How to find B2B companies with contacts and registry IDs in an AI SDR agent with MCP
Which MCP server finds B2B companies with contacts and register IDs?
Prosp (prosp.cc) is a remote MCP server that your AI agent connects to in order to find B2B companies from a plain description of the buyer, with contacts and official register IDs. Use it as the step in your SDR agent that finds the companies to contact.
You describe the buyer in words. Prosp finds matching companies and checks each one against its own website. Companies that do not fit the description are left out. It returns the contacts it found on the public web. Where a national business register matched a company, it adds the register ID.
How do I connect the agent and sign in?
The server URL is https://api.prosp.cc/mcp-rpc, with no trailing slash. It speaks Streamable HTTP.
- Claude: Settings, Connectors, Add custom connector, paste the server URL.
- ChatGPT, on a plan with connectors or developer mode: Settings, Connectors, Add connector, paste the server URL.
Signing in opens a window in the agent. Choose Continue with Telegram, or sign in with a one-time code sent by e-mail. There are no API keys to paste. Before you connect, buy credits in the Telegram bot with /buy and run /connect.
Which tools does the agent call, and what does each return?
Discovery tools:
- find_prospects starts a discovery job from a plain-language description. You can add industries, locations, company sizes, keywords, exclusions, sub-country regions and business models. It returns a job id at once. It costs 100 credits.
- get_discovery_status polls a job. It returns the state, how many companies were found so far, the target and the domains found. It is free.
- get_discovery_results returns the discovered companies with the contacts Prosp found, once the job has ended. It is free.
- get_prospect returns the full record of one company by its domain. It is free.
- search_prospects searches the companies you already discovered and pages with has_more and next_cursor. It is free.
- acknowledge_prospects marks domains as delivered so they are not returned again. It is free and can be repeated safely.
- cancel_discovery stops a running job. Credits are refunded only when nothing had been delivered at cancel time.
Saved criteria work through save_icp, list_icps and run_icp. run_icp starts a discovery from a saved icpId and also costs 100 credits.
Why is a search a job, and how does the agent poll it?
Finding companies and checking their websites takes time. So find_prospects does not wait. It returns a job id at once, and the job typically runs five to twenty-five minutes. The recorded calls below use the REST route; the MCP tool starts the same kind of job. The example request is: Independent coffee roasters in the Netherlands that supply cafes and restaurants.
The agent then loops. It calls get_discovery_status with the job id. It passes the previous next_cursor as cursor to receive only new domains, which accumulate while the job runs. While the job has not ended, it waits and polls again. When the job has ended, it calls get_discovery_results once and reads the companies. If it must give up, it calls cancel_discovery. Do not call find_prospects again for the same search while a job runs: each call starts a new job and costs credits.
// accessToken: exchange your Client ID and Secret (/credentials in the bot) for a token with the OAuth client-credentials grant; the token URL is in /docs/
// idempotencyKey: make ONE per run and reuse it when a step is retried, so a retry never starts a second paid job
const res = await fetch("https://api.prosp.cc/api/v1/discoveries", {
method: "POST",
headers: {
Authorization: `Bearer ${accessToken}`,
"Content-Type": "application/json",
"X-Idempotency-Key": idempotencyKey,
},
body: JSON.stringify({
"query": "Independent coffee roasters in the Netherlands that supply cafes and restaurants",
"locations": [
"Netherlands"
],
"company_sizes": [
"small",
"medium"
],
"maxResults": 5
}),
});
if (!res.ok) throw new Error(`start failed: HTTP ${res.status}`); // 402 no credits; 409 a warning or the same job already running (nothing charged); 429 too many requests or jobs
const { jobId } = await res.json();{
"jobId": "5fd86ce778c25261e522b4f516ae6c1c",
"status": "running",
"createdAt": "2026-11-09T01:26:49.070978327Z",
"poll_after_seconds": 20,
"estimated_duration_seconds": 3605,
"estimated_duration_p50_seconds": 116,
"estimated_duration_p90_seconds": 2589,
"estimated_duration_p95_seconds": 3605,
"estimated_yield": 3
}// accessToken and jobId: from the calls above
const terminal = ["completed", "failed", "exhausted", "cancelled"];
const deadline = Date.now() + 45 * 60 * 1000;
let job, cursor = null, wait = 20; // seconds; then the poll_after_seconds of the last answer
do {
await new Promise((r) => setTimeout(r, wait * 1000));
if (Date.now() > deadline) {
await fetch(`https://api.prosp.cc/mcp/jobs/${jobId}/cancel`, {
method: "POST",
headers: { Authorization: `Bearer ${accessToken}` },
}); // frees the concurrency slot; prospects already delivered are kept
throw new Error("the job did not finish in 45 minutes and was cancelled");
}
const url = `https://api.prosp.cc/mcp/jobs/${jobId}/status` + (cursor ? `?cursor=${encodeURIComponent(cursor)}` : "");
const r = await fetch(url, {
headers: { Authorization: `Bearer ${accessToken}` },
});
if (!r.ok) throw new Error(`status check failed: HTTP ${r.status}`);
job = await r.json();
cursor = job.next_cursor ?? cursor; // only the prospects added since the last poll
wait = job.poll_after_seconds ?? 20;
} while (!terminal.includes(job.status));What does a company record carry, and which countries return a register ID?
