Quickly launch a Hyperbrowser agent and tool in a Meshagent cloud room.

​
Environment Variables

  • HYPERBROWSER_API_KEY: <app or user token>

​
Steps

  1. Install Meshagent

    pip install "meshagent[all]"

  2. Authenticate with Meshagent

  3. Start the MCP Server in a Meshagent Room

    meshagent service test --room=test --role=agent --image=meshagent/mcp-hyperbrowser:latest --env MESHAGENT_PORT=8001 --env HYPERBROWSER_API_KEY=<your-api-key> --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" --name=mcp-hyperbrowser-service
    • This command launches a Meshagent Room (test) with the Hyperbrowser MCP server running inside.
    • Replace <your-api-key> with your actual API token.
    • Rooms close if they go inactive.

  4. Join the Room as a Chatbot Agent

    meshagent chatbot join --room=test --agent-name=sample --name=sample --toolkit=mcp-hyperbrowser
    • Starts a chatbot agent in the room capable of using Hyperbrowser tools.
    • Multiple toolkits (besides mcp-hyperbrowser) can be attached to the same agent.
    • A link to the Meshagent Room will be shown in the output—use it to interact with your agent and tools.

​
Tools Available

​
Tools Summary

Tools provided by this ServerShort Description
browser_use_agentFast, cost-effective browser automation for explicit tasks using an open-source agent.
claude_computer_use_agentExecutes complex browser tasks using Anthropic’s Claude model for nuanced reasoning.
crawl_webpagesCrawl a website from a URL, following links and exploring multiple pages.
create_profileCreate a persistent Hyperbrowser browser profile.
delete_profileDelete an existing persistent Hyperbrowser profile.
extract_structured_dataExtract structured data from webpages using prompts and optional schemas.
list_profilesList all persistent browser profiles, with pagination.
openai_computer_use_agentGeneral-purpose browser automation using OpenAI models.
scrape_webpageScrape content from webpages in various formats.
search_with_bingSearch the web using Bing.

​
Tools Details

​
Tool: browser_use_agent

Employs an open-source browser automation agent for fast, efficient, and cost-effective browser tasks using a cloud browser. Requires detailed, explicit instructions for precise, high-speed automation.
Best for: rigid workflows like logins, navigation, and explicit scraping.
Parameters:

  • task (string)
  • maxSteps (integer, optional)
  • returnStepInfo (boolean, optional)
  • sessionOptions (object, optional)

​
Tool: claude_computer_use_agent

Uses Anthropic’s Claude model to run browser tasks that require sophisticated reasoning or multi-step logic in the cloud.
Best for: complex, reasoning-intensive interactions, research, or onboarding flows.
Parameters:

  • task (string)
  • maxSteps (integer, optional)
  • returnStepInfo (boolean, optional)
  • sessionOptions (object, optional)

​
Tool: crawl_webpages

Crawls a website from a given URL, traversing linked pages for systematic content collection.
Parameters:

  • followLinks (boolean)
  • outputFormat (array)
  • url (string)
  • ignoreSitemap (boolean, optional)
  • maxPages (integer, optional)
  • sessionOptions (object, optional)

​
Tool: create_profile

Creates a new persistent Hyperbrowser profile.

​
Tool: delete_profile

Deletes an existing persistent Hyperbrowser profile.
Parameters:

  • profileId (string)

​
Tool: extract_structured_data

Extracts structured data from webpages using provided schema and prompts.
Parameters:

  • prompt (string)
  • urls (array)
  • schema (string, optional)
  • sessionOptions (object, optional)

​
Tool: list_profiles

Lists persistent Hyperbrowser profiles, with optional pagination.
Parameters:

  • limit (integer, optional)
  • page (integer, optional)

​
Tool: openai_computer_use_agent

Utilizes OpenAI’s model for balanced, flexible browser automation—good for general-purpose tasks with moderate complexity.
Parameters:

  • task (string)
  • maxSteps (integer, optional)
  • returnStepInfo (boolean, optional)
  • sessionOptions (object, optional)

​
Tool: scrape_webpage

Fetches and extracts webpage content in various formats (HTML, text, screenshots, etc).
Parameters:

  • outputFormat (array)
  • url (string)
  • sessionOptions (object, optional)

​
Tool: search_with_bing

Performs searches using Bing and returns web results.
Parameters:

  • query (string)
  • numResults (integer, optional)
  • sessionOptions (object, optional)