What your AI can do
with graph8.
420+ tools behind one server URL. Search the buyer graph, enrich anything, run Launches, watch competitors, book meetings. Here is what that looks like in practice.
https://be.graph8.com/mcp/ 
Ask it like this.
Copy one and paste it into any connected AI. The connection pays off on the first question.
Selling
Building
One-click Instant Connect
Pick your client. Click once.
Nine clients, one server URL. OAuth handles the rest.
Claude Desktop
Just paste the server URL, no config file needed
Setup steps
- 1 Copy the server URL
- 2 In Claude Desktop, go to Settings > MCP Servers > Add
- 3 Paste the URL
- 4 Authorize when prompted (one time)
https://be.graph8.com/mcp/
Claude.ai
Just paste the server URL, no config file needed
Setup steps
- 1 Copy the server URL
- 2 In Claude.ai, go to Customize > Connectors > Add custom connector
- 3 Paste the URL
- 4 Authorize when prompted
https://be.graph8.com/mcp/
ChatGPT
Connector availability varies by ChatGPT plan
Setup steps
- 1 Copy the server URL
- 2 In ChatGPT, go to Settings > Apps > Developer Mode
- 3 Add MCP server and paste the URL
- 4 Authorize via OAuth
https://be.graph8.com/mcp/
Gemini CLI
OAuth is auto-discovered; a browser opens to sign in once
Setup steps
- 1 Copy the config (note: uses "httpUrl", not "url")
- 2 Open ~/.gemini/settings.json (create it if it does not exist)
- 3 Paste the config
- 4 Restart Gemini CLI, then authorize in the browser when prompted
{
"mcpServers": {
"graph8": {
"httpUrl": "https://be.graph8.com/mcp/"
}
}
} Cursor
40-tool limit: use dev or gtm mode if you have other MCP servers
Setup steps
- 1 Copy the config
- 2 Create .cursor/mcp.json in your project root
- 3 Paste the config
- 4 Restart Cursor
{
"mcpServers": {
"graph8": {
"url": "https://be.graph8.com/mcp/"
}
}
} VS Code
Agent Mode required in GitHub Copilot
Setup steps
- 1 Copy the config (note: uses "servers" key, not "mcpServers")
- 2 Create .vscode/mcp.json in your project root
- 3 Paste the config
- 4 Enable Agent Mode in GitHub Copilot
{
"servers": {
"graph8": {
"url": "https://be.graph8.com/mcp/"
}
}
} Windsurf
Works in Cascade mode
Setup steps
- 1 Copy the config
- 2 Open ~/.codeium/windsurf/mcp_config.json
- 3 Paste the config
- 4 Restart Windsurf
{
"mcpServers": {
"graph8": {
"url": "https://be.graph8.com/mcp/"
}
}
} CLI
Setup steps
- 1 Run: pip install g8-mcp-server
- 2 Run: g8 login (opens browser to sign in via OAuth)
- 3 Or: g8 login --api-key (paste your API key manually)
pip install g8-mcp-server g8 login
SDK
Setup steps
- 1 Run: npm i @graph8/sdk
- 2 Generate an API key in the graph8 app
- 3 Initialize: g8.init({ apiKey: 'your_key' })
npm i @graph8/sdk
What your AI can do
420+ tools do four jobs.
Your AI reads the Buyer Graph, runs Launches, and moves Pipeline.
Find
- Search 700M+ contacts and 100M+ companies, free.
- Enrich any person or company. 1 credit.
- Preview prospect lists before you spend.
Engage
- Run Launches with 20 campaign tools.
- Draft, launch, and monitor from chat.
- Sequences built where you already work.
Close
- Book meetings. 51 scheduling tools, one search away.
- Track rivals with Radar. 42 discoverable tools.
- Walk into calls with fresh intel.
Automate
- Compose workflows. 24 tools.
- Switch orgs mid-conversation.
- 119 tools always on. 300+ on demand.
41 tools render interactive widgets inside Claude.
See it work
One connection powers four demos.
How auth works
Click once, authorize once, and you're in.
Pick your client
Choose your AI above and hit Instant Connect. One click, no config files for Claude or ChatGPT.
Approve once in your browser
A graph8 sign-in opens. Approve it one time. New here? Sign up inline, right in that window.
Tools appear in your client
Your AI can now search, enrich, and act. No API keys to generate, paste, or rotate.
Pricing, laid out straight
Here's what costs what.
No surprise spend. Your AI searches free and only spends credits on lookups and saves.
Credit pricing →Search across 700M+ contacts and 100M+ companies costs nothing.
Each person or company enrichment costs 1 credit.
Preview lists cost nothing. Saving a list spends credits.
More ways in
Prefer code? Call the same engine.
REST API
Every tool as an HTTP endpoint.
api-reference →TypeScript SDK
Typed client for your product.
npm i @graph8/sdk →Python CLI
The same server, from your terminal.
pip install g8-mcp-server →Docs
Auth, tools, and examples.
build/docs →Building agents on graph8? Read the developer MCP page.
Common questions, answered
Six questions come up every time.
Which AI clients does this work with? +
Claude Desktop, Claude.ai, ChatGPT, Gemini CLI, Cursor, VS Code, and Windsurf. Any client that speaks streamable HTTP MCP connects with the same URL.
What does it cost me? +
Search costs nothing. Person and company lookups cost 1 credit each. Prospecting previews cost nothing; saving a list spends credits.
Do I need an API key? +
No. The server uses OAuth 2.0. Click connect, approve once in your browser, done. New to graph8? Sign up inline during that same step.
How many tools does my AI actually get? +
420+ in total. 119 stay always on, and 300+ more load on demand via tool search. 41 render interactive widgets inside Claude.
I run more than one workspace. Does that work? +
Yes. Multi-org switching is built in. Tell your AI to switch orgs mid-conversation.
I'd rather call an API. +
Same engine, different door. The REST API, TypeScript SDK, and Python CLI live at graph8.com/build.
Ready when you are
Your AI is ready. Connect it.
One click to connect. One prompt to prove it.
OAuth 2.0 · 420+ tools · No API keys