mcpField
S
Stophy

Stophy

Remote
github.com/stophydotdevBrowser Automation0Updated 3w ago
Repository

About Stophy

Stophy is a web data API for AI agents, exposed as an MCP server. Its three tools help agents find endpoints, inspect their inputs and outputs, and call them for web, video, place, job, and housing data.

Key features

  1. 01Find data endpoints
  2. 02Inspect endpoint inputs and outputs
  3. 03Call endpoints and select result fields
  4. 04Authenticate with an API key or browser sign-in
  5. 05Run a local forwarding package

Use cases

  1. 01Search the web and retrieve results
  2. 02Look up videos, places, or users
  3. 03Find job and housing listings

MCP support

As an MCP server, it exposes three tools for endpoint discovery, endpoint descriptions, and API calls. Clients can connect to the hosted service or run a local package that forwards requests.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://api.stophy.dev/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://api.stophy.dev/mcpAuth required
Streamable HTTPhttps://mcp.stophy.dev/mcp
Packages
npm@stophy/mcp2.0.9

Tools

3
ParameterTypeRequiredDescription
querystringNoWhat you want in plain words, such as "restaurant reviews in Austin".
categorystringNoSource name, such as reddit or youtube.

Details

From the project README

Stophy MCP

smithery badge

Web scraping API for AI agents, as an MCP server. One API to search the web, read what people say, and look up products, places, jobs and homes. Pay only for answers. Your agent reaches every Stophy endpoint through three MCP tools.

Add to Cursor

Connect to the hosted server

Most apps can connect to the hosted server directly. There is nothing to install.

URLHow you connectWhat your agent can use
https://api.stophy.dev/mcpNo keyThe endpoints marked keyless: true in the endpoint list, within a free allowance
https://api.stophy.dev/mcpAuthorization: Bearer <key> headerEvery endpoint
https://mcp.stophy.dev/mcpSign in with your browserEvery endpoint

Get an API key at stophy.dev/signup.

Claude

In Claude on the web, Desktop or mobile, open Customize > Connectors, choose Add custom connector, and paste https://mcp.stophy.dev/mcp. Sign in when Claude asks.

Claude Code

To sign in with your browser, add the server, then run /mcp and choose Authenticate:

claude mcp add --transport http stophy https://mcp.stophy.dev/mcp

To use an API key:

claude mcp add --transport http stophy https://api.stophy.dev/mcp --header "Authorization: Bearer <key>"

To start without a key:

claude mcp add --transport http stophy https://api.stophy.dev/mcp
Codex

To use the key in your STOPHY_API_KEY environment variable:

codex mcp add stophy --url https://api.stophy.dev/mcp --bearer-token-env-var STOPHY_API_KEY

To sign in with your browser instead, add the sign-in URL, then log in:

codex mcp add stophy --url https://mcp.stophy.dev/mcp
codex mcp login stophy
Cursor

Use the Add to Cursor button above to sign in with your browser, or add this to .cursor/mcp.json to use a key:

{
  "mcpServers": {
    "stophy": {
      "url": "https://api.stophy.dev/mcp",
      "headers": { "Authorization": "Bearer <key>" }
    }
  }
}

To start without a key, leave out headers.

Run it as a local server

Some apps can only start a local MCP server. For those, use this package. It runs on your computer and passes every request to the hosted server, so it always has the same tools.

npx -y @stophy/mcp

With STOPHY_API_KEY set, your agent can use every endpoint. Without it, your agent can use the endpoints that need no key, within a free allowance.

Claude Desktop

Add this to claude_desktop_config.json, then restart Claude Desktop:

{
  "mcpServers": {
    "stophy": {
      "command": "npx",
      "args": ["-y", "@stophy/mcp"],
      "env": { "STOPHY_API_KEY": "<key>" }
    }
  }
}

To start without a key, leave out env.

The package needs Node.js 18 or later.

VariableWhat it does
STOPHY_API_KEYYour Stophy API key. Optional.
STOPHY_MCP_URLThe server to connect to. The default is https://api.stophy.dev/mcp.

Tools

ToolWhat it does
stophy_search_endpointsFinds the endpoint for a task, with its cost
stophy_describe_endpointShows the input an endpoint takes and what it returns
stophy_callRuns an endpoint with id, input and optional fields, and returns JSON

Your agent usually searches, then describes, then calls. Endpoint ids are camelCase, such as googleSearch or youtubeSearch. Lists come back as results, and fields keeps only the keys you name on each row. Each result says how many credits it used.

To point at one thing, send its link or its id, never both: videoUrl or videoId, placeUrl or placeId, userUrl or username. The endpoint description lists the pair.

One call returns one page from the site at the price shown for that endpoint. Most calls cost 1 credit, some cost 2, and long lists cost 1 credit per 10 results. Some list endpoints take page and return page: ask for the next number, and stop when results comes back empty. Others return a cursor when there is more: pass it back unchanged as cursor to get the next page.

Network and credentials

  • The hosted server lives at https://mcp.stophy.dev/mcp (sign in) and https://api.stophy.dev/mcp (no key or an API key). Your client sends requests only there.
  • Credentials: browser sign-in handled by your MCP client, or an optional STOPHY_API_KEY that you set yourself. The local package also reads an optional STOPHY_MCP_URL.
  • The plugin manifests (.mcp.json, .grok-plugin/plugin.json, mcp.json, plugin.json) only point at the hosted server. They run no code on your machine.
  • No telemetry.

More

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.