Company data and web search for AI agents

Give your agent company lookups and web search as tools, over MCP or plain HTTP. JSON in and out, a fixed price per call, and no subscription.

The tools your agent gets#

Connect over MCP#

Connect your agent to our MCP server with your API key. Apps that only take a server address, like Claude's connectors, can carry the key in the address; keep that address private, because it holds your key. Every call your agent makes shows up in your ledger like any other.

In Claude (web, desktop and mobile): Settings > Connectors > Add custom connector, with this URL. Create a key just for the connector:

https://api.happyfetch.net/mcp?apiKey=YOUR_API_KEY

claude mcp add --transport http happyfetch https://api.happyfetch.net/mcp \
  --header "Authorization: Bearer $HAPPYFETCH_API_KEY"

{
  "mcpServers": {
    "happyfetch": {
      "url": "https://api.happyfetch.net/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Any other client that speaks MCP over Streamable HTTP connects to https://api.happyfetch.net/mcp with the same Authorization header, or with ?apiKey= in the URL when it takes only a URL. The MCP docs cover every client and list every tool.

Or define the tools yourself#

Any framework with tool calling works. This is the domain lookup as a tool definition, in the format of Anthropic's Messages API. OpenAI's function calling takes the same schema.

{
  "name": "domain_lookup",
  "description": "Look up the company behind a website. Returns its name, industry, size, headquarters and more as JSON.",
  "input_schema": {
    "type": "object",
    "properties": {
      "domain": { "type": "string", "description": "A website domain or any URL on it, such as stripe.com" }
    },
    "required": ["domain"]
  }
}

When the model calls the tool, send the input to the API and hand the JSON back to the model, errors included:

import os

import requests

API = "https://api.happyfetch.net"


def domain_lookup(domain: str) -> dict:
    res = requests.post(
        f"{API}/v1/domain",
        headers={"Authorization": f"Bearer {os.environ['HAPPYFETCH_API_KEY']}"},
        json={"domain": domain},
        timeout=120,
    )
    return res.json()  # errors are JSON too, with a code and whether retrying helps

Why it suits agents#

• No missing fields. Every field is always there, and empty when a value isn't known, so the model never trips over one that's missing.
• Errors a model can act on. Each error has a stable code and a plain message, and says whether trying again helps.
• One key, one bill. Company data and web search on the same key and the same ledger.
• A spending cap built in. Credits are prepaid, so an agent stuck in a loop can never spend more than your balance. When it runs low we email you, and you choose the level in the console.
• Repeat questions cost less. Agents ask about the same companies again and again. Lookups from our cache cost half.

What it costs#

Company lookups by LinkedIn company URL#

Every lookup fetched fresh, at 2 credits each. Lookups our cache answers cost half. A confirmed not-found costs 1 credit, and nothing from our cache. Starter: $10 for 20,000 credits. Business: $500 for 2,000,000 credits. Credits never expire.

Company lookups by website domain#

Every lookup fetched fresh, at 4 credits each. Lookups our cache answers cost half. A confirmed not-found costs the same, and nothing from our cache. Starter: $10 for 20,000 credits. Business: $500 for 2,000,000 credits. Credits never expire.

Web searches#

1 credit each, fresh or from our cache. Starter: $10 for 20,000 credits. Business: $500 for 2,000,000 credits. Credits never expire.

New accounts get 200 credits, enough to wire up a tool and watch your agent use it.