Public API Access — AI Compatible

1,850 curated public APIs — trust-rated, AI-ready, and 470 KB small. The first API directory built for both humans AND AI agents.

---

🚀 Quick Start

For AI Agents (primary use case)

# Load directly into any LLM context
curl -s https://raw.githubusercontent.com/shivtcdfinance/public-api-access-ai-compatible/main/api_access.min.json

Claude / Cursor / Copilot:

Load this file: https://raw.githubusercontent.com/shivtcdfinance/public-api-access-ai-compatible/main/ai_prompt.txt
Then ask: "Find me a free HTTPS API for currency exchange rates"

Python:

import json, urllib.request

url = "https://raw.githubusercontent.com/shivtcdfinance/public-api-access-ai-compatible/main/api_access.min.json"
apis = json.loads(urllib.request.urlopen(url).read())

# Safe APIs: trusted + no auth + HTTPS
safe = [a for a in apis if a['trust'] == 'trusted' and a['auth'] == 'no' and a['https']]
weather = [a for a in safe if 'weather' in a['name'].lower()]
print(weather[0]['url'])  # Ready to call

For Humans — Live Demo

---

✨ Why This Exists

Every other API directory (public-apis, RapidAPI, APIList) is built for human browsing — long markdown tables, vague descriptions, no filtering. AI agents can't use those.

API Canon is different:

Feature	Other directories	API Canon
Format	Markdown tables	Structured JSON
Trust signal	None	Domain-level rating (trusted/likely/unknown/suspicious)
Fits in LLM context	5+ MB (too big)	472 KB (fits everywhere)
Live demo	None	Searchable web UI
AI agent ready	No	Prompt templates included
Health-checked	Dead links rot silently	CI pings 100 random APIs weekly

---

🚀 Quick Start

Web Browser — Live Demo

Open the demo, type "weather" or "crypto", filter by trust rating. Instant.

For Humans

# Search for APIs by keyword
curl -s https://raw.githubusercontent.com/shivtcdfinance/public-api-access-ai-compatible/main/api_access.min.json \
  | jq '.[] | select(.name | test("weather"; "i")) | {name, trust, auth}'

For AI Agents

Claude / Cursor / Copilot:

Load this file into your context: https://raw.githubusercontent.com/shivtcdfinance/api-canon/main/ai_prompt.txt
Then ask: "Find me a free HTTPS API for currency exchange rates"

ChatGPT Custom GPT:

Upload api_canon.min.json to your Knowledge.
Then prompt: "Using the API Canon, find the best API for [task]"

Python:

import json, urllib.request

url = "https://raw.githubusercontent.com/shivtcdfinance/api-canon/main/api_canon.min.json"
apis = json.loads(urllib.request.urlopen(url).read())

# Safe APIs: trusted + no auth + HTTPS
safe = [a for a in apis if a['trust'] == 'trusted' and a['auth'] == 'no' and a['https']]
weather = [a for a in safe if 'weather' in a['name'].lower()]
print(weather[0]['url'])  # Ready to call

---

📊 Data Structure

Every API entry:

{
  "name": "OpenWeatherMap",
  "desc": "Current weather forecasts, and historical data",
  "cat": "weather",
  "domain": "openweathermap.org",
  "url": "https://openweathermap.org/api",
  "auth": "apikey",
  "https": true,
  "cors": "unknown",
  "trust": "trusted",
  "source": "marcelscruz"
}

Field	Values	Description
name	string	API name
desc	string	Short description
cat	string	Category (finance, weather, geocoding, etc.)
domain	string	Provider domain
url	string	API documentation URL
auth	no / apikey / oauth	Authentication required
https	true / false	HTTPS support
cors	yes / no / unknown	Cross-origin support
trust	trusted / likely_safe / marketplace / unknown / suspicious	Trust rating
source	marcelscruz / public-api-lists	Data provenance

---

🔒 Trust Rating System

Every API domain is classified automatically:

Rating	Count	Criteria
trusted	252	Google, Microsoft, Stripe, .gov/.edu, known providers
likely_safe	61	Hosted on Vercel/Render/GitHub Pages — reputable platforms
marketplace	75	RapidAPI, APYHub — vet individual APIs
unknown	1,453	Valid SSL, active DNS — probably fine, check before uploading data
suspicious	10	.xyz, .cf, .gq TLDs — excluded from trust tiers

Methodology

1. Domain whitelist — ~200 known trusted organizations (Google, Microsoft, Stripe, GitHub, etc.)
2. TLD blocklist — Free/cheap TLDs commonly used for disposable services
3. Gov/edu detection — .gov, .edu, .mil domains auto-trusted
4. Platform detection — vercel.app, netlify.app, github.io = likely_safe
5. SSL validation — Spot-checked: 98% of unknown domains have valid SSL
6. Health check — Weekly CI pings 100 random APIs, flags dead domains

---

📂 Files

File	Size	Use
api_access.min.json	470 KB	Primary — load this into your app/AI agent
api_access.full.json	770 KB	Full version with category/auth indexes
docs/index.html	—	GitHub Pages searchable demo
examples/	—	Python, curl, Node.js, AI prompt examples
scripts/	—	Rebuild from upstream sources
tests/	—	Schema + trust + health validation

🔄 Auto-Sync Pipeline

This repo is kept in sync from a central curated database:

Central DB (local, private)
  └─ health_check.py   → Weekly: pings APIs, prunes dead ones
  └─ sync_to_github.py → Sanitizes & pushes clean data here
      ✓ Strips personal API keys / private entries
      ✓ Only marcelscruz + public-api-lists sources
      ✓ Zero personal data ever touches this repo

---

⚡ Integration

Data Sources

Built from two community-maintained API directories:

• marcelscruz/public-apis — 1,605 APIs in structured JSON
• public-api-lists/public-api-lists — 730+ community-vetted APIs

Both MIT-licensed. Clean, no malware, no telemetry. Full audit →

Rebuilding

cd scripts
python3 build_db.py  # Pulls latest from upstream, rebuilds api_canon.json

---

📋 Contributing

Found a dead API? Wrong trust rating? Missing category?

1. Open an issue with the API name and what's wrong
2. Or: edit api_canon.min.json directly and open a PR
3. PRs are auto-validated by CI for JSON schema + trust consistency

---

⭐ Credits

• Shivansh Rao (@shivtcdfinance) — curation, trust system, build pipeline
• marcelscruz/public-apis — upstream data source
• public-api-lists/public-api-lists — upstream data source
• Jarvis (Hermes Agent) — built, audited, and deployed end-to-end

---

License: MIT — use freely in any project, commercial or personal.