Integrate ZeroClaw With Bright Data for a Web-Aware AI Assistant

In this article, you will learn:

• What ZeroClaw is and what it brings to the table.
• Why adding web access elevates your AI assistant’s capabilities.
• How Bright Data enables ZeroClaw integration via web scraping, search, discovery, and browser automation tools.
• How to connect Bright Data to ZeroClaw using MCP.
• How to provide your ZeroClaw AI assistant with Bright Data knowledge through dedicated Agent Skills.
• What the ZeroClaw + Bright Data integration supports, with a complete example.

Let’s dive in!

What Is ZeroClaw?

ZeroClaw is a lightweight, Rust-based AI agent runtime engineered to run entirely on your own machine.

It acts as a personal AI assistant that connects to multiple LLM providers, interacts across dozens of channels, and executes real-world tasks through tools like the shell, browser, and APIs. What makes it stand out to competitors is its privacy-first and fully local approach.

ZeroClaw has been gaining rapid traction in the open-source community, reaching over 30k GitHub stars in less than a couple of months.

The core features it provides are:

• Single binary runtime: Runs as a compact Rust binary with fast startup and minimal resource usage.
• Provider-agnostic: Works with 50+ AI providers, including OpenAI, Anthropic, Ollama, and more.
• Multi-channel support: Connects to 20+ platforms like Discord, Telegram, email, and CLI.
• Local-first privacy: Everything runs on your machine, so you get full data ownership (no cloud, no telemetry).
• Tool execution engine: Interacts with shell, browser, HTTP APIs, hardware, and MCP servers.
• Security and sandboxing: Built-in safeguards, permissions, and workspace isolation for safe automation.
• Automation and SOP (Standard Operating Procedure) engine: Supports workflows via cron jobs, webhooks, and event-driven procedures.
• Hardware integration: Works with devices like Raspberry Pi, Arduino, and ESP32.
• Extensible architecture: Easily add new providers, tools, channels, and custom integrations.

Find out more in the official docs.

Why Give Your ZeroClaw AI Assistant Access to the Web

Even the most advanced LLMs powering ZeroClaw share a fundamental limitation: their knowledge is frozen in time. Since they rely on static training data, they cannot naturally keep up with real-world changes.

In a fast-moving environment, that is a serious drawback. Without access to up-to-date information, an AI assistant may produce outdated answers, hallucinate facts, or take incorrect actions.

To remain reliable, your ZeroClaw AI agents need access to the live web. This is why ZeroClaw comes with built-in tools for web scraping and search. Still, these are not always production-ready and may face blocks or inconsistent results.

This is where Bright Data comes in!

Bright Data stands out with a network of over 400 million residential IPs across 195 countries, offering 99.99% uptime and a 99.95% success rate. It turns your AI agent into a setup with scalable, production-grade web data access.

By integrating Bright Data’s AI-ready infrastructure, your ZeroClaw assistant can go beyond its training limits by:

• Performing real-time research: Access fresh information through live search queries on Google, Bing, Yandex, and more
• Validating information accuracy: Cross-check outputs against reliable online sources
• Collecting live data: Extract up-to-date content such as prices, reviews, and market insights
• Supporting many other use cases…

How to Extend ZeroClaw With Bright Data: 2 Approaches

Bright Data supports ZeroClaw through two complementary integrations:

• Bright Data Web MCP: The official MCP server exposing 70+ tools to interact with Bright Data’s APIs and services.
• Bright Data Skills: A collection of Agent Skills that teach ZeroClaw how to effectively use Bright Data solutions.

Important: These two approaches are not alternatives. Instead, they work best together. In particular, the Bright Data skills include a dedicated skill to help your AI agents get the most out of Web MCP tools.

Bright Data Web MCP

The Bright Data Web MCP provides 70+ tools for automated web data collection, structured extraction, and browser automation.

Even on the free tier (5k requests/month), you get access to core tools such as:

Then, Pro mode unlocks the full capabilities of Web MCP. This includes advanced tools for structured extraction from platforms like Amazon, LinkedIn, Yahoo Finance, YouTube, TikTok, Zillow, and Google Maps. Also, it exposes browser automation tools.

Bright Data Skills

The Bright Data skills extend ZeroClaw’s knowledge via:

Common Steps

Before seeing how to integrate Bright Data into ZeroClaw, take care of the common setup steps.

Prerequisites

To follow this tutorial, make sure you have:

• A UNIX-based operating system (Linux or macOS).
• Node.js 20+ installed locally (LTS version recommended).
• Rust installed locally.
• An API key from one of the LLM providers supported by ZeroClaw (here, we will use an OpenAI API key).
• A Bright Data account with an API key configured. To generate a Bright Data API key, follow the official guide.

Step #1: Install ZeroClaw

In your terminal, run the command below to retrieve and execute the ZeroClaw installation script:

curl -fsSL https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

You will output similar to the following:

Now restart your shell. The zeroclaw command should now be available globally. Verify that with:

zeroclaw --version

The result will look something like:

zeroclaw 0.7.3

Well done! ZeroClaw has been installed on your machine.

Step #2: Complete the Setup

For a guided setup, run the following command:

zeroclaw onboard

The ZeroClaw onboarding process will start:

You will be guided through a 9-step wizard to configure ZeroClaw.

Start by selecting your target AI provider. In this case, choose OpenAI and paste your API key:

Then fetch the latest models and select one (in this case, gpt-5-mini).

You will be prompted to connect a channel. This is optional for this tutorial. Next, make sure to enable “Sovereign Mode” for full control over the AI assistant’s tools:

Finally, personalize your AI assistant by providing its name, timezone, and preferred tone:

ZeroClaw will now be configured, and the required daemon will be launched automatically.

To verify everything is working, start the CLI chat with:

zeroclaw agent

Try sending a message like “Hey!” If everything is set up correctly, you should see a response like:

Awesome! ZeroClaw is now properly running on your machine.

Connecting Your ZeroClaw AI Assistant to the Bright Data MCP

In this guided section, you will learn how to configure a local instance of the Bright Data Web MCP in ZeroClaw.

Prerequisites

To follow along more easily, it is recommended that you have:

• Familiarity with how the MCP standard works.
• Knowledge of the tools provided by the Bright Data Web MCP.

Also note that the prerequisites listed in the “Common Steps” section still apply here.

Step #1: Set Up Bright Data’s Web MCP

Before configuring the Web MCP in your ZeroClaw setup, first verify that the MCP server runs correctly on your machine.

Start by logging in to your Bright Data account. For a quick setup, consider the instructions in the “MCP” section of the control panel:

Otherwise, for more guidance, follow the steps below.

First, install the Web MCP by globally adding the @brightdata/mcp package:

npm install -g @brightdata/mcp

Next, check that the MCP server starts with:

API_TOKEN="<YOUR_BRIGHT_DATA_API>" npx -y @brightdata/mcp

Replace the <YOUR_BRIGHT_DATA_API> placeholder with your actual Bright Data API key. The above command sets the required API_TOKEN environment variable and launches the Web MCP server locally.

If everything works correctly, you should see logs similar to:

On the first run, the @brightdata/mcp package automatically creates two zones in your Bright Data account:

• mcp_unlocker: A zone for Web Unlocker.
• mcp_browser: A zone for Browser API.

These zones power the 70+ tools available in Web MCP. You can also configure custom zone names, as described in the official repository.

To check that the default zones were created, reach the “Proxies & Scraping Infrastructure” page in the control panel. You should see both zones appearing in the “My Zones” table:

Now, remember that on the Web MCP free tier, only a limited set of tools is available.

To unlock all 70+ tools, you must enable Pro mode by setting the PRO_MODE="true" env:

API_TOKEN="<YOUR_BRIGHT_DATA_API>" PRO_MODE="true" npx -y @brightdata/mcp

Note: Pro mode is not included in the free tier and incurs additional charges.

Excellent! You just made sure that the Bright Data Web MCP works on your system. Now, you will configure ZeroClaw to connect to it.

Step #2: Add the Web MCP Configuration

ZeroClaw supports MCP configuration through a special setting in the ~/.zeroclaw/config.toml file. That represents the global configuration file used by ZeroClaw.

To enable the Bright Data Web MCP, make sure your config.toml includes the following:

# ~/.zeroclaw/config.toml

[mcp]
enabled = true
deferred_loading = true
servers = [{ name = "bright-data", command = "npx", transport = "stdio", args = ["@brightdata/mcp"], env = { API_TOKEN = "<YOUR_BRIGHT_DATA_API_KEY", PRO_MODE = "true" }, headers = {} }]

In detail, enabled must be set to true to activate MCP support (in recent versions, it is disabled by default). Also, the servers array in [mcp] must contain a valid Bright Data MCP configuration as shown above.

That configuration mirrors the npx command seen earlier, with credentials passed via environment variables:

• API_TOKEN (required): Your Bright Data API key.
• PRO_MODE (optional): Enable advanced features. Remove it or set it to "false" if you do not intend to rely on Pro mode.

Note: MCP servers in ZeroClaw are lazy-loaded, meaning they are not started at daemon startup.
Instead, ZeroClaw connects to the local Web MCP instance only when you launch your AI assistant, triggering the configured npx command at that time.

Cool! Let’s verify that.

Step #3: Check the Connection

Start a new ZeroClaw CLI session with:

zeroclaw agent

In the startup logs, you should now observe:

Since the Web MCP is the only one configured, and the logs refer specifically to it, we can state that ZeroClaw successfully connected to a local Web MCP instance.

To clear up any doubts, try a prompt like this:

Which Bright Data MCP tools do you have access to?

You should receive a response listing all tools exposed by the Web MCP. With Pro Mode enabled, the output will include the full set of available tools, as below:

Congratulations! The Web MCP integration is working correctly. Later in this piece, you will see the Web MCP in action together with Bright Data Skills.

Giving Your ZeroClaw AI Assistant Knowledge of Bright Data’s Infrastructure via Skills

Here, you will see how to add Bright Data skills to your local ZeroClaw setup.

Prerequisites

To go through this chapter, make sure you have:

• A basic understanding of the Agent Skills standard.
• Familiarity with the Bright Data skills.
• Git installed locally.

In addition to the “Common Steps” prerequisites, you will also need:

• A Web Unlocker API zone configured in your Bright Data account.
• The jq library installed locally.

To install jq (a lightweight JSON processor similar to sed) on Debian-based operating systems, run:

sudo apt-get install curl jq

Or, on macOS, run:

brew install curl jq

For a quick setup of a Web Unlocker API zone, check out the “Create Your First Unlocker API” guide. Alternatively, continue with the step below.

Step #1: Create a Web Unlocker API Zone

Begin by logging in to your Bright Data account. Go to the “Proxies & Scraping Infrastructure” page, and take a look at the “My Zones” table:

If a Web Unlocker zone already exists, such as web_unlocker in this case, move to the next step.

Otherwise, create one by scrolling to the “Unblocker API” card and pressing “Create zone”:

Choose a clear name for your zone and complete the setup wizard until the zone becomes active.
Fantastic!

Step #2: Finalize the Skills Setup

To connect to the Bright Data infrastructure, the official skills require these two environment variables:

• BRIGHTDATA_API_KEY: Required to authenticate requests to the Bright Data APIs.
• BRIGHTDATA_UNLOCKER_ZONE: Contains your Web Unlocker API zone name, enabling web scraping and search capabilities (as Web Unlocker API can also act as a SERP API).

Set the required envs in your system with:

export BRIGHTDATA_API_KEY="<YOUR_BRIGHT_DATA_API_KEY>"
export BRIGHTDATA_UNLOCKER_ZONE="<YOUR_BRIGHT_DATA_WEB_UNLOCKER_API_ZONE_NAME>"

Replace the placeholders with the actual values. Once set, the Bright Data skills are ready to be employed by ZeroClaw!

Step #3: Add the Bright Data Skills in ZeroClaw

At the time of writing, Bright Data skills require shell script execution to work. So first, set allow_scripts to true in the [skills] section of your ZeroClaw configuration file:

# ~/.zeroclaw/config.toml

[skills]
# ...
allow_scripts = true
# ...

To install new skills in ZeroClaw, you need to place them inside the ~/.zeroclaw/workdpsace/skills directory. Start by cloning the official Bright Data skills repository:

git clone https://github.com/brightdata/skills

Inside the repository, you will find all available skills under the inner skills/ folder:

skills/
  ├── ...
  └── skills/
        ├── brd-browser-debug/
        ├── bright-data-best-practices/
        ├── bright-data-mcp/
        ├── brightdata-cli/
        ├── competitive-intel/
        ├── data-feeds/
        ├── design-mirror/
        ├── python-sdk-best-practices/
        ├── scrape/
        └── scraper-builder/

Copy all the skills/skills/ folders into your ZeroClaw workspace:

cp -r skills/skills/* ~/.zeroclaw/workspace/skills/

This is it! The Bright Data skills are now successfully added to ZeroClaw.

Step #4: Verify That the Skills Are Available

To confirm that ZeroClaw recognizes the Bright Data official skills, run:

zeroclaw skills list

The output should display the list of Bright Data skills you previously added:

Mission accomplished! In the next section, you will experience the ZeroClaw AI assistant enhanced with the Bright Data Web MCP and Agent Skills.

ZeroClaw + Bright Data in Action

Assume you are a business owner who regularly wants to receive a message from your ZeroClaw AI assistant summarizing reviews of your Google Maps listing. Achieve that with a prompt like this:

Retrieve the Google Maps reviews for the "NY Pizza Suprema" business in New York. Focus on reviews that were posted in the last 2 days. Also scrape the main business information from Google Maps (e.g., name, category, location, rating, number of reviews, etc.).

Then, aggregate everything into a concise report that includes 
- Updated average review score + udpated review count 
- A sentiment analysis of the most recent reviews 

For the sentiment analysis:
- Highlight common positive and negative themes 
- Include real quotes from customers 
- Emphasize what customers appreciated most and what they suggested could be improved

Important: The task is not something ZeroClaw can reliably achieve on its own. That is because it requires production-grade tools for web search (to locate the business on Maps), scraping (to extract business data), and structured extraction from Google Maps.

Tip: To ensure ZeroClaw uses Bright Data for this workflow (instead of unstable built-in tools), you can explicitly instruct it in the prompt. Or, optionally, send a message asking to store this preference in the agent’s memory so it persists across sessions.

To verify that the prompt works, test it by starting the chat with the agent and sending the message:

zeroclaw agent

Or send the prompt directly with:

zeroclaw agent --message "<YOUR_PROMPT>"

This is what you should see:

The expected workflow is:

1. It performs a web query via the Web MCP search_engine tool (powered by Bright Data SERP API).
2. It extracts the Google Maps URL from structured search results.
3. It passes the URL to the web_data_google_maps_reviews tool (backed by the Bright Data Google Reviews API).
4. It retrieves business metadata (name, rating, category, reviews) by scraping the Maps business page using scrape_as_markdown (via Web Unlocker API).
5. It aggregates everything into a structured report.

Note: Before running any tool, the AI agent will ask for permission. Grant it with “A” to allow the agent to remember this permission for future tool usage. Otherwise, for the current run only, approve execution with “Y”.

The result is a contextual report containing review scores and sentiment analysis based on real Google Maps data:

This includes real-world insights derived directly from live Google Maps info and review data:

Et voilà! The only step left is to enable scheduling, connect a channel, and you will start receiving automated reports with all the insights you need. This is a simple example showcasing the power of ZeroClaw + Bright Data, though many other use cases are possible as well!

Conclusion

In this tutorial, you explored what ZeroClaw is and the value it provides. In particular, you learned why and how to enhance it by integrating Bright Data through the Web MCP and its official skills.

This setup elevates your ZeroClaw AI assistant by enabling direct web access. It unlocks capabilities such as web search, content discovery, structured data extraction, and automated interactions with websites.

If you want to go further, explore the full suite of AI-ready services available in Bright Data’s ecosystem.

Sign up for a free Bright Data account today and start integrating our powerful web data tools!