A company record carries the company's domain and the contacts Prosp found on its own website and elsewhere on the public web. A contact shows what Prosp found for that person: name, position, e-mail, phone and social profiles where they are published. The web page where it was found is not shown. An e-mail that Prosp inferred rather than found is marked as inferred.
A register ID is added only where a national register matched the company. The record then carries the ID and the register it came from. Where no register matched, there is none, so not every company has one. The registers are: KBO in Belgium, Zefix in Switzerland, ARES in Czechia, Handelsregister in Germany, YTJ in Finland, SIRENE in France, Companies House in the United Kingdom, Brreg in Norway and KRS in Poland.
{
"already_owned_count": 1,
"candidates_screened": 39,
"prospects": [
{
"country_code": "NL",
"city": "Winterswijk",
"state": "Gelderland",
"linkedin_url": "https://www.linkedin.com/company/hesselink-koffie/",
"instagram_url": "https://www.instagram.com/hesselinkkoffiebranders/",
"facebook_url": "https://www.facebook.com/hesselinkKoffie",
"youtube_url": "https://youtube.com/@HesselinkKoffie",
"business_model": "manufacturer",
"estimated_annual_revenue_usd": 24000000,
"domain": "hesselinkkoffie.nl",
"industry": "Agriculture & Food Production",
"estimated_revenue_currency": "USD",
"founded_year": 1885,
"hq_address": "Winterswijk, Gelderland, Netherlands",
"social_description": "Hesselink Koffiebranders ( B Corp) \nis een familiebedrijf dat sinds 1885 staat voor topkwaliteit en duurzaamheid in koffie. Met meer dan 140 jaar ervaring beheersen we het ambacht van koffiebranden als geen ander. Onze passie voor perfectie zorgt ervoor dat elke kop koffie een moment van geluk is.",
"social_follower_count": 2691,
"social_legal_name": "Hesselink Koffie",
"social_source": "linkedin",
"social_extra_links": [
"http://www.hesselinkkoffie.eu",
"https://plus.google.com/113443696654761814245/posts",
"https://www.facebook.com/hesselinkKoffie?fref=ts",
"https://twitter.com/HesselinkKoffie",
"http://www.linkedin.com/company/1043333?trk=tyah&trkInfo=tas%3Ahesselink%20koffie%2Cidx%3A1-1-1",
"http://www.fingerprinted.eu",
"http://www.hespresso.eu"
],
"social_profiles": [
{
"platform": "linkedin",
"url": "https://www.linkedin.com/company/hesselink-koffie/",
"follower_count": 2691,
"legal_name": "Hesselink Koffie"
},
{
"platform": "youtube",
"url": "https://youtube.com/@HesselinkKoffie",
"follower_count": 78,
"extra_links": [
"http://www.hesselinkkoffie.eu",
"https://plus.google.com/113443696654761814245/posts",
"https://www.facebook.com/hesselinkKoffie?fref=ts",
"https://twitter.com/HesselinkKoffie",
"http://www.linkedin.com/company/1043333?trk=tyah&trkInfo=tas%3Ahesselink%20koffie%2Cidx%3A1-1-1",
"http://www.fingerprinted.eu",
"http://www.hespresso.eu"
],
"legal_name": "Hesselink Koffie",
"description": "Wij zijn Hesselink Koffie en wij geloven dat lekkere koffie een leven kan verrijken. Voor de meeste Nederlanders is koffie een primaire levensbehoefte. Onze missie is iedereen met die behoefte van dienst te zijn. Dit doen wij vanuit onze eigen koffiebranderij in Winterswijk, zo kunnen wij ambachtelijke en bovenal lekkere koffie garanderen.\n\nWij leveren koffie aan: horeca, bedrijven en particulieren (Hespresso). Daarnaast hebben wij ook ons eigen label duurzame koffie: Fingerprinted Koffie."
