> ## Documentation Index
> Fetch the complete documentation index at: https://docs.airscale.io/llms.txt
> Use this file to discover all available pages before exploring further.
# How to use the Airscale MCP (+Claude demo)
> Prospect through conversation with Airscale search, enrichment, and export tools.
Prospect without leaving your AI chat. Airscale connects Claude, ChatGPT, and other compatible assistants to search, enrichment, research, and export tools, so you can build a list through natural-language conversation.
## What is the Airscale MCP?
The Airscale MCP gives your assistant 37 typed tools that can:
* Check your workspace credit balance.
* Search for people and companies, including with Leads Finder filters.
* Find companies similar to reference domains with Company lookalikes.
* Discover valid Find Companies filters before searching.
* Enrich contact details and person or company profiles.
* Find a company's LinkedIn URL from its domain and list the people engaging with a LinkedIn post.
* Verify emails, check WhatsApp and US Do Not Call status, and look up Meta ads.
* Run Airsearch for structured web research.
* Create downloadable CSV or JSONL exports.
See the [MCP tool catalog](/mcp/tools) for every tool and its inputs.
## How credits work
Find People costs 0.1 credit per returned person. Find Companies costs 0.1 credit per returned company. Airsearch costs 1 credit per successful call.
Company lookalikes costs 0.5 credits per unique returned company. Set an explicit small `limit`; the API reserves `limit × 0.5` credits and refunds unused credits. `airscale_company_lookalikes` returns synchronously and shares 15 starts per rolling minute and 5 active searches per workspace with the API. Do not automatically retry: a search can complete and be charged even when its response is lost, and repeating it can cause another charge. See [Company lookalikes](/api-reference/company-lookalikes).
Credit checks, people counts, and Find Companies filter discovery are free planning tools. Other enrichment follows linked API pricing in the related references for [contact data](/api-reference/email-finder), [people profiles](/api-reference/extract-people-profile), and [company profiles](/api-reference/extract-company-profile).
Before a paid export starts, review the final filters, row limit, and maximum spend, then explicitly confirm that paid action. No paid search, enrichment, or export should run before your approval.
## Run a search in natural language
Describe the cohort and keep the first result small. The example prompts below are illustrative and non-executing; they instruct your assistant to pause before every paid action.
> Find up to five cybersecurity companies in France with 51–200 employees. Before you run it, show me the proposed tool, filters, row limit, and maximum cost, then wait for my approval.
Your assistant can use a free filter-discovery tool first when a value needs validation. Approve only the bounded search you reviewed.
### Review your results
Review the company name, domain, website, country, and filter precision before expanding the search. At 0.1 credit per returned company, a five-company sample costs at most 0.5 credits.
### Enrich contacts through conversation
Begin with a bounded role-discovery request for only the companies you approved:
> Discovery step: For the approved companies, discover at most one CEO, one CFO, and one Head of Growth per approved company. Show me the proposed tool, per-step maximum cost, and cumulative workflow maximum, then wait for my approval.
Review the discovered people and approve the decision-makers you want to continue with. Then request only their professional work emails:
> Professional work email enrichment step: For only the approved decision-makers, find professional work emails. Show me the proposed tool, contact count, per-step maximum cost, and new cumulative workflow maximum, then wait for my approval.
Review the requested fields and linked pricing before approving each stage. Keep each follow-up bounded, and do not add unapproved companies or contacts to the request.
## Export your final list
People and company exports perform a fresh paid search from the final filters. They do not export the selected chat rows, may return different rows, and add a new charge in addition to earlier paid steps in the workflow.
Before starting, require the assistant to repeat the exact filters, fields, format, `max_rows`, additional maximum credits, and cumulative workflow maximum, then wait for your approval.
Exports are asynchronous: start one export, check its status, then retrieve the file after completion. Save the returned `export_id` and reuse the same `export_id` for status and file requests.
Do not start a duplicate export while the existing job is `queued` or `running`; follow the returned `poll_after_seconds` interval instead. When status is `completed` or `failed`, stop polling; retrieve the file only after completion.
## Get started
* [Connect Airscale MCP to ChatGPT](/mcp/connect-airscale-mcp-to-chatgpt)
* [Connect Airscale MCP to Claude](/mcp/connect-airscale-mcp-to-claude)
* [Read the Airscale MCP server reference](/mcp/airscale-mcp-server)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.