# looot [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/loootai/looot-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 1181  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/looot

## Description
Search, price and run 2,500+ data endpoints from 90+ providers with one key and one prepaid balance: work emails, phones, company and people search, Google results, web pages and news. Price shown before every run.

## Tools
Capabilities this server exposes over MCP:

- **catalog_overview** — categories, platforms and jobs with counts and cheapest prices
- **discover_smart** — ranks a shortlist against a plain-English use case
- **inspect** — input and output schema, price formula, estimated max cost
- **run** — holds the estimated cost, runs, settles; accepts `job:<id>` and `fallback

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "looot": {
    "command": "npx",
    "args": ["-y","looot-mcp"]
  }
}
```

## Documentation

## What looot MCP server does

The looot MCP server provides one MCP interface to about 2,500 operations from more than 90 data providers. Available work includes finding and verifying email addresses, enriching company records, searching for people and companies, retrieving Google and SERP data, scraping web pages, accessing social profiles, and collecting news data. The exact jobs available depend on the hosted catalog and workspace capabilities.

Agents can browse categories, platforms, and jobs; compare available endpoints; inspect required inputs and outputs; and see pricing information before a paid run. A single looot account uses one token and one prepaid balance across the available provider operations.

## How it works

The hosted endpoint is `https://api.looot.ai/mcp` and uses streamable HTTP. Most supported clients connect to it directly and open a browser-based sign-in flow. The server can also authenticate headless agents with an agent token supplied through `LOOOT_TOKEN`.

A typical workflow starts with `catalog_overview`, followed by catalog search or discovery. `inspect` returns the input and output schema, pricing formula, and estimated maximum cost for a selected job. `run` reserves the estimated amount, executes the job, and settles the charge. It accepts identifiers such as `job:<id>` and supports a `fallback` option. Run status, history, cancellation, and per-attempt receipts are available through the run management tools.

Browsing tools are free, while `discover_smart` makes a small paid AI call and `run` executes paid work. New workspaces start with a zero balance and do not receive trial credit; the balance tools can provide a minimum top-up amount and payment link.

## Setup and configuration

Connect a client directly to `https://api.looot.ai/mcp` using its HTTP or streamable-HTTP MCP configuration. The repository documents setup for Claude Code, Claude Desktop connectors, Cursor, VS Code, Codex, Gemini CLI, and Windsurf. Gemini CLI can also install the repository as an extension that includes a usage guide.

For clients that only support stdio, the repository provides a bridge package named `looot-mcp`, but it is not yet published to npm according to the provided documentation. Until release, the documented alternative runs `mcp-remote@0.14.3` with the hosted URL. Browser authentication is the default. Headless use requires creating an agent token with `catalog.read`, `runs.read`, and `runs.execute` permissions and keeping the token out of source control.

## Tools and capabilities

The hosted server lists these capabilities:

- `catalog_overview` for category, platform, job-count, and cheapest-price information.
- `search_catalog`, `search`, and `discover` for finding endpoints, including ranking by preference.
- `discover_smart` for ranking a shortlist against a plain-English use case.
- `inspect` for schemas, pricing formulas, and estimated maximum cost.
- `run` for executing a job with cost holding, settlement, and fallback support.
- `runs_get`, `runs_list`, `runs_cancel`, and `runs_evidence` for run operations and receipts.
- `balance` and `top_up` for account balance and Stripe payment-link workflows.
- `my_tools` and `capability_request` for workspace capabilities and requesting a missing job.

## Limitations and notes

The repository primarily configures access to a hosted service; it does not add another local execution layer beyond the stdio bridge. The hosted service has separate terms at looot.ai/terms. ChatGPT developer-mode setup is documented but not tested end to end, and the referenced public ChatGPT page is described as not live. Paid discovery and execution require an available prepaid balance.

_Full upstream README: https://allmcps.com/mcp/looot/readme_