pywebtask 0.15.0

LLM-powered web automation library with autonomous agents and natural language selectors

webtask

LLM-powered web automation with autonomous agents and natural language selectors.

📚 Documentation | 🐍 PyPI | 📊 Benchmarks

---

Quick Start

pip install pywebtask
playwright install chromium
export GEMINI_API_KEY="your-key"  # or OPENAI_API_KEY

Autonomous mode - Give it a task, let the agent figure out the steps:

from webtask import Webtask
from webtask.integrations.llm import GeminiLLM

wt = Webtask(headless=False)
llm = GeminiLLM.create(model="gemini-2.5-flash")
agent = await wt.create_agent(llm=llm)

result = await agent.execute("search for cats and click the first result")

Direct control - Natural language selectors, you control the flow:

await agent.navigate("https://example.com")
search = await agent.select("search box")
await search.fill("cats")
button = await agent.select("search button")
await button.click()

No CSS selectors. No XPath. Just describe what you want.

---

Features

• Multimodal by default - Sees screenshots with bounding boxes + DOM text
• Extensible - Pluggable LLM and browser interfaces
• Batteries included - OpenAI, Gemini LLMs and Playwright browser provided
• Isolated sessions - Separate cookies and storage per agent

---

Documentation

Full Documentation

• Getting Started
• Examples
• API Reference

---

Benchmarks

webtask-benchmarks - Evaluation on Mind2Web and other benchmarks

---

License

MIT