Agent Skills · v2.0.1 · MIT · Zero config
43 Skills. One install. Zero waste.
Your AI agent shouldn't scrape Wikipedia to tell you the weather. 43 structured skills. Instant data. No server needed.
$
gemini extensions install https://github.com/nikhilreddy3888/public-api-toolkit
Weather
Crypto
Wikipedia
Open Data
The problem
AI is solving easy problems the hard way.
Every time an AI assistant browses for weather, hallucinates exchange rates, or scrapes news HTML for a one-line factual answer — it burns tokens and time that a structured API call would solve in 200ms.
What agents do today (expensive)
- Search the web for BTC price — burns 800+ tokens parsing finance pages
- Reason about FX rates from stale training data — answer might be months old
- Scrape weather.com HTML for tomorrow's forecast — brittle, slow, wasteful
- Fetch full Wikipedia pages just to summarize one paragraph
- Install a different setup for every single capability
What they should do (instant)
- Use
cryptoskill — structured JSON in <300ms - Use
currencyskill — live rates, zero hallucination - Use
weatherskill — clean forecast, 0 HTML parsing - Use
wikipediaskill — structured summary, minimal tokens - One toolkit covers all 43 skills
User asks
What AI often does
What it should do
Skill name
What's the weather in Tokyo?
Search, scrape, summarize — 800+ tokens
Call forecast endpoint, return clean conditions
weather
Convert €450 to USD
Guess from training data or browse
Use live FX APIs with exact rates
currency
What's Bitcoin trading at?
Search finance sites or refuse
Fetch live quote with market context
crypto
Is today a holiday in France?
Google search + HTML parse. Unreliable.
Call holiday API, get definitive answer
holidays
Total install time
43 separate configs
One install. Every skill.
43-in-1
Why it matters
Less tokens. Less compute. More signal.
Every avoidable web search is wasted compute — on the GPU that generates it, the data center that processes it, and the energy that powers it. Structured API calls return exactly what the agent needs in a fraction of the tokens, time, and electricity.
⚡ 10–40x fewer output tokens
Structured JSON replaces pages of parsed HTML. The model spends tokens reasoning — not reading noise.
🌱 Less compute, less energy
Fewer generated tokens means less GPU time, less data center load, and a smaller carbon footprint per lookup.
🔒 Safe by default
Every upstream is vetted. No scraping brittle websites, no secrets in prompts, no user data in transit. Pure markdown-based skill definitions.
🤝 Built with contributors
This project thrives because contributors keep adding safe, useful, free APIs. Every new skill means one fewer reason for agents to scrape the web. Add yours →
43 skills · 12 categories
Stop scraping. Start calling.
Related capabilities are grouped into domain skills. No server code required — agents discover and use them directly.
weather
Weather & Sky
Forecasts, current conditions, alerts, sun times
air_quality
Air & Carbon
Air quality, AQI, carbon intensity
currency
Currency Exchange
Live FX rates, 150+ currencies, historical series
crypto
Crypto Markets
Bitcoin, altcoins, market cap, 24h change
country_data
Countries & Regions
Country profiles, flags, populations, capitals
geocoding
Geocoding
Forward geocoding, reverse geocoding, postal lookups
ip_geolocation
IP Geolocation
IP lookup for city, region, timezone, network
wikipedia
Wikipedia
Structured summaries, search, Wikidata
dictionary
Dictionaries & Words
Definitions, synonyms, rhymes, autocomplete
dev_tools
Developer Tools
npm registry, GitHub search, CDNJS libraries
entertainment
Movies & Fandom
Star Wars, Rick & Morty, Harry Potter, MCU
space
Space & Earth
NASA, ISS tracking, Mars photos, SpaceX
holidays
Holidays
Public holidays and bank holidays
transport
Transport
Flight tracking, bike share, EV charging
security_intel
Security Intel
Vulnerability, Shodan, and threat feeds
food_recipes
Food & Drink
Recipes, nutrition, breweries
mock_data
Mocking & Prototyping
Fake users, products, posts for testing
anime
Anime & Ghibli
Quotes, facts, and Studio Ghibli films
coding_contests
Competitive Coding
LeetCode, Codeforces, HackerRank schedules
avatar_generator
Avatars
Robots, pixel art, identicons from seeds
tech_news
Tech News
HackerNews top stories, Dev.to articles
...and 22 more
See all 43 skills
Books, universities, name insights, random facts, animals, pokemon, and more
Client support
Works where you already work.
Designed for the agent ecosystem people are actually using. Gemini CLI, Claude Code, OpenClaw, Codex — each has a native install path.
Get started in 30 seconds
One install. Every client.
No servers to run. No keys needed. Pure agent-ready skills.
Gemini CLI
gemini extensions install https://github.com/nikhilreddy3888/public-api-toolkit
Claude Code
claude plugin marketplace add nikhilreddy3888/public-api-toolkit
OpenClaw
openclaw plugins install https://github.com/nikhilreddy3888/public-api-toolkit
Codex
codex install-skill https://github.com/nikhilreddy3888/public-api-toolkitFAQ
Common questions.
Do I need API keys for all 43 skills?
+
No. All 43 tools work without any API keys — they use free public APIs
that don't require authentication. The toolkit is built to be zero-config.
How is this different from just letting the agent use web search?
+
Web search returns noisy HTML that costs 500–2000+ tokens per lookup.
Structured API calls return clean JSON in <300ms using 50–100 tokens.
It's more accurate and significantly cheaper.
Is this an MCP server?
+
No. It is a Skill Pack. While MCP servers require running a local Node/Python process,
Skills are pure Markdown instructions that agents interpret. This means zero overhead,
zero local execution risks, and instant setup.
Can I contribute new APIs?
+
Yes! Adding a skill takes about 10 minutes. Check the Contributing Guide.
Get started
Stop scraping. Start calling.
43 agent skills for structured public data. Give your agents the data they need without the token waste.