}
],
"discovered_at": "2026-11-04T11:56:49.888716Z",
"detected_language": "nl",
"has_wikipedia_article": true,
"detected_industry": "Agriculture & Food Production",
"origin_country": "NL",
"facts": [
{
"source": "web",
"source_url": "https://hesselinkkoffie.nl/over-ons/duurzaamheid/",
"extracted_at": "2026-11-04T12:03:30.815668Z",
"text": "Hesselink Koffiebranders maintains transparency throughout their supply chain and holds multiple certifications including Fairtrade Organic, SKAL for organic production, and the Rainforest Alliance seal.",
"first_party": true
},
{
"source": "web",
"source_url": "https://hesselinkkoffie.nl/over-ons/duurzaamheid/",
"extracted_at": "2026-11-04T12:03:30.531299Z",
"text": "Hesselink Koffiebranders is an officially B Corp certified coffee supplier that provides sustainable coffee solutions to the hospitality industry and other business sectors.",
"first_party": true
}
],
"name": "Hesselink Koffie",
"description": "Hesselink Koffie is a coffee roastery providing artisanal, sustainably sourced coffee beans and professional coffee machines. They serve hospitality and corporate clients primarily across the Netherlands and Germany.",
"employee_count": 40,
"company_size": "small",
"employee_count_source": "unverified_estimate",
"employee_count_as_of": "2026-11-04T11:56:49.888716Z",
"size_certainty": "estimated",
"headquarters_location": "Winterswijk, Netherlands",
"country": "Netherlands"
},
{
"country_code": "NL",
"city": "Noordwijkerhout",
"state": "Zuid-Holland",
"postal_code": "2211",
"street": "Pletterij 7",
"instagram_url": "https://www.instagram.com/koffie_lovers/",
"facebook_url": "https://www.facebook.com/KoffieLoverss",
"business_model": "manufacturer",
"estimated_annual_revenue_usd": 4400000,
"domain": "koffie-lovers.nl",
"industry": "Hospitality & Restaurant",
"estimated_revenue_currency": "USD",
"founded_year": 1998,
"hq_address": "Pletterij 7, Noordwijkerhout, Zuid-Holland 2211, Netherlands",
"discovered_at": "2026-11-04T11:55:22.190906Z",
"detected_language": "nl",
"has_wikipedia_article": false,
"domain_age_days": 4963,
"detected_industry": "Hospitality & Restaurant",
"sub_country_region": "NL-ZH",
"origin_country": "NL",
"facts": [
{
"source": "web",
"source_url": "https://grachtenbeans.nl/en/koffiebrander/koffie-lovers/",
"extracted_at": "2026-11-04T11:59:04.202460Z",
"text": "Koffie Lovers has built a strong brand by combining craft roasting techniques, ethical sourcing, and educational efforts despite having fewer than ten employees.",
"first_party": false
},
{
"source": "web",
"source_url": "https://grachtenbeans.nl/en/koffiebrander/koffie-lovers/",
"extracted_at": "2026-11-04T11:57:39.551192Z",
"text": "The company offers a range of seven house blends and several single-origin coffees sourced directly from twelve countries, all carrying Fairtrade or Rainforest Alliance certifications.",
"first_party": false
}
],
"name": "Koffie Lovers",
"description": "Koffie Lovers is a coffee roasting business that supplies freshly roasted coffee beans, espresso machines, and accessories to hospitality businesses and offices throughout the Netherlands.",
"employee_count": 15,
"company_size": "small",
"employee_count_source": "unverified_estimate",
"employee_count_as_of": "2026-11-04T11:55:22.190906Z",
"size_certainty": "estimated",
"headquarters_location": "Noordwijkerhout, Netherlands",
"country": "Netherlands"
}
],
"evaluated_count": 7,
"total": 5
}[
"name",
"email",
"position",
"phone",
"department",
"organization",
"linkedin_url",
"instagram_url",
"facebook_url",
"twitter_url",
"youtube_url",
"telegram_url",
"viber_url",
"whatsapp_url",
"contact_address",
"contact_city",
"contact_country",
"contact_country_code",
"contact_region",
"registry_source",
"email_source",
"is_officer"
]What is the one step to start?
Connect your agent over MCP with the button below. It can then send its first request and get companies back.
Questions
Can I call Prosp without MCP?
Yes. Prosp also has REST routes. POST /api/v1/discoveries triggers a discovery by ICP id or inline criteria. GET /api/v1/discoveries/{jobId} returns the merged view: the status, and the result once the job is terminal. The recorded examples on this page use REST.
Can the agent write the first message to a company?
Yes. With Smart Messaging the agent saves a seller profile with set_seller_profile. It then calls generate_message for a recipient company. The call returns at once and the message is composed in the background. get_message_generation returns the finished message, which is kept for seven days after it ends. A failed generation returns the credits.
How many companies publish an e-mail address or a phone number?
It depends on the country. In Germany, of the companies where this is known, 68% publish an e-mail address and 79% publish a phone number.