B2B prospect discovery with registry IDs

Prosp finds, validates and enriches B2B company prospects from the live web. Describe the companies you want to reach; Prosp searches the web, crawls each candidate's website, checks it against your criteria and returns the companies that pass, with the funnel that explains the count. Use it from a REST API, from ChatGPT or Claude over MCP, or from a Telegram bot.

Start in Telegram Read the API reference

How a discovery run works

  1. Describe the companies you want. A sentence of intent plus optional industries, countries or regions, company sizes, keywords and exclusions. Save it as a reusable ICP if you will run it again.
  2. Prosp searches and checks every candidate. Candidates come from several web search and language-model sources. Each one's domain is resolved and its website crawled, then it is scored against your criteria and, where an official company register is available, checked against it.
  3. You get the companies that passed, and the funnel behind them. Each delivered company carries its firmographics, location and the business contacts Prosp found for it. The result also reports how many candidates were screened and evaluated and why the run stopped, so a thin result explains itself.

A run costs 100 credits and typically takes 5–25 minutes. Results arrive progressively while it runs.

Three ways in, one API underneath

REST API

OAuth2 client-credentials, an OpenAPI 3.1 contract and idempotent job submission. Start a discovery, poll it, page through prospects, acknowledge what you have consumed.

MCP for ChatGPT and Claude

Add https://api.prosp.cc/mcp-rpc as a remote MCP server and ask for prospects in a chat. Sign in with Telegram; there is no key to paste.

Telegram bot

Buy credits with Telegram Stars, run discoveries with /find, read results, and get your API credentials, in 14 languages.

All three surfaces return the same customer-safe field set, enforced at the API gateway: internal scores and raw crawled content never leave the platform.

Use cases from real runs

Each article is rendered from one real run: the exact request, the companies that came back, how many candidates were rejected and for which reasons, and what went wrong.