The cheat code · AI agent tool

Plug Gorilla into your agent.

Gorilla is open-source on the Model Context Protocol. Drop the server into Claude, Cursor, Codex, or any MCP-compatible agent. Your agent runs the demand hunt for you — high-intent posts across Reddit, X, YouTube, TikTok and LinkedIn — and drafts the replies.

ADD MCP CONFIG ~/.claude.json
{
  "mcpServers": {
    "gorilla": {
      "command": "npx",
      "args": ["-y", "github:opusforge/gorilla-mcp"],
      "env": { "GORILLA_API_KEY": "grla_..." }
    }
  }
}
ASK YOUR AGENT
Find high-intent conversations
for <your idea>.
Draft 2 short replies.

What the MCP server does

The server exposes Gorilla's lead-discovery pipeline as tools your agent can call. Your agent decides when to expand a theme, when to scan, and when to draft a reply — same engine as the web app, driven from chat.

  • Theme expansion — turns a one-line idea into the keywords, pain phrases, and competitor names buyers actually use.
  • Multi-platform search — runs a single hunt across Reddit, X, YouTube, TikTok and LinkedIn in parallel.
  • Intent scoring — ranks every post 0–100 on signal, fit, and recency.
  • Reply drafting — produces short, sub-aware drafts. You read, edit, and post yourself. No bots, no auto-reply.

When this beats the web app

The MCP server is for founders who already live in an agent. Skip the dashboard, keep the work where you are — your codebase, your terminal, your chat. Pipe the results into a doc, a CSV, a Notion table, whatever your agent already knows how to write to.

What you need

  • A Gorilla API key — sign up free, copy it from your dashboard.
  • An MCP-compatible agent. Claude Desktop, Claude Code, Cursor, and Codex all work. Any client that speaks MCP can run the server.
  • npx on the box you're running the agent from. The first run pulls the package straight from GitHub.

YOUR IDEA NEXT?

DROP IT IN. SEE WHO WANTS IT.

▶ RUN YOUR IDEA