← Catalog
agent resources
Free tools for AI agents
A roundup of free tools an AI agent or its developer can use today, no signup on any of them. Most sit in this catalog: an MCP server, JSON endpoints, a machine-readable index. Two are outside resources we use ourselves and link after checking they answer. One item is a paid script pack, marked as such. Every link on this page returned HTTP 200 on 2026-09-09.
Tools in this catalog, free
mcp
dataset-mcp, an MCP server with 5 tools
An MCP server that wraps this catalog. Five tools: list_datasets, get_dataset_info, get_sample, query_dataset, and get_stats. query_dataset filters and paginates the full released files (19 MB across 22 files), 100 rows per call, with =, contains, gt, and lt operators, so a query against 90,169 rows downloads nothing on your side. get_stats returns row counts, numeric min/max/mean, and top-N frequencies.
Access: npx --allow-git=all -y github:jayjex/dataset-mcp (the flag is for npm 12+, npm 10 and 11 run the plain form). Listed in the official MCP registry as io.github.jayjex/dataset-mcp, status active. Repo: github.com/jayjex/dataset-mcp. Setup walkthrough and a captured session: MCP server for datasets. Every tool with real output examples: MCP tools list. Config JSON and registry entries for all three of our servers in one place: MCP tools for Claude.
Price: free.
http
Five sample.json endpoints
Each of the 5 datasets exposes /data/<slug>/sample.json, which returns {slug, name, columns[], row_count, records[]}: the schema plus the first 10-25 rows. Same rows as CSV at sample.csv. No key, no auth, plain static files.
The slugs: nfl-games, hud-fmr-2026, airbnb-six-cities, earn-bounties, scraper-pack.
curl -s https://jayjex.github.io/data-vault/data/hud-fmr-2026/sample.json | jq '.columns, .records[0]'
Price: free. Endpoint docs with a Python example: agent access. Each sample listed with its columns and row count on the sample data downloads page.
http
catalog.json, a machine-readable index
The discovery document: every dataset with slug, niche, stats (rows, tables, columns), sample URLs, license, and the full_data link where a paid pack exists. An agent that fetches this one file knows the whole catalog without rendering a page.
curl -s https://jayjex.github.io/data-vault/catalog.json | jq '.datasets[] | {slug, niche, rows: .stats.rows}'
Price: free. Direct fetch: catalog.json.
http
llms.txt at the site root
A plain-text catalog for LLM crawlers: what the site is, how to fetch the catalog, one line per dataset with row counts and sample URLs. If your crawler reads llms.txt, this site is already described in it.
Curious about the format? The llms.txt guide breaks the structure down with this file as the example.
Price: free. Direct fetch: llms.txt.
Outside resources, verified live
Two links we rely on ourselves. Both checked by direct HTTP fetch on 2026-09-09.
registry
Official MCP registry
registry.modelcontextprotocol.io is the official registry of MCP servers. It has an HTTP API (/v0/servers) so an agent can search it programmatically. Our own server is in there as io.github.jayjex/dataset-mcp, status active, published 2026-09-07.
Price: free.
docs
Model Context Protocol documentation
modelcontextprotocol.io hosts the protocol docs: what MCP is, how clients and servers talk, and how to wire a server into Claude Desktop or another MCP client. Start here if dataset-mcp is your first server.
Price: free.
One paid item, marked as such
dev-tools
Web Scraping Script Pack
Ten self-contained Playwright scripts for pagination, infinite scroll, login sessions, price monitors, sitemap crawling, and retries. Node 18+. The sample.json endpoint lists every script with its purpose, so you can check the inventory before buying. If you would rather build the scraper yourself from free parts, the web scraping tools comparison covers the free and self-hosted options.
Price: on Getly. Free sample: data/scraper-pack/sample.json.