Hire from inside Claude or ChatGPT with Tapflow

You already describe roles to an AI assistant. Now the assistant can do something with the description.

Tapflow runs an MCP server, the open standard assistants use to call outside tools. Connect it once and Claude, ChatGPT, Cursor or VS Code can create a role, show you the price, start a search when you say yes, and work the shortlist with you — without leaving the chat.

This walkthrough shows the setup and a first session, step by step.

Pick your assistant

Everything starts from one address: www.tapflowai.com/mcp. Open it in a browser and you get the setup page; add it to an assistant and you get the server.

• Claude Desktop, Claude Code, Cursor, VS Code and Windsurf: paste the address and your API key from Settings → API. Cursor and VS Code have one-click install buttons on the setup page.
• Claude.ai and ChatGPT: add the address as a custom connector. Then sign in with your Tapflow account and approve; each teammate connects with their own account.
• Anything else that speaks MCP over Streamable HTTP: the generic configuration on the setup page works the same way.

Try it without an account

Add the address with no key and ask: what roles do I have? The assistant answers from sample roles and a sample shortlist. Nothing is saved and nothing is charged. A sandbox key (taptest) goes one step further: you can create roles, approve and move people, all against fixed sample data, with no credits used.

A first session

With a live key in place, paste a job description and say what you want:

Here's the job description for our senior data engineer. Set up the role in Tapflow and tell me what it would cost to find 10 people.

The assistant creates the role. Tapflow reads the description into an ideal profile — title, seniority, must-have skills, location — and prepares a few questions where the description was unclear. The assistant shows you those questions with their default answers, and the price for the search.

You correct what is wrong in plain words (Pulsar is a must-have, EU remote is fine) and say go. Only then does the assistant start the search. Starting a search is the one paid step, and the tool refuses to run without an explicit confirmation after the price has been shown.

Following the search

Ask how it is going whenever you like:

How is the search for the data engineer going?

You get the people read so far, how many are on the shortlist, and anything that needs your answer. The search runs on Tapflow's side; closing the chat does not stop it.

Working the shortlist

Show me the top five on the shortlist and why each one fits.

Each person comes with the reasons they fit the role. Ask for a Deep-dive on anyone you are unsure about and the assistant reads back the evidence for every claim. When you have decided, ask it to move people forward or pass on them, and to find personal emails for the ones you want to contact. Emails are only looked up when you ask.

What it costs

A search your assistant starts is an API start, priced at the API share of the app's price for the same search; the setup page shows today's figures. Reading roles and the shortlist, Deep-dives and email lookups are free.

Same account, same rules

Every tool call is a call to the public API: same key scopes, same rate limits, and every call appears in your request log in Settings → API, marked with the assistant's name. Whatever your assistant does shows up in the app, and whatever you do in the app your assistant can read.

Setup for every assistant: www.tapflowai.com/mcp.