Instantly MCP Server: Connect Your AI Assistant Directly to Your Cold Email Platform

Instantly MCP Server connects AI assistants like Claude directly to your cold email platform for automated campaign analysis and management. It eliminates hours of manual CSV exports by giving your AI live access to read stats, create campaigns, and manage leads using natural language.

Instantly MCP Server: Connect Your AI Assistant Directly to Your Cold Email Platform

Updated February 24, 2026

TL;DR: The Instantly MCP Server is a community bridge that connects AI assistants like Claude Desktop to your Instantly account through the Model Context Protocol. Instead of copying campaign data into ChatGPT, you give your AI direct access to read stats, create campaigns, and manage leads using natural language. Setup requires Python 3.10+, an Instantly API key, and basic command-line comfort. Once running, you automate analysis, hygiene checks, and sequence drafts that used to take hours.

The Model Context Protocol (MCP) solves this by letting AI assistants talk directly to your tools. The Instantly MCP Server is a community-maintained package that exposes our API v2 to MCP-compatible clients like Claude Desktop. You ask questions in plain English and the AI reads live data, identifies patterns, and executes changes without you touching a spreadsheet.

What is the Model Context Protocol (MCP)?

Model Context Protocol (MCP) is an open standard introduced by Anthropic in November 2024 to standardize how large language models integrate with external data sources and tools. Think of it as a universal adapter. Before MCP, every AI tool needed custom code to access your CRM, email platform, or database. MCP creates a standard handshake so any MCP-compatible AI can connect to any MCP-compatible server.

The architecture is simple. The Server (Instantly, in this case) exposes actions like "get campaign stats" or "add lead" as structured tools. The Client (Claude Desktop or another AI assistant) calls those tools when you ask a question. The AI decides which tools to use, executes them, and returns results in natural language.

The Instantly MCP Server is built with FastMCP, a Python framework that handles protocol compliance, serialization, and error handling automatically. This means the server works without you needing to understand the spec internals. You install it, point your AI client at it, and start issuing commands.

Why this matters for agencies is real-time access. When you paste a CSV into ChatGPT, the data is already stale. With MCP, the AI reads current campaign performance, reply content, and lead status on demand. No exports, no version control headaches, no human error in copy-paste transfers.

Why the Instantly MCP Server changes agency workflows

For agency operators managing 10 to 150 inboxes across multiple client domains, time is the scarcest resource. Every hour spent in dashboards is an hour not spent closing deals or onboarding clients. The MCP server removes three friction points.

Speed. Analyzing dozens of campaigns used to mean opening each one, exporting metrics, and building comparison tables. Now you prompt the AI with a question about campaign performance across workspaces and the AI scans all campaigns in seconds, returning ranked results.

Accuracy. Manual data transfer introduces errors. You misread a number, paste into the wrong cell, or forget to update a status field. When the AI reads directly from our API, the data is current and structured. No transcription mistakes.

Scale. One operator managing five clients is manageable. Managing 15 clients with 10 campaigns each means 150 data points to track. The MCP server lets you treat your AI as a junior analyst who monitors all workspaces, flags issues like rising bounce rates, and drafts remediation steps without supervision.

"I use Instantly for Cold Outreach Emails for me and my clients. It pretty much covers everything I need like domains, emails, deliverability, AI Integration, and API." - Riccardo C. on G2

The result is you book more meetings per hour of work because the AI handles the grunt analysis while you focus on strategy and client conversations.

For a visual walkthrough of setting up cold email systems with Instantly, watch this full tutorial!

What makes MCP different from Instantly Copilot?

We offer a native AI feature called Copilot, which is built into the UI and helps you find leads, generate campaign ideas, write sequences, and summarize analytics. Copilot is purpose-built for Instantly workflows, requires no setup, and uses Instantly credits.

The Instantly MCP Server is a different tool for a different job. You bring your own AI (Claude, ChatGPT, or any MCP-compatible client), write custom prompts, and chain multiple API operations together for complex orchestration. MCP is better for agencies that want full programmatic control, need to integrate Instantly data with other tools, or prefer their AI's reasoning for multi-step workflows.

Feature

Instantly Copilot

Instantly MCP Server

Setup

None (built into UI)

Command-line installation

AI provider

Pre-configured

Bring your own (Claude, ChatGPT, etc.)

Best for

Fast content generation, templates

Cross-workspace analysis, custom automation

Cost model

Instantly credits

API access (included in Growth+)

Use Copilot for fast campaign generation inside our platform. Use the MCP server when you manage multiple client workspaces and need custom automation that Copilot does not support, or when you already use Claude Desktop for other workflows and want one AI interface.

