Find and Personalize Leads Without Leaving Claude: DeepReachAI's MCP Server

The DeepReachAI MCP Server lets you connect DeepReachAI to Claude as a Custom Connector so research and outreach tasks can be started from a Claude workspace. The goal is to keep the conversation and the sales workflow closer together without making research context disappear into disconnected tabs.

MCP stands for Model Context Protocol. In this setup, Claude can call the DeepReachAI endpoint using an authenticated connector. Access requires a DeepReachAI account with the appropriate plan and an API key; it is not an anonymous public endpoint.

What the connector is designed to support

DeepReachAI’s MCP setup is intended to let you prompt Claude to work with DeepReachAI capabilities such as:

• Drafting personalized outreach from researched prospect context.
• Starting or managing outreach sequences.
• Running bulk research jobs for a focused list.

Use the connector to prepare and review work, not to bypass the same quality checks you would use in the DeepReachAI app. A useful AI SDR workflow still needs a clear ICP, evidence, human review, and a stop rule for unwanted outreach.

How to connect DeepReachAI to Claude

The setup flow is straightforward, but the API key should be treated like a password:

1. Sign in to DeepReachAI and confirm that your account has access to the MCP setup.
2. Open the MCP Setup & API Keys page.
3. Create a named API key and copy it somewhere safe. The key is shown for you to save when it is created.
4. In Claude, open Settings, then Connectors, then Add custom connector.
5. Use DeepReachAI as the connector name and the endpoint https://deepreachai.tech/api/mcp.
6. Leave OAuth Client ID and OAuth Client Secret blank when following DeepReachAI’s current setup instructions.
7. Complete authorization with the DeepReachAI API key.

Never paste an API key into a public prompt, commit it to a repository, or include it in a blog post or screenshot. Revoke a key if you believe it has been exposed.

A useful prompt pattern

A connector is only as useful as the instruction around it. Start with the market, evidence standard, output, and review requirement:

Research companies that fit [ICP] and are showing [signal].
Return the evidence and source for each account.
Draft one email per reviewed contact, but do not invent pain points.
Show the research and draft before any sending or sequencing action.

This keeps the request aligned with the market-to-first-email workflow. Ask Claude to expose the reasoning and source context so a human can correct it.

Why use an MCP server for sales outreach?

Sales research often starts as a question: which companies fit, what changed, and who should I understand next? A conversational interface can make that exploration faster. The MCP Server provides a path for Claude to use DeepReachAI’s research and outreach workflows while you keep the question in context.

It is especially useful when you want to move from an account-level question to a prepared set of drafts without manually copying every research note between tools. It does not eliminate the need to check source quality, targeting, or message relevance.

Security and access boundaries

• API keys should be created for a specific workspace use and revoked when no longer needed.
• Do not share keys across people or put them in source control.
• Review account and subscription access before expecting the connector to work.
• Keep sending and sequencing decisions behind a human approval step.
• Use the connector only with data and outreach practices your team is authorized to use.

How this fits the AI SDR workflow

An MCP server is an interface layer, not a sales strategy. The strategy still comes from your market definition, qualification rules, and message quality bar. Cold email personalization explains how to turn evidence into a message; the MCP Server gives you another way to ask DeepReachAI to help prepare that work.

Get started carefully

Read the MCP setup page, create an API key only when you are ready to connect Claude, and test the workflow with a small set of accounts. Keep the research visible, review the drafts, and expand only after the output is accurate enough to trust.