Ask an AI assistant for a prospect's email address and it will confidently give you the results - but our research on LLM email accuracy found that 76% of the email addresses that AI models with no live data connection gave us were wrong.
Hunter MCP fixes that by giving your AI a verified B2B data source.
This guide is for what comes after you connect: what each action costs, the limits your plan sets, what AI can and can't do without asking you, and how to write prompts that get good results.
When you're ready to put it to work, start with Hunter MCP for lead generation, or the other use case guides near the end, which walk through real workflows for each team.
What is Hunter's MCP?
Hunter's MCP is the official Model Context Protocol server that turns your Hunter account into a tool for your AI. Instead of answering from memory, your AI searches Hunter's data, saves leads, and runs email sequences for you, within your plan and credits.
MCP itself is an open standard that Anthropic released in November 2024 for connecting AI assistants to the tools where data lives.
Hunter's server sits in front of the Hunter API and exposes it as around 100 individual tools. You describe the job in plain English, and the assistant picks which tools to run and in what order.
MCP server, plugin, app, or API: which one to use
Hunter offers several ways into the same data.
The Claude plugin and the ChatGPT app both run on the official MCP server, so they use the same account, plan, and credits.
The difference is where you work and how much setup you want.
Option | Best for | Where it works | How you connect |
|---|---|---|---|
Anyone using an MCP-compatible assistant | Claude.ai, Claude Desktop, Gemini, Perplexity, Cursor, Codex, and other MCP clients | Add https://mcp.hunter.io/mcp as a custom connector | |
Claude Code and Claude Cowork users | Claude Code, Claude Cowork | Install the plugin, which bundles Hunter's tools with skills | |
ChatGPT users | ChatGPT on web, desktop, and mobile | Add the app from ChatGPT | |
Developers writing their own code | Any system that can make HTTP requests | API key and endpoints |
If you work in a chat assistant, start with the plugin or app for your assistant, or the MCP server if your assistant has no dedicated option.
Pick the API only when you're writing code and want full control over every request.
Our guide to MCP vs API for lead generation goes deeper on that choice.
Get connected first
If you haven't connected yet, do that before anything else. It takes a couple of minutes through the Hunter MCP page linked above, or the guided pages for Claude, ChatGPT, and Gemini.
A free Hunter account is enough to start and gives you 50 credits a month.
For a full prospecting walkthrough once you're connected, see how to use Claude for prospecting, prospect inside ChatGPT, or run research-led prospecting with Perplexity.
Check if connection works
Run this prompt right after connecting to confirm the connection without spending a credit.
Check your connection
Hunter, how many credits do I have left this month, and which plan am I on?
If the answer shows your plan and credit balance, you're connected.
If the assistant answers in general terms instead, it didn't call Hunter.
Use the troubleshooting section below to address any challenges you have.
What Hunter's MCP can do
The MCP covers most of what you can do in the Hunter web platform.
These are the tool groups your assistant can use, by their names in Hunter:
Tool group | What your assistant can do |
|---|---|
Discover | Find companies from a plain-English description, count available contacts, and save searches |
Domain Search | List the people Hunter knows at a company, filtered by department, seniority, and email type |
Email Finder | Find one person's email from their name and company |
Email Verifier | Check whether an address will deliver, with a status and confidence score |
Lead Enrichment and Company Enrichment | Fill in a person's role and profiles, or a company's industry, size, technologies, and funding |
Leads and company lists | Save contacts without duplicates, organize lists and folders, add tags and custom attributes |
Sequences | Build sequences and follow-ups, manage recipients and templates, pick the sending inbox, start, pause, and read stats |
Integrations and account | Push leads to a connected CRM, manage webhooks, check credits, and manage API keys |
The MCP pulls ahead of the app when the assistant chains these together.
One prompt can take you from a company search to a drafted sequence, with verified contacts saved as leads along the way. In the Hunter app, that's five screens and a few CSV exports.
Use cases by job
Each team uses the MCP differently. These guides walk through real workflows with copy-paste prompts, what comes back, and what each step costs.
Team | What they use the MCP for | Guide |
|---|---|---|
Lead generation | Building target account lists, finding decision makers, verifying contacts, and turning them into sequences | |
Sales | Account research, call prep, multi-threading deals, and keeping the CRM current | |
SEO and link building | Finding the right editor at each publication and re-engaging quiet contacts | Coming soon |
Recruiting | Sourcing candidates by department and seniority, and reaching hiring managers | Coming soon |
Partnerships | Finding partnership and affiliate contacts at companies that share your audience | Coming soon |
For ready-made prompts across every role, browse the Prompt library.
If you're new to the approach itself, the AI lead generation guide helps get you up to speed on how it works.
MCP pricing
The MCP is free. Actions use the same credits they'd use in the Hunter app, and many use none.
Action | Cost |
|---|---|
Discover searches, contact counts, and saved searches | Free |
Domain Search through the MCP | 1 credit per 1-10 emails returned |
Email Finder | 1 credit per email found |
Email Verifier | 0.5 credits per verification |
Lead, company, or combined enrichment | 0.2 credits per enriched email or domain |
Leads, lists, Sequences, CRM pushes, and account checks | Free |
A few rules:
- You pay nothing when Hunter can't find or verify an address.
- Repeating the same search in the same billing period doesn't cost you twice, and
- Sending emails through Sequences uses no credits at all.
Data Platform plans work differently, with separate Search and Verification credit pools.
Plan limits that affect the MCP
Your plan's limits apply in the chat exactly as they do in the app.
The Free plan is enough to try every workflow, with a few caps:
- Credits: 50 a month, shared across all tools.
- Discover: up to 10 AI-powered searches a month, up to 100 saved companies, and no premium filters like Technologies, Year founded, or companies similar to.
