MCP

The Autobound MCP (Model Context Protocol) server lets AI assistants like Claude, Cursor, and any MCP-compatible client query your signal data directly. Ask natural language questions about companies and contacts - the AI agent calls our API autonomously.

What is MCP?

Model Context Protocol is an open standard from Anthropic that lets AI assistants connect to external tools and data sources. Instead of copying API responses into chat, your AI assistant queries Autobound directly.

Example: Ask Claude "What hiring signals has Salesforce shown in the last 90 days?" - it calls our search API, gets structured results, and reasons about them in real time.

Two Ways to Connect

You need an Autobound API key for either path. Create one on the API Keys page, or contact [email protected].

Option 1 - Hosted Server (recommended)

Connect to https://mcp.autobound.ai/sse and pass your key as an x-api-key header.

Claude Code

claude mcp add -s user --transport sse autobound-signals https://mcp.autobound.ai/sse --header "x-api-key: YOUR_API_KEY"

-s user makes the server available in every directory. Without it, the server only exists in the directory where you ran the command. Restart Claude Code, then run /mcp to confirm the connection.

Cursor

Cursor supports remote servers natively. Add to .cursor/mcp.json:

{
  "mcpServers": {
    "autobound-signals": {
      "url": "https://mcp.autobound.ai/sse",
      "headers": {
        "x-api-key": "YOUR_API_KEY"
      }
    }
  }
}

Claude Desktop

Claude Desktop needs the mcp-remote bridge for remote servers. Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "autobound-signals": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.autobound.ai/sse",
        "--header",
        "x-api-key:${AUTOBOUND_API_KEY}"
      ],
      "env": {
        "AUTOBOUND_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

The key passes through env because Claude Desktop splits arguments on spaces. Restart Claude Desktop after saving.

Option 2 - Local Server (npm)

Prerequisites

• Node.js 18+
• An Autobound API key

Install

npm install -g @autobound-ai/mcp-server

Or run directly with npx (no install needed):

npx @autobound-ai/mcp-server

Claude Desktop / Cursor

Use the same config shape for both (claude_desktop_config.json or .cursor/mcp.json):

{
  "mcpServers": {
    "autobound-signals": {
      "command": "npx",
      "args": ["-y", "@autobound-ai/mcp-server"],
      "env": {
        "AUTOBOUND_API_KEY": "your-api-key-here"
      }
    }
  }
}

Restart the client after saving.

Other MCP Clients

Any MCP-compatible client works. Set the command to npx -y @autobound-ai/mcp-server and provide AUTOBOUND_API_KEY as an environment variable.

Environment Variables

Available Tools

The MCP server exposes 25 tools across 7 categories:

Enrichment

Search

Signals

Buyer Intent

Email Verification

Monitoring

⚠️

LinkedIn post types require a LinkedIn URL. linkedin-post-contact records need contact_linkedin_url and linkedin-post-company records need company_linkedin_url; email- or domain-only records return a 400. Batches are all-or-nothing: one invalid record rejects the whole batch (nothing is added, nothing is charged). Duplicates and rejected records are free.

Account

Example Prompts

Once connected, try asking your AI assistant:

• "Search for 10-K filing signals for Apple in the last 6 months"
• "What hiring signals has Stripe shown recently?"
• "Find companies showing buyer intent for 'data enrichment'"
• "Export intent contacts for my top 100 topics"
• "Enrich the contact [email protected] with all available signals"
• "Which companies are showing intent for AI infrastructure?"
• "Verify these 50 email addresses"
• "How many credits do I have left?"

Troubleshooting

What's Next

• Streaming responses for large result sets
• Multi-agent workflow support
• Pre-built prompt templates for common sales workflows

Need help? Email [email protected] or reach out in your Slack Connect channel.