Connect Your AI Agent to Rotifer Protocol

Your AI agent is already good at writing code — but what if it could also discover, evaluate, and install modular logic units on its own? That's exactly what connecting to the Rotifer gene ecosystem enables.

With 50+ genes already in the registry, your agent can search for a web-scraping gene, compare alternatives by fitness score, and pull the best one into your project — all within a single conversation. No browser tab, no copy-paste, no context switch.

This tutorial walks you through setting it up in under five minutes.

Prerequisites

• Node.js 20+ — the MCP server runs via npx, so a modern Node runtime is required.
• One of these AI clients:
  • Claude Desktop (macOS / Windows)
  • Cursor IDE
  • OpenClaw AI Agent Platform

That's it. No API key, no account creation — the Rotifer public registry is open.

Claude Desktop Setup

Claude Desktop loads MCP servers from a JSON config file on startup.

1. Open the config file

On macOS the path is:

~/Library/Application Support/Claude/claude_desktop_config.json

On Windows:

%APPDATA%\Claude\claude_desktop_config.json

Create the file if it doesn't exist yet.

2. Add the Rotifer server

{
  "mcpServers": {
    "rotifer": {
      "command": "npx",
      "args": ["@rotifer/mcp-server"]
    }
  }
}

If you already have other MCP servers configured, just add the "rotifer" key alongside them.

3. Restart Claude Desktop

Quit and reopen the app. You should see a hammer icon (🔨) in the chat input area indicating that MCP tools are available. Click it to confirm rotifer tools appear in the list.

Cursor Setup

Cursor reads MCP configuration from a .cursor/mcp.json file in your project root.

1. Create the config file

mkdir -p .cursor
touch .cursor/mcp.json

2. Add the Rotifer server

{
  "mcpServers": {
    "rotifer": {
      "command": "npx",
      "args": ["@rotifer/mcp-server"]
    }
  }
}

3. Reload the window

Open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and run Developer: Reload Window. Cursor will start the MCP server in the background. You can verify it's running from the MCP panel in Settings.

OpenClaw Setup

OpenClaw uses a skill-based extension model. Rotifer ships an official OpenClaw Skill.

1. Clone the skill repo

git clone https://gitlab.com/rotifer-protocol/rotifer-openclaw-skill.git \
  ~/.openclaw/skills/rotifer

2. Restart OpenClaw

The platform discovers new skills on startup. After restarting, the /rotifer slash commands become available in any conversation.

3. Try a command

/rotifer search web scraping

OpenClaw will call the Rotifer registry and return matching genes inline.

Try It Out

Once the server is connected, just talk to your agent naturally. Here are a few prompts to get started:

Search the registry

"Search for web scraping genes"

Your agent calls list_genes with the query "web scraping" and returns a list of matching genes with their names, descriptions, and fitness scores.

Find the best option

"Which search gene has the highest fitness?"

The agent calls get_leaderboard or list_genes with a sort parameter and tells you which gene ranks highest in the Arena.

Compare alternatives

"Compare gene-http-fetch and gene-browser-scraper"

The agent calls compare_genes with both gene IDs and presents a side-by-side comparison: fitness scores, input/output schemas, fidelity level (Native vs Wrapped), and download counts.

Get detailed stats

"Show me the download trend for gene-http-fetch over the last 30 days"

The agent calls get_gene_stats with a time range and summarizes the trend.

Available MCP Tools

Here's the full set of tools your agent gains access to:

Tool	What It Does
list_genes	Search and browse genes by keyword, category, or tag
get_gene	Retrieve full details for a gene: README, phenotype schema, fitness
run_gene	Execute a gene with custom input and return the output
compare_genes	Side-by-side comparison of two genes
get_gene_stats	Download and usage statistics over 7d / 30d / 90d
get_leaderboard	Global developer reputation rankings
get_developer_profile	Developer profile and contributed genes

All tools are read-only except run_gene, which executes a gene in Rotifer's sandboxed Cloud Binding.

Troubleshooting

"Tool not found" or MCP server won't start

• Verify Node.js version: node -v should print v20.x or higher.
• Make sure npx is available: npx --version.
• Try running the server directly to see errors: npx @rotifer/mcp-server.

"Connection timeout"

• The server connects to cloud.rotifer.dev. Make sure your network can reach it.
• If you're behind a corporate proxy, set HTTPS_PROXY in your environment before launching the AI client.

"No results" for a search

• The gene ecosystem is in alpha — the registry has 50+ genes but coverage is growing. Try broader terms (e.g. "http" instead of "REST API JSON parser").
• Use list_genes without a query to browse all available genes.

What's Next

You now have a live connection between your AI agent and the Rotifer gene ecosystem. Every gene in the registry — current and future — is instantly available to your agent without any plugin updates.

Try building something: ask your agent to find a gene, inspect it, and integrate it into your project. The more you use it, the more natural it becomes.

Deep Dive: See the full MCP Server Setup guide for advanced configuration and WebMCP details.

---

This article was originally published on rotifer.dev. Follow the project on GitLab or install the CLI: npm i -g @rotifer/playground.