"Deliverability tools that actually move the needle: warmup, inbox rotation, and smart sending windows help us land in Primary instead of Promotions/Spam." - Anthony V on G2

Core capabilities: What your AI assistant can control

The Instantly MCP Server exposes 38 tools across six categories, mirroring the full Instantly API v2 surface. Each tool corresponds to a specific action your AI can take.

Reading data:

  • Campaign analytics: List all campaigns and retrieve detailed status, analytics, and summary statistics
  • Lead search: Search leads by email, company, or custom variables
  • Reply monitoring: Fetch reply content from the unified inbox
  • Account health: Check email account health, warmup status, and sending limits
  • Job tracking: View background job progress for bulk operations

Writing data:

  • Campaign creation: Create new campaigns with configured sequences and schedules
  • Campaign management: Launch, pause, or archive campaigns
  • Lead operations: Add leads to campaigns and update lead status (interested, meeting booked, out of office)
  • Personalization: Update lead custom variables for personalization
  • Inbox rotation: Add or remove sending accounts from inbox rotation
  • Send controls: Set campaign send windows and daily limits
  • Blocklist management: Manage blocklist entries to avoid accidental sends
  • Warmup settings: Configure account warmup parameters

The server supports dual transport: stdio (local mode where the server runs on your machine) and HTTP (remote deployment). For agencies, we recommend local mode because it keeps API keys on your device. The server also uses lazy loading, meaning it only exposes the tool categories you need, which reduces the context window your AI consumes and speeds up responses.

One limitation to note is the server operates within Instantly's API rate limits. The server automatically tracks rate limit headers and surfaces remaining quota, so your AI can pace itself.

Step-by-step: How to connect Claude Desktop to Instantly

Prerequisites:

  • Python 3.10 or higher installed on your machine
  • An Instantly account on the Growth plan or above (API access is included)
  • Claude Desktop (download from Anthropic) or another MCP-compatible AI client
  • Your Instantly API key (instructions below)

Step 1: Get your Instantly API key

Log in to Instantly, click your account icon in the upper right, navigate to Settings → Integrations → API, and generate a new key for API V2. Copy the key and treat it like a password.

Step 2: Install the MCP server package

Open your terminal and clone the repository:

git clone https://github.com/bcharleson/Instantly-MCP.git
cd Instantly-MCP
pip install -e .

If you don't have Git installed, download it here or download the repository as a ZIP file and extract it before running the pip command from that directory.

This installs the server and its dependencies (fastmcp, httpx, pydantic, python-dotenv).

Step 3: Configure Claude Desktop

Claude Desktop reads configuration from a JSON file. The location depends on your OS:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Linux: ~/.config/Claude/claude_desktop_config.json

Open (or create) that file and add this configuration:

{
  "mcpServers": {
    "instantly": {
      "command": "python",
      "args": ["-m", "instantly_mcp.server"],
      "env": {
        "INSTANTLY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Replace "your-api-key-here" with the key from Step 1. Save the file.

Step 4: Verify the connection

Quit Claude Desktop completely (not just close the window) and restart it. Open a new conversation and ask:

"What campaigns do I have running in Instantly right now?"

If the connection works, Claude will call the MCP server, fetch your campaign list, and respond with names, statuses, and basic stats. If you see an error, check the troubleshooting section below.

See how to build automation workflows with Instantly and external tools.

Three agency workflows you can automate today

1. The autonomous A/B testing analyst

You run multiple campaigns for a SaaS client and want to identify which subject lines drive the highest open rates, then create a new variant using the winner.

Prompt:

"Analyze my last five campaigns in Workspace SaaSClient. Compare open rates by subject line. Identify the top performer and draft a new subject line variation that uses the same pattern but different wording."

What happens: The AI fetches campaign analytics, ranks subject lines, identifies the winner (e.g., "Quick question about {{company}}"), and suggests a new variation ("Thoughts on {{company}}'s Q1 goals?"). You review the suggestion, approve it, and the AI can create a new campaign draft with that subject line.

2. The hyper-personalized sequence writer

You imported 200 leads for an HR tech campaign. You want the first email step to reference something specific about each lead's company.

Prompt:

"Look at leads in Campaign HRTech_Jan. For leads at companies with 500+ employees, update the {{custom_intro}} variable to mention their company size and suggest how our tool helps at scale."

What happens: The AI searches leads, filters by company size (if that data exists or you've enriched it via Clay integration or SuperSearch), and updates the {{custom_intro}} variable for each lead. Your sequence now reads: "I noticed CompanyX has 1,200 employees. Most HR teams at that scale struggle with..." instead of a generic opener.

3. The automated lead hygiene guard

You run multiple campaigns and some leads reply with "out of office" or "no longer with the company." You want those paused automatically.

