Use Prosp in n8n
You describe the companies you want in one sentence. The workflow starts one Prosp search, waits until it ends, and adds one row per company to a Google Sheet. It calls the same REST API as every other Prosp client, with the plain HTTP Request node.
Cost. Every run starts one paid search: 100 credits. Reading the result costs nothing.
What the workflow does
- Describe. You write the request. Countries and company sizes are optional.
- Start the search. One retry-safe call. This is the paid step.
- Wait until it ends. It checks every 20 to 120 seconds, at most 30 times (about 55 minutes).
- Read the companies. 200 per page.
- Write the rows. One row per company in your sheet.
Two more steps are included and switched off: a weekly schedule trigger, and an example call that sends each row to your own CRM.
What you need
- An n8n instance. See the version note below.
- A Prosp account with credits. Open the Telegram bot
@prosp_mcp_credits_bot and buy a credit pack with
/buy. The first purchase creates the account. - API credentials. Send
/credentialsto the bot. It returns a Client ID and a Client Secret. - A Google account, for the Google Sheets step.
Your company does not use Telegram? Write to us and we set up API access for you.
What it costs
- One search costs 100 credits. The workflow starts one search per run.
- Credits come in packs, paid with Telegram Stars. Needs a Telegram account · introductory offer: about €2 for 200 searches (100 Stars). All packs are on the pricing page.
- The schedule trigger is off for a reason: every tick starts a new paid search.
Set it up
- Import the file. In n8n, create a workflow, open the menu and choose Import from File.
- Create one credential of type OAuth2 API:
- Grant type:
Client Credentials - Access token URL:
https://prosp-mcp-641053169627.auth.eu-central-1.amazoncognito.com/oauth2/token - Scope:
mcp/execute - Authentication:
Header - Your Client ID and Client Secret
- Grant type:
- Select the credential on the three Prosp HTTP steps: Start the search, Check the search and Read the companies.
- Pick your Google Sheet in the last step. Row 1 must hold the column names listed below.
- Edit the first step, Describe the companies you want.
- Run by hand.
The first step has five fields:
| Field | What to write |
|---|---|
request |
The companies you want, in one sentence. |
countries |
Optional. Comma-separated, for example Czechia. |
company_sizes |
Optional. Comma-separated, for example small,medium. |
max_companies |
How many companies you want. |
run_even_if_few_new |
false by default. See "Held before the start" below. |
The columns
The workflow writes these columns. They are fields the API returns; contacts are business contact details found on public pages.
| Column | Meaning |
|---|---|
name |
The company's name. |
website, domain |
The company's website and its domain. |
country, city |
Where the company is based. |
size, employees |
Size band and number of employees, where known. |
registration_number |
The company's number in its government register, where found. |
company_emails, company_phones |
The company's own published e-mails and phones. |
contact_name, contact_position |
One named contact: the first one with an e-mail, else the first one. |
contact_email, contact_phone, contact_linkedin |
That contact's details. |
contact_email_inferred |
true when the e-mail was inferred from the company's address pattern, not found on a page. |
emails_count, phones_count, contacts_count |
How many of each the company has. |
search_id |
The search the row came from. |
What happens when
- Out of credits. The workflow stops with "Prosp refused the start". Nothing is started.
- Held before the start. Prosp can hold a search and say why, for example when it expects few
new companies. The workflow stops and nothing is charged. To run it anyway, set
run_even_if_few_newtotrue. - The same search is already running. The workflow follows the running search. No second charge.
- Too many requests. It waits, then retries with the same key, up to 4 times. A rejected request never costs credits.
- Fewer companies than asked. The rows found are still written. The search is credited back as the pricing page describes.
Version note
We ran this workflow in n8n 2.48.7: the import, the start, the waiting, the paging and every case above, against a stand-in for the Prosp API. The Google Sheets step and waits of 80 seconds or longer have not had a full run with a live account yet. If a step fails on your version, tell us.
Use it from an assistant instead
No workflow needed: ChatGPT or Claude can run the same search from a chat. See the MCP server page.