Skip to content

Model Context Protocol

Give your AI agent a lead-finding tool.

Leadiver speaks MCP, so Claude Code, Claude Desktop, Cursor, Codex, and any other MCP client can drive it directly. Your agent runs the searches and reads the threads through the same stealth browser you see in the app.

Quick install for Claude Code

Using another client? 
claude mcp add --transport http leadiver http://127.0.0.1:41789/mcp

AI agents such as Claude Code, Claude Desktop, Cursor, Codex and VS Code connect to Leadiver over the Model Context Protocol, and Leadiver reaches Reddit, YouTube, LinkedIn and Facebook with its stealth browser.

Claude Code claude mcp add … Claude Desktop Edit claude_desktop_config.json Cursor Add to ~/.cursor/mcp.json Codex Add to ~/.codex/config.toml Any MCP client Add to any .mcp.json
Leads DB

Two ways to put it to work

Hand the whole job to Leadiver, or borrow its stealth browser and let your agent do the thinking.

Mode 1

Your agent drives Leadiver

Ask your agent, in plain language, to find leads. It calls the high-level tools, and Leadiver does the work end to end with its own judge, then hands back scored leads and drafts.

  • find_leads

    Start a lead search on a topic. Returns fast with a session id while the run works in the background.

  • find_more

    Keep searching an existing session for more leads on top of the ones it already found.

  • get_run_status

    Check a run: running, blocked, or done, with the lead count and whether it hit a wall.

  • list_leads

    Read the qualified leads back: score, url, the passage, and why each one is a fit.

  • draft_reply

    Draft a brand-voiced reply for a lead, with a "likely to land" score and a short critique.

  • search

    Full-text search across a brand's saved sessions and every lead inside them.

Mode 2

Your agent borrows the hands

The differentiator. Leadiver hands back the raw thread and comment text its stealth browser pulled, and your agent supplies its own judgment. A web-research source that survives the anti-bot walls a plain fetch tool cannot.

  • browse_search

    The candidate threads and videos the human-behavior browser finds for a query. No judging, just what a person would see.

  • read_thread

    The raw post and comments behind a url, after settling the page and clearing consent and login walls, exactly as a human reads them.

  • fetch_passages

    A full traverse pass that returns raw, unscored passages so your agent can qualify them with its own judgment.

  • add_candidate

    Hand a url back and the stealth browser scrapes it, scores it against the brand, and stores it as a lead in the session.

Plus read-only resources for your brands and saved sessions, and reusable prompts for qualifying leads and drafting outreach.

Works with the agent you already use

MCP is a shared standard, so one connection works everywhere.

Connect in one command

With Leadiver running, point Claude Code at its local endpoint: 

claude mcp add --transport http leadiver http://127.0.0.1:41789/mcp

Using another client, or want it to run with the app closed? Each setup guide above has copy-paste steps, including the stdio bridge that starts the engine on demand.

Questions

What is MCP, and why does Leadiver support it?
MCP, the Model Context Protocol, is the open standard AI agents use to call external tools. Because Leadiver already runs a long-lived local service that owns the browser and the run engine, it exposes those same capabilities to any MCP client. Your agent gets a lead-finding tool that browses like a human, and you did not have to build an integration.
Does my data leave my machine?
The MCP connection is local. Your agent talks to Leadiver over a localhost endpoint that is bound to your own computer, and it drives the same browser and sessions you see in the app. The server only accepts loopback connections, so a web page cannot reach it.
Do I need the desktop app open?
You can run it either way. When the app is open, your agent reuses the running service over HTTP. When it is closed, a small stdio bridge starts the engine on demand, so an agent can use it with no window on screen. The browser itself only launches on the first search.
What happens if a run hits a CAPTCHA in the background?
The status reports that the run is blocked, with the signals about what it hit, so your agent can tell you to open the app and clear the wall. The run picks up from where it stopped once the challenge is solved.

Put it in front of your agent

Install Leadiver, connect your MCP client, and let your agent find warm leads on Reddit, YouTube, LinkedIn and Facebook, then draft the first reply.

Get started