context Protocol) servers extracted from the curated awesome-mcp-servers list. Covers 49 categories with language, scope, OS, and officiality metadata for each server.

Updated monthly. Last updated: 2026-05.

Dataset Summary

This dataset catalogs what each MCP server is: its name, repository URL, category, description, implementation language, deployment scope (cloud/local), supported operating systems, and whether it carries an official or community-quality signal (official flag, Glama badge, npm install hint).

Existing MCP directories (glama.ai, mcp.so) are web-browsable but not machine-queryable. This dataset fills the gap: a flat, Parquet-native catalog an AI agent or researcher can load in one line and filter, rank, or embed without scraping.

Intended Uses

Agent tooling at inference time. An AI agent selecting an MCP server for a task can load this catalog as context — filter by category, language, and scope before recommending an install command.

LLM training and fine-tuning. Ground models in the current MCP server landscape. Reduces hallucination of server names, capabilities, and install paths.

Research. Quantitative analysis of the MCP ecosystem: language distribution, category coverage, official vs. community server ratio, cloud vs. local split.

Developer reference. "What MCP servers exist for database access?" becomes a one-liner (see Sample Queries).

Dataset Statistics

Metric	Count
Total servers	2,223
Categories	49
Official servers	122
Servers with Glama quality badge	857
Servers with npm install hint	84

Top categories: Developer Tools (327), Finance & Fintech (273), Other Tools and Integrations (162), Knowledge & Memory (159), Search & Data Extraction (140)

Languages: TypeScript (1,114), Python (776), Go (131), Rust (56), Java (23), C# (17)

Scope: Cloud (1,243), Local (1,124), Embedded (16)

Files

File	Description
servers.json	Canonical record-per-server, UTF-8 JSON array
servers.parquet	Columnar output, Snappy-compressed, HuggingFace dataset viewer-ready
extract.py	Extraction script - run to regenerate

Schema

Field	Type	Notes
name	string	Server name as listed (often owner/repo format)
url	string	Primary repository or project URL
source_host	string	github, gitlab, npm, or other
category	string	Top-level category from the curated list (e.g. Developer Tools, Databases)
description	string	One-line description from the listing
languages	list[string]	Implementation languages: typescript, python, go, rust, csharp, java, c_cpp, ruby
scopes	list[string]	Deployment scope: cloud, local, embedded
operating_systems	list[string]	Supported OS: macos, windows, linux
is_official	bool	true if the server carries the official implementation flag
has_glama_badge	bool	true if a Glama quality score badge is present
glama_server_path	string	Glama server path (e.g. owner/repo) — empty if no badge
npm_install	string	Extracted npx ... install command from description — empty if none

Note on list fields in Parquet: languages, scopes, and operating_systems are stored as JSON strings (e.g. '["typescript","python"]'). Parse with json.loads().

Sample Queries

Find all Python MCP servers for databases (pandas):

import pandas as pd, json
df = pd.read_parquet("servers.parquet")
df["langs"] = df["languages"].apply(json.loads)
db_python = df[
    (df["category"] == "Databases") &
    df["langs"].apply(lambda l: "python" in l)
]
print(db_python[["name", "description"]].to_string())

List official servers with npm install commands:

from datasets import load_dataset
ds = load_dataset("automatelab/mcp-servers-catalog", split="train")
official_npm = [r for r in ds if r["is_official"] and r["npm_install"]]
for r in official_npm:
    print(r["npm_install"], "—", r["description"][:60])

Count servers by category and language (DuckDB):

SELECT category,
       COUNT(*) AS total,
       COUNT_IF(has_glama_badge) AS quality_badged
FROM read_parquet('servers.parquet')
GROUP BY category
ORDER BY total DESC
LIMIT 15;

Find all cloud-native TypeScript servers:

import pandas as pd, json
df = pd.read_parquet("servers.parquet")
df["scopes"] = df["scopes"].apply(json.loads)
df["langs"] = df["languages"].apply(json.loads)
cloud_ts = df[
    df["scopes"].apply(lambda s: "cloud" in s) &
    df["langs"].apply(lambda l: "typescript" in l)
]
print(f"{len(cloud_ts)} cloud TypeScript servers")

Methodology

The catalog is extracted from the awesome-mcp-servers README using extract.py. The script:

1. Reads the README (local path or fetched from GitHub).
2. Walks the ## Server Implementations section, tracking ### category headers.
3. For each - [name](url) ... entry, parses emoji flags for language/scope/OS metadata, strips Glama badge markup, and splits the description from the flags region.
4. Deduplicates on URL (keeps first occurrence of any URL collision).
5. Emits servers.json (UTF-8 JSON array) and servers.parquet (Snappy-compressed columnar).

The script is idempotent: re-running on the same README produces identical output.

Update Cadence

Updated monthly in sync with the upstream awesome-mcp-servers list. The Last Updated field at the top of this card tracks the most recent extraction run.

Source

Upstream data: AutomateLab-tech/awesome-mcp-servers — a curated community list of MCP server implementations.

License

Our additions (catalog format, extract.py, this dataset card, and editorial framing) are licensed under CC-BY-4.0.

Upstream server metadata is derived from the awesome-mcp-servers community list. Each server's source code is under its own license (typically MIT). This dataset is a catalog/index of publicly listed server metadata, not a redistribution of source code.

Attribution: Upstream server listings from the awesome-mcp-servers community list. Individual server source code under each project's own license.

Maintainer

AutomateLab - AI automation guides and tools.