Genudo MCP Client

Genudo is the platform to build AI agents for any communication or sequence-based channel — pipeline-aware agents with integration capabilities and company-knowledge access, all managed from one platform, one inbox, and one analytics dashboard. This lightweight bridge connects Claude and other MCP clients to your Genudo workspace so you can build, run, and improve those agents directly from chat.

📖 Which install is right for you? See CONNECT.md — every way to connect Genudo to Claude (plugin, connector, desktop extension, other MCP clients, remote OAuth), weighted by value with copy-paste install commands.

Overview

Genudo MCP Client bridges the communication gap between Claude Code's stdio-based MCP implementation and Genudo's Streamable HTTP MCP server. This allows you to leverage Genudo's powerful business automation tools directly within your Claude Code conversations.

Features

Access Genudo's business automation tools directly from Claude Code. The server exposes 29 tools across five areas:

• 📊 Analytics — account summary, messaging volume stats, AI performance & cost (get_account_summary, get_messaging_stats, get_ai_performance)
• 🔎 Discover — list pipelines, stages, actions, variables, contacts, opportunities, messages, and knowledge tables; search knowledge; read follow-up configs; valid build options (list_pipelines, list_pipeline_stages, list_actions, list_variables, list_contacts, list_opportunities, list_messages, list_knowledge_tables, search_knowledge_table, get_stage_followup, get_pipeline_options)
• 🚀 Build — a guided step-by-step builder plus direct creation of pipelines, stages, variables, webhook actions, follow-up sequences, and knowledge tables (start_pipeline_journey, create_pipeline, create_stage, create_variable, create_action, create_followup, create_knowledge_table, upsert_knowledge_points)
• 🔧 Update — tune live pipelines, stages, actions, variables, opportunities (bulk), and follow-ups (update_pipeline, update_stage, update_action, update_variable, update_opportunities, update_followup)
• 🗑️ Delete — remove knowledge rows by stable id (delete_knowledge_points)

Three workflows to start with

1. Audit my AI business operations — Claude pulls your account, messaging, and AI-cost stats and surfaces the top fixes.
2. Analyze my agent performance — spot expensive agents, low-completion stages, and actions that aren't triggering.
3. Create or improve a pipeline — e.g. "Create a Messenger sales agent for a summer camp; collect name, phone, branch, child age; escalate on special-needs questions."

Guided instruction editing (built in)

The connector doesn't just expose tools — it teaches the agent how to write good agent instructions, entirely client-side (no extra setup):

• Guide tools — get_instruction_guides (authoring rules + persona/global + stage templates + a token-aware QA checklist) and get_editing_playbook (a safe load → edit → diff → confirm → push workflow). The agent pulls these on demand.
• Prompts (slash-commands in any MCP client) — edit_instructions, build_pipeline, audit_pipeline.
• Safety — before changing a live agent, the agent reads current text from list_pipelines / list_pipeline_stages, shows you a before/after diff with expected impact, and pushes with update_pipeline / update_stage only after you confirm.

Quick Start

One command — no clone, no manual config:

claude mcp add --env GENUDO_TOKEN=YOUR_TOKEN --transport stdio genudo -- npx -y genudo-mcp-client

Replace YOUR_TOKEN with your Genudo token (API Keys & Tokens → Create token with the mcp:use scope — see Get Your Token). Restart Claude Code and start using Genudo tools!

Prerequisites

• Node.js 18.0.0 or higher
• A Genudo account with API access
• Your Genudo token

Installation

Recommended: one command

claude mcp add --env GENUDO_TOKEN=YOUR_TOKEN --transport stdio genudo -- npx -y genudo-mcp-client

npx fetches and runs the published package — no clone, no local path. Get your token from API Keys & Tokens (see below), then restart Claude Code.

Alternative: from source

For local development or contributions:

git clone https://github.com/genudo-ai/genudo_mcp.git
cd genudo_mcp
npm install

Then configure Claude Code manually (see Configuration section below).

Configuration

Step 1: Get Your Token

1. Log in to your Genudo account
2. In the sidebar, under Developer, open API Keys & Tokens
3. Click Create token
4. Name it (e.g. claude-mcp)
5. Under Scopes, scroll the list and check mcp:use (Access MCP server — Streamable HTTP + JSON-RPC tool calls)
6. Pick an Expiry (30 days, 90 days, 1 year, or No Expiry)
7. Click Create token and copy the token — it's shown only once; Genudo keeps only a hashed copy

Step 2: Connect your MCP client

Recommended — Claude Code, one command:

claude mcp add --env GENUDO_TOKEN=YOUR_TOKEN --transport stdio genudo -- npx -y genudo-mcp-client

Other clients (Codex, Cursor, Windsurf, Claude Desktop): same package, add this block to the client's MCP config:

{
  "mcpServers": {
    "genudo": {
      "command": "npx",
      "args": ["-y", "genudo-mcp-client"],
      "env": {
        "GENUDO_TOKEN": "your_token_here"
      }
    }
  }
}

{
  "mcpServers": {
    "genudo": {
      "command": "node",
      "args": ["/absolute/path/to/genudo_mcp/index.js"],
      "env": { "GENUDO_TOKEN": "your_token_here" }
    }
  }
}

Step 3: Restart Claude Code

Exit your current Claude Code session and start a new one:

# Exit current session
exit

# Start new session
claude

Usage

Once configured, Claude Code will automatically have access to your Genudo tools. You can use them in your conversations:

Example:

User: "Show me my account summary"
Claude: [Uses get_account_summary tool to fetch your data]

User: "Create a new sales pipeline using GPT-4"
Claude: [Uses get_pipeline_options and create_pipeline tools]

Testing the Connection

You can test the bridge manually:

# Set your token
export GENUDO_TOKEN="your_token_here"

# Run the bridge
node index.js

# Send a test message (paste this and press Enter):
{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}

# You should receive a response with server info

Troubleshooting

"GENUDO_TOKEN environment variable is required"

• Make sure you've set the token in your Claude Code config
• Verify the token is valid in your Genudo account

"MCP handshake failed" / no response

• Check that GENUDO_BASE_URL is correct (defaults to https://api.genudo.ai) — the bridge POSTs to <base>/mcp
• Verify your server is running and accessible
• Check firewall/network settings

"HTTP 401: Unauthorized"

• Your token is invalid or expired
• Generate a new token in your Genudo account settings

Connection Issues with Self-Signed Certificates

• The bridge is configured to accept self-signed certificates for local development
• For production, use proper SSL certificates

Environment Variables

Development

Project Structure

genudo-mcp-client/
├── index.js           # Main bridge script
├── package.json       # Node.js project config
├── .env.example       # Environment variables template
└── README.md          # This file

How It Works

1. Bridge performs the MCP handshake against <base>/mcp (Streamable HTTP)
2. Reads JSON-RPC requests from stdin (from Claude Code)
3. Forwards each one as an HTTP POST to that same endpoint
4. Returns responses via stdout back to Claude Code

Since v2.5.0 there is no SSE step: Streamable HTTP is a single endpoint that takes every message as a POST. If you don't need the bridge's local guide tools, prompts or response slimming, you can skip it entirely and point your client at the server: claude mcp add --transport http genudo https://api.genudo.ai/mcp

Claude Code → (stdin) → Bridge → (HTTPS) → Genudo MCP Server
                ↑                              ↓
                └─────── (stdout) ←────────────┘

Security Notes

• Never commit your token to version control
• Store tokens securely (environment variables, secrets manager)
• Rotate tokens regularly
• Use HTTPS in production
• Limit token scopes to minimum required access

Privacy Policy