Prompt:

"Check recent replies in all active campaigns. Find leads whose replies contain 'out of office', 'no longer with', or 'unsubscribe'. Pause those leads and give me a summary of how many were paused in each campaign."

What happens: The AI scans reply content, identifies matches, updates lead status to paused, and returns a report summarizing the paused leads by campaign.

"I find Instantly incredibly beneficial for centralizing all my inboxes... The platform allows me to check my inboxes from a single location without needing to log into each one individually, which saves me a lot of time." - Eli A on G2

For more examples of automating Instantly workflows with AI and external tools, see this cold email system walkthrough.

Security and API key management

Your API key is a master key to your Instantly account. Someone with that key can read all campaigns, leads, and replies, and can create or delete data. Handle it like a password.

The key is stored locally in your Claude Desktop configuration file as an environment variable. It is not sent to Anthropic's servers. When you prompt the AI, Claude Desktop starts the local MCP server process, which uses your key to call our API.

Best practices:

  • Rotate API keys every 90 days or when a team member leaves
  • Use separate keys for production vs. testing if possible
  • Never commit keys to version control or share them in Slack
  • If you suspect a key is compromised, regenerate it immediately in Instantly settings

For agencies handling client data, confirm that using AI tooling complies with your service agreements and client privacy policies.

Troubleshooting common connection errors

"Connection Refused" or "Server not running"

Cause: The local MCP server process did not start when Claude Desktop launched.

Fix: Open your terminal and manually start the server to check for errors:

python -m instantly_mcp.server

If you see a traceback, the issue is usually a missing dependency or incorrect Python version. Confirm you have Python 3.10+ with python --version and reinstall dependencies with pip install -e . from the repository directory.

"Unauthorized" or "Invalid API key"

Cause: The API key in your configuration file is incorrect, expired, or lacks permissions.

Fix: Verify your API key in Instantly settings. Regenerate a new key if needed and update the INSTANTLY_API_KEY value in your claude_desktop_config.json file. Restart Claude Desktop after saving changes.

"Rate limit exceeded" or "429 error"

Cause: You sent too many API requests in a short window. Our API rate limits apply per workspace and are shared across all API keys in that workspace.

Fix: Ask the AI: "What is my current rate limit status?" It will report remaining quota and reset time. Wait for the reset period before issuing more commands. If you consistently hit limits, batch operations instead of individual calls.

Claude Desktop does not recognize the MCP server

Cause: Configuration file syntax error or file in the wrong location.

Fix: Validate your JSON syntax using an online JSON validator. Confirm the file path matches your OS. On macOS, the full path is ~/Library/Application Support/Claude/claude_desktop_config.json. After fixing, fully quit and restart Claude Desktop.

Start connecting AI to your campaigns

The Instantly MCP Server turns your AI assistant from a passive chatbot into an active operator. You gain hours back by automating analysis, hygiene checks, and sequence updates that used to require manual clicking and copy-pasting.

If you manage multiple client workspaces, start with the A/B testing workflow. Prove the time savings on one campaign, then expand to lead hygiene and personalization. If you hit rate limits, batch your prompts or upgrade to a higher Instantly plan with more API headroom.

We built Instantly to be developer-friendly and open. The MCP server is a community example of that philosophy in action. Ready to set this up? Grab the server from GitHub, follow the steps above, and connect your AI. You'll need an Instantly account on the Growth plan or above for API access. If you're not using Instantly yet, start with a Instantly free trial to test this workflow.

FAQs

Does this cost extra Instantly credits?
No. The MCP server uses our API, which is included in the Growth plan and above. You are not charged per API call.

Can I use this with the Instantly Growth plan?
Yes. API access is available on Growth plans as of the latest pricing updates.

Is my campaign data shared with the AI provider for model training?
For Claude commercial plans (API, Team, Enterprise), no. API data is never used for training. The API key itself stays local on your machine.

What happens if I exceed API rate limits?
The API returns a 429 error. The MCP server tracks rate limits automatically and reports remaining quota. Wait for the reset period before retrying.

Can I use this with ChatGPT or other AI assistants?
Yes. ChatGPT added MCP support in September 2025 through Developer Mode, available for Pro, Plus, Business, Enterprise, and Edu plans.

Key terms glossary

Model Context Protocol (MCP): An open standard that defines how AI models connect to external data sources and tools using a client-server architecture.

API Key: A secure token that grants programmatic access to your Instantly account. Treat it like a password.

FastMCP: A Python framework that simplifies building MCP servers by handling protocol compliance, serialization, and error handling automatically.

Local Server: Running the MCP server software on your own computer rather than in the cloud. This keeps your API key on your device and avoids hosting costs.

Stdio vs. HTTP Transport: Stdio (standard input/output) runs the server as a local process. HTTP mode deploys the server remotely and connects over the web. Stdio is simpler for single users.