Dendrite provides a Python SDK for building AI agents capable of interacting with websites like humans. It enables agents to authenticate, extract structured data, upload/download files, and browse the web without being blocked, offering an alternative to API-dependent automation.

How It Works

Dendrite leverages an asynchronous Python client that wraps browser automation. It uses natural language prompts to identify and interact with web elements, abstracting away complex selectors. For data extraction, it converts page content to Markdown and uses LLMs to parse structured data, with caching for generated scripts to improve efficiency and resilience against website changes.

Quick Start & Requirements

• Install: pip install dendrite && dendrite install
• Prerequisites: Requires an LLM API key (e.g., Anthropic) configured via .env or code.
• Docs: https://docs.dendrite.systems/
• Examples: https://docs.dendrite.systems/examples

Highlighted Details

• Authentication: Supports cookie-based authentication by opening a browser for user login.
• Data Extraction: Can extract specific data points or entire pages as Markdown.
• Remote Browsing: Integrates with Browserbase for scalable, parallel agent execution.
• Self-Healing Scripts: Automatically caches and attempts to reuse/repair interaction scripts.

Maintenance & Community

The project is not under active development but remains open source for learning and forking. The team is available on Discord for questions.

Licensing & Compatibility

The license is not explicitly stated in the README, but the project is presented as fully open source for forking and adaptation. Commercial use implications are not detailed.

Limitations & Caveats

The SDK is explicitly stated as no longer under active development. While functional, users should not expect new features or bug fixes from the original maintainers.