- Sequences: one connected email account and up to 500 recipients per sequence. Link tracking and Sequences reporting aren't included.
Plain-English searches through the MCP count as AI-powered Discover searches, so on the Free plan, 10 a month is the maximum.
Permissions and safety
Hunter's MCP is the official server, and it only reaches your Hunter data within the permissions you grant. Your assistant never stores your Hunter credentials. If you're vetting any data source for an agent, start with our checklist on data provenance for AI agents.
Actions that send email or change outside systems ask for your explicit confirmation first, including:
- Starting a sequence, because it sends real email to real people.
- Pushing leads to your CRM, because it writes to another system.
- Creating or deleting API keys.
Reading data, searching, and drafting happen without a prompt, and nothing is saved to your Hunter Leads unless you ask.
For bulk changes to your leads, add "show me the count before you do anything" to the prompt.
How you connect also sets what's possible:
- OAuth connections can't manage API keys
- API key connections can, which is useful when you're setting up an agent.
To revoke access, disconnect Hunter in your assistant's settings, or delete the API key it uses from your Hunter account.
How to write prompts that work
Good prompts have the same four parts:
- The job: one specific piece of work, stated first. "Find senior marketing people at 10 UK fintechs" is a job. "Help me with outreach" is a topic, and the assistant will fill the gap with guesses.
- The inputs: the list name, domain, ICP, or numbers only you know. Anything you leave out, the assistant decides for you.
- The output shape: a table with named columns, a ranked list, or a draft held for review. Naming the shape stops it from returning a wall of prose.
- The stop condition: where it hands control back to you, like "show me the credit estimate before you search" or "don't start the sequence."
Here's a prompt that has all four parts:
A prompt with all four parts
Hunter, find 10 UK fintech companies with 51-200 employees and up to 3 senior marketing people at each. Return a table with company, name, title, email, and confidence score. Show me the credit estimate before you search for people.
Start prompts with "Hunter, help me..." so the assistant reaches for Hunter instead of its training data. Name the department and seniority, since those map straight to filters. Ask for confidence scores every time, and count contacts before you reveal them.
In case you’re looking for prompt inspiration, we built a library of AI prompts for sales and lead generation that helps you with most common jobs-to-be-done.
Troubleshooting
Most problems come down to the assistant not calling Hunter, or a plan limit:
Problem | Fix |
|---|---|
The assistant answers from general knowledge | Start the prompt with "Hunter," and check the Hunter connector is turned on for that chat |
Hunter doesn't show up as a connector | On Claude's Free plan, only one custom connector fits, so remove another first. On Team and Enterprise, ask an Owner to add it |
An authentication error | Disconnect Hunter and connect again. With an API key connection, check the key still exists in your Hunter account |
A filter or feature isn't available | Some Discover filters and Sequences features need a paid plan. See the plan limits above |
You ran out of credits | Ask "How many credits do I have left?" and check your plan before large runs |
Build your own agent with Hunter's MCP
Developers can connect any MCP-compatible agent framework or coding tool to the same server, from an internal research agent to an AI SDR.
Use an API key connection: add https://mcp.hunter.io/mcp as a remote server and pass the key in an X-API-Key header.
It helps to give each agent its own key.
With an API key connection, your assistant can create a named key for a new agent and delete old ones from the chat.
The full key value is shown once, so store it somewhere safe. You can't delete your last remaining key.
In Claude Code, the Hunter plugin is the faster route, since it bundles skills that tell the agent when and how to use each Hunter tool. If you need fine-grained control over each request, build on the Hunter API directly.
What Hunter's MCP doesn't do yet
A few things still live in the Hunter app, and it's better to know before you try them in a chat:
- Signals: Hunter's monitoring of funding rounds, job openings, and company changes isn't exposed as MCP tools.
- Bulk tasks: for lists of thousands, the app's bulk finder and verifier are the faster route. The MCP works through rows one by one.
- Connecting a new inbox: the first connection of a sending account happens in the Hunter dashboard. After that, the assistant can use it.
- Creating webhooks: the assistant can list and repoint existing webhooks, not create new ones.
- Exact saved-search reruns: reruns are rebuilt from the stored filters as a fresh query, so review the query before running it.
FAQ
Will the assistant make up email addresses?
Not when it's using Hunter. In our test, 45% of the addresses GPT-5.5 gave on its own were invalid. With Hunter connected, that dropped to 3%, because the assistant reports when it can't verify a contact instead of guessing.
Can Hunter's MCP send emails for me?
Yes, through Hunter Sequences. It always asks for your explicit confirmation before a sequence starts.
How many credits does a typical workflow use?
Most single actions cost 1 credit or less. A workflow that finds 10 companies and up to 3 contacts at each usually costs 10-20 credits for the searches, plus 0.5 per address you verify. Counting contacts is free, so ask for an estimate before the assistant searches.
Can I use Hunter's MCP on the Free plan?
Yes. You get 50 credits a month, up to 10 AI-powered Discover searches, 100 saved companies, one connected email account, and 500 recipients per sequence. Premium Discover filters, link tracking, and Sequences reporting need a paid plan.
Should I connect with OAuth or an API key?
Use OAuth for chat assistants like Claude and ChatGPT. It's the quickest setup and there's no key to manage. Use an API key for agents and coding tools, and when you want to create or delete API keys from the chat.
How do I disconnect Hunter from my assistant?
Remove the Hunter connector in your assistant's settings. If you connected with an API key, delete that key in your Hunter account so it stops working everywhere.
Start with one workflow
Connect Hunter, run the free credit check, then pick the use case guide for your team and run its first workflow. Most cost a handful of credits and show you the whole loop in one sitting.
Send cold emails with Hunter