How to Make OpenClaw Actually Useful: A Practical Guide

You installed OpenClaw. Maybe you ran through the onboarding, connected it to Telegram or Discord, asked it a few questions. It answered them well enough. Then you tried to do something real, like look up a restaurant, send an email, or pull recent news, and hit a wall. The agent is smart but blind. It can reason about anything you paste into it, but it can’t go get information on its own.

This is where most people stall. They installed the most popular open-source AI agent (247K+ GitHub stars, for good reason), poked around for an afternoon, and then moved on because it felt like a chatbot instead of an assistant.

The fix is tools. OpenClaw becomes a different product once you connect it to external data sources and actions. Here’s how to get there.

What OpenClaw Does Out of the Box

The defaults are solid for basic work. You get:

  • Chat across messaging apps. Telegram, Discord, Signal, Slack. One agent, multiple channels.
  • File management. Read, write, and organize files on your local machine.
  • Shell commands. Run scripts, manage processes, interact with your system.
  • Brave Search. Basic web search ships natively, though the free tier has tight rate limits and the index is thinner than Google’s.

For casual Q&A and local automation, this covers a lot of ground. The problem shows up when you ask it to do something that requires external services.

What It Can’t Do Without Tools

Without connecting additional tools, OpenClaw has no way to:

  • Search Google. Brave is the default, but Google’s index is deeper and more current for most queries.
  • Send or read email. No SMTP, no inbox access, nothing.
  • Generate images. It can describe images but not create them.
  • Pull financial data. No stock quotes, no SEC filings, no FRED economic data.
  • Use Google Maps. No place lookups, no reviews, no directions.
  • Search news. Google News, not just general web results.
  • Get YouTube transcripts. Useful for summarizing videos or extracting information.
  • Track trends. Google Trends data for market research and content planning.

Each of these requires either a dedicated MCP server or a skill that provides the connection. You could set them up one by one, each with its own API key and configuration. Or you could connect a tool marketplace that bundles them behind a single key.

How Skills and MCP Tools Fill the Gap

OpenClaw supports two ways to add capabilities:

Skills are installed with clawhub install <skill-name>. They teach the agent when and how to use specific tools. A skill wraps an MCP connection with context that helps the agent make better decisions about when to invoke which tool.

MCP servers are configured directly in ~/.openclaw/openclaw.json under the mcp.servers key. This is the lower-level approach. You point OpenClaw at an MCP endpoint and it discovers the available tools at runtime.

Skills are the better starting point. They handle the MCP configuration for you and add routing hints that improve tool selection.

Top 5 Things to Set Up First

If you only do five things after installing OpenClaw, do these:

1. Web search via Google. Replace or supplement Brave with Google Search. The results are better for almost every query type, and you won’t hit free-tier rate limits during normal use.

2. Email. Give your agent the ability to send emails. This unlocks a huge category of useful workflows: research-then-email, digest-then-send, draft-then-deliver.

3. Image generation. Your agent can create images from text descriptions. Useful for content creation, mockups, social media graphics, and creative brainstorming.

4. News search. Google News gives you recent, source-attributed articles on any topic. Pair this with email for daily digests.

5. YouTube transcripts. Ask your agent to summarize a video, extract key points, or compare what two people said about the same topic. Transcripts make video content searchable.

Setup

AgentPatch bundles all five of these (and 50+ more tools) behind a single MCP connection. One API key, one skill install, and your agent has access to the full marketplace.

The AgentPatch CLI is designed for AI agents to use via shell access. Install it, and your agent can discover and invoke any tool on the marketplace.

Install (zero dependencies, Python 3.10+):

pip install agentpatch

Set your API key:

export AGENTPATCH_API_KEY=your_api_key

Example commands your agent will use:

ap search "web search"
ap run google-search --input '{"query": "test"}'

Get your API key from the AgentPatch dashboard.

Install the AgentPatch skill from ClawHub — it teaches OpenClaw when to use AgentPatch and how to use the CLI:

clawhub install agentpatch

MCP Server (Alternative)

If you prefer raw MCP tool access instead of the skill, add AgentPatch to ~/.openclaw/openclaw.json:

{
  "mcp": {
    "servers": {
      "agentpatch": {
        "transport": "streamable-http",
        "url": "https://agentpatch.ai/mcp",
        "headers": {
          "Authorization": "Bearer YOUR_API_KEY"
        }
      }
    }
  }
}

Replace YOUR_API_KEY with your actual key from the AgentPatch dashboard. Restart OpenClaw and it discovers all AgentPatch tools automatically.

Restart your gateway after making config changes.

Wrapping Up

OpenClaw is a capable agent framework that ships with a limited toolset. The value comes from what you connect to it. Start with search, email, and image generation. Add news and transcripts. Once those are working, you’ll find yourself reaching for the agent instead of opening browser tabs, and that’s when it starts being useful. Browse the full tool list at agentpatch.ai.

← Back to blog