CrewAI Integration

Give your CrewAI agents web browsing superpowers - Plasmate provides structured SOM pages at ~10x fewer tokens than raw HTML scraping.

Source: integrations/crewai/

Installation

pip install plasmate crewai crewai-tools

Quick Start

from crewai import Agent, Task, Crew
from plasmate.integrations.crewai import PlasmateWebTool

# Create the Plasmate browsing tool
browse = PlasmateWebTool()

# Create an agent with web access
researcher = Agent(
    role="Web Researcher",
    goal="Find and summarize information from the web",
    backstory="Expert at extracting key information from web pages.",
    tools=[browse],
)

# Define a task
task = Task(
    description="Research the top stories on Hacker News and summarize them.",
    expected_output="A bullet-point summary of the top 5 stories.",
    agent=researcher,
)

# Run the crew
crew = Crew(agents=[researcher], tasks=[task])
result = crew.kickoff()
print(result)

Available Tools

`PlasmateWebTool`

Fetches a URL and returns SOM text. Drop-in replacement for ScrapeWebsiteTool with dramatically fewer tokens.

from plasmate.integrations.crewai import PlasmateWebTool

tool = PlasmateWebTool()
# Agents invoke it automatically when they need web content

`PlasmateBrowseTool`

Persistent browser session with navigate, click, and type actions for multi-step workflows.

from plasmate.integrations.crewai import PlasmateBrowseTool

tool = PlasmateBrowseTool()
# Supports: navigate(url), click(index), type(index, text)

Why Plasmate for CrewAI?

	ScrapeWebsiteTool	PlasmateWebTool
Output	Raw HTML/text	Structured SOM
Tokens per page	~20,000-40,000	~2,000-4,000
Interactive elements	Lost	Indexed `[N]`
Multi-step browsing	❌	✅
Dependencies	requests + beautifulsoup	`plasmate` binary

Over a typical crew run with 5-10 page loads, you save 50,000-150,000 tokens - significant cost reduction at scale.