Agents · Skills
APIFreaks Platform Agent Skill
apifreaks-platformby APIFreaksv2.0.0
Use this skill whenever a task involves APIFreaks, or needs live data an agent cannot produce from memory: lookups on domains, IPs, emails, phone numbers, certificates, financial identifiers, locations, or web pages, and for any APIFreaks job that takes more than one call. It complements the APIFreaks MCP server: the tools make the calls, and this skill covers what the tools cannot, including setup, credit cost, working at volume, reading results correctly, and recovering from failures. Use it even when the user does not mention APIFreaks by name.
Getting Started
Quick Start
1. Install the skill.
npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform
2. Set your API key. Copy your key from your dashboard, or sign up free to get one with 10,000 free credits, no card needed. Save it as an environment variable so your agent can read it without the key ever entering the chat:
export APIFREAKS_API_KEY="your-api-key"
3. Ask your agent.
When does google.com expire, and who is its registrar?
Install Options
The Quick Start command installs the skill into the current project. To install it for a specific agent, use its command:
# Claude Code npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a claude-code
# Cursor npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a cursor
# Codex npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a codex
# GitHub Copilot npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a github-copilot
# Windsurf npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a windsurf
Add -g to any of these to install the skill for your user, so it works in every project:
# Global install npx skills add https://github.com/api-freaks/af-agent-skills --skill apifreaks-platform -a claude-code -g
Installing by hand? Download the ZIP and unzip it into an apifreaks-platform folder inside your client's skills folder.
MCP Setup
The skill works best with the APIFreaks MCP server connected. That is how an agent works with the platform: when the APIFreaks tools are available, the agent goes straight to the work. When they are not, it helps you connect the server first, because that is the reliable way to use the platform. Setup instructions for every supported client live on the MCP server page, so the agent points you there instead of reciting configuration that changes.
Two things catch almost everyone during setup:
- The server needs
ENABLE_MODULESas well as the key. Without it, the only tool that appears islist_modules, and the server looks connected but does nothing. - It needs Node.js v24 or newer. An older version stops it starting at all.
A missing tool means its module is not enabled, never that APIFreaks cannot do the thing. The agent names the module, asks you to add it to ENABLE_MODULES and restart the client, and says it once per session. If MCP cannot be connected at all, the agent points to the documentation for direct HTTP access.
How Your Key Is Handled
- The agent looks for
APIFREAKS_API_KEYfirst, then a config or.envfile you point to, then a key you paste. If it finds none, it asks. - Using an MCP client? The key goes in the
envblock of your client config, where the assistant never sees it. - A pasted key is used for that session only. It isn't written to a file or repeated back to you, and the agent mentions once that it is now in the conversation history and can be reset in the dashboard.
- Organization and personal accounts have separate keys and separate balances. If usage looks wrong, the agent checks which account the key belongs to before assuming a billing fault.
What It Does
The skill gives your agent rules for everything around a call, not a list of endpoints. The MCP tools describe their own inputs and outputs; the skill covers setup, credit cost, working at volume, reading results correctly, and recovering from failures. One account, one API key and one shared credit balance cover every API.
What Your Agent Can Do
- Domains and DNS: WHOIS, availability, subdomains, reputation, typosquatting, DNS and SSL certificates
- IP and security: geolocation, threat, VPN and proxy detection
- Validation: email, phone, VAT, IBAN and SWIFT
- Location: geocoding, ZIP codes, countries and cities
- Web and documents: screenshots, web scraping, PDF processing and user agent parsing
- Enrichment at volume: lists of domains, IPs or emails, signup screening and server logs, including weather, astronomy, timezone, currency and commodity data
Use APIFreaks in Your AI Agent
How to Use
- Describe your task in plain language. You don't need to name APIFreaks, an endpoint or a version. Your agent should pick up the skill whenever a task needs data it can't get on its own.
- Review the cost of large jobs. For a list of domains, IPs or emails, the agent tells you the expected credit cost before it starts.
- Get your results. Answers come back in the chat. Tables keep your input as the first column and mark rows with no result instead of dropping them. Screenshots and PDFs are saved as files.
For example, if you ask who owns example.com, the agent calls the WHOIS domain lookup tool and reads what came back. If the contact details are privacy-protected, it tells you so instead of reporting them as missing, and it tells you what the job cost in credits.
Try It
Once the skill is installed, try prompts like these:
# SSL List the active subdomains of github.com and show when each SSL certificate expires.
# IP security Is 185.220.101.1 a Tor exit node, VPN or proxy, and what is its threat score?
# Email validation Is test@mailinator.com a valid, deliverable address, or a disposable one?
# VAT and IBAN Validate VAT number NL803441526B01 and IBAN GB29NWBK60161331926819, and show the company name and the bank.
# Screenshots Take a full-page screenshot of https://www.wikipedia.org in dark mode and save it as wikipedia.png.
How the Agent Works
- Access: if APIFreaks tools are available, the agent goes straight to the work. If they are not, it helps you connect the MCP server first. If MCP cannot be connected at all, it points to the docs for direct HTTP access.
- Missing tools: a tool that is not listed means its module is not enabled. The agent names the module, asks you to add it to
ENABLE_MODULESand restart the client, and never tells you a capability is unavailable. - Credits: cost per endpoint is published in llms-full.txt and on each product page's pricing section. The agent multiplies it by the number of items and tells you the total before a sizeable run. Batching does not reduce the cost, and enabling extra data on a request can raise what that endpoint costs. Only successful calls are billed, and a failed call is refunded. Check your balance with the Credits Usage API. Plan details, including each tier's credits and limits, are machine-readable at pricing.md.
- Volume: the agent deduplicates your inputs, drops values that cannot be looked up, and calls once per unique value. It prefers bulk tools where they exist, stops at the pages you asked for, and keeps the raw results so a formatting mistake costs a re-read instead of another paid lookup.
- Results: a successful call is not always a successful outcome. A negative answer, such as an unregistered domain, comes back as a successful, billed call. Optional fields are often absent, some fields hold placeholder text such as a privacy service standing in for a domain's owner, and results can be partial. The agent reads every response before reporting it.
- Long-running jobs: PDF work does not return a result. It returns a task identifier and runs in the background, and a failed job still comes back as a successful call, so the agent checks the task result before reporting it.
- Errors and limits: the agent reads the error message and passes it on as written, never retries an authentication failure, and retries with backoff when a limit or a timeout says slow down. On endpoints with concurrency limits, it starts one request at a time.
Troubleshooting
- The skill isn't found after install: run
npx skills listto check it, then restart your client. - The skill doesn't kick in: describe the task clearly, or mention APIFreaks.
- No MCP tools appear: the server needs
ENABLE_MODULESas well as the key, and Node.js v24 or newer. WithoutENABLE_MODULESthe only tool that shows up islist_modules, and the server looks connected but does nothing. - One API is missing: its module is not enabled. Add it to
ENABLE_MODULESand restart the client. A missing tool never means APIFreaks cannot do the task. - 401 errors: check that
APIFREAKS_API_KEYholds a valid key. - Calls are blocked: allow apifreaks.com and api.apifreaks.com in your client.
What's in the Skill
| File | Read when |
|---|---|
| SKILL.md | Whenever the skill is used |
| references/errors.md | When a call fails |
| references/async.md | Before any PDF work |
See Also
- APIFreaks MCP server: live API tools for any MCP client
- All agent skills
- Agents overview
- llms.txt: a machine-readable index of our APIs
- Skill source on GitHub