What is MCP?
The Model Context Protocol (MCP) is an open standard for connecting AI clients to external tools and data. Instead of copying context into a chat window, the model calls PG:AI tools directly when your prompt needs account data, search, or content generation.MCP is supported by Claude (Connectors), Cursor, and other MCP-compatible clients. PG:AI hosts the production server at
https://mcp.getpg.ai.What you get
22 tools
Research, search, accounts, contacts, jobs, content generation, and optional agent sessions
2 resources
Passive lists of tracked companies and territories
Hosted by PG:AI
No self-hosting required for production - connect with OAuth or an API key
How it works
company tool (and others if needed). You do not need to name tools in your prompt - the client picks them from the tool list.
Authentication
API keys start with
pgai_live_. Send them as x-api-key: pgai_live_… or Authorization: Bearer pgai_live_….
Tool groups (at a glance)
See Tools reference for every parameter. See Examples for copy-paste prompts.
Supported clients
Claude
Claude.ai Pro / Team / Enterprise - Custom connector with OAuth (setup guide)
Cursor
Project or global MCP config with an API key
Other MCP clients
Streamable HTTP at
https://mcp.getpg.ai/mcp or SSE at https://mcp.getpg.ai/sseGet started
1
Choose authentication
Claude users: OAuth via a Custom connector. Developers: create an API key in Settings → API Keys.
2
Connect your client
Follow the Quickstart for Claude step-by-step instructions.
3
Test with a simple prompt
Try “List my tracked accounts” or “Get the profile overview for [company name]”.
Quickstart - connect Claude
Step-by-step connector setup with